Ophthalmology science最新文献_第6页

Predicting Choroidal Nevus Transformation to Melanoma Using Machine Learning 利用机器学习预测脉络膜痣向黑色素瘤的转化

IF 3.2 Q1 OPHTHALMOLOGY

Ophthalmology science

Pub Date : 2024-07-20 DOI: 10.1016/j.xops.2024.100584

Prashant D. Tailor MD , Piotr K. Kopinski MD, PhD , Haley S. D’Souza MD , David A. Leske MS , Timothy W. Olsen MD , Carol L. Shields MD , Jerry A. Shields MD , Lauren A. Dalvin MD

Purpose

To develop and validate machine learning (ML) models to predict choroidal nevus transformation to melanoma based on multimodal imaging at initial presentation.

Design

Retrospective multicenter study.

Participants

Patients diagnosed with choroidal nevus on the Ocular Oncology Service at Wills Eye Hospital (2007–2017) or Mayo Clinic Rochester (2015–2023).

Methods

Multimodal imaging was obtained, including fundus photography, fundus autofluorescence, spectral domain OCT, and B-scan ultrasonography. Machine learning models were created (XGBoost, LGBM, Random Forest, Extra Tree) and optimized for area under receiver operating characteristic curve (AUROC). The Wills Eye Hospital cohort was used for training and testing (80% training–20% testing) with fivefold cross validation. The Mayo Clinic cohort provided external validation. Model performance was characterized by AUROC and area under precision–recall curve (AUPRC). Models were interrogated using SHapley Additive exPlanations (SHAP) to identify the features most predictive of conversion from nevus to melanoma. Differences in AUROC and AUPRC between models were tested using 10 000 bootstrap samples with replacement and results.

Main Outcome Measures

Area under receiver operating curve and AUPRC for each ML model.

Results

There were 2870 nevi included in the study, with conversion to melanoma confirmed in 128 cases. Simple AI Nevus Transformation System (SAINTS; XGBoost) was the top-performing model in the test cohort [pooled AUROC 0.864 (95% confidence interval (CI): 0.864–0.865), pooled AUPRC 0.244 (95% CI: 0.243–0.246)] and in the external validation cohort [pooled AUROC 0.931 (95% CI: 0.930–0.931), pooled AUPRC 0.533 (95% CI: 0.531–0.535)]. Other models also had good discriminative performance: LGBM (test set pooled AUROC 0.831, validation set pooled AUROC 0.815), Random Forest (test set pooled AUROC 0.812, validation set pooled AUROC 0.866), and Extra Tree (test set pooled AUROC 0.826, validation set pooled AUROC 0.915). A model including only nevi with at least 5 years of follow-up demonstrated the best performance in AUPRC (test: pooled 0.592 (95% CI: 0.590–0.594); validation: pooled 0.656 [95% CI: 0.655–0.657]). The top 5 features in SAINTS by SHAP values were: tumor thickness, largest tumor basal diameter, tumor shape, distance to optic nerve, and subretinal fluid extent.

Conclusions

We demonstrate accuracy and generalizability of a ML model for predicting choroidal nevus transformation to melanoma based on multimodal imaging.

Financial Disclosures

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

目的开发并验证机器学习（ML）模型，根据初次发病时的多模态成像预测脉络膜痣向黑色素瘤的转化。方法获取多模态成像，包括眼底照相、眼底自动荧光、光谱域 OCT 和 B-scan 超声波检查。创建了机器学习模型（XGBoost、LGBM、随机森林、Extra Tree），并对接收者工作特征曲线下面积（AUROC）进行了优化。威尔斯眼科医院队列用于训练和测试（80%训练-20%测试），并进行五倍交叉验证。梅奥诊所队列提供了外部验证。模型性能以 AUROC 和精确度-召回曲线下面积 (AUPRC) 为特征。使用 SHapley Additive exPlanations (SHAP) 对模型进行了检验，以确定哪些特征最能预测痣向黑色素瘤的转化。使用 10,000 个带替换的 Bootstrap 样本对模型之间的 AUROC 和 AUPRC 差异进行了测试，并得出了结果。简单人工智能痣转化系统（SAINTS; XGBoost）是测试队列中表现最好的模型[汇总 AUROC 0.864（95% 置信区间 (CI)：0.864-0.865），集合 AUPRC 0.244（95% 置信区间：0.243-0.246）]，在外部验证队列中[集合 AUROC 0.931（95% 置信区间：0.930-0.931），集合 AUPRC 0.533（95% 置信区间：0.531-0.535）]表现最好。其他模型也具有良好的判别性能：LGBM（测试集集合 AUROC 0.831，验证集集合 AUROC 0.815）、随机森林（测试集集合 AUROC 0.812，验证集集合 AUROC 0.866）和额外树（测试集集合 AUROC 0.826，验证集集合 AUROC 0.915）。仅包括随访至少 5 年的痣的模型在 AUPRC 方面表现最佳（测试：集合 AUROC 为 0.592（95% CI：0.590-0.594）；验证：集合 AUROC 为 0.656 [95% CI：0.655-0.657]）。根据 SHAP 值，SAINTS 中的前 5 个特征是：肿瘤厚度、最大肿瘤基底直径、肿瘤形状、与视神经的距离以及视网膜下积液范围。

{"title":"Predicting Choroidal Nevus Transformation to Melanoma Using Machine Learning","authors":"Prashant D. Tailor MD , Piotr K. Kopinski MD, PhD , Haley S. D’Souza MD , David A. Leske MS , Timothy W. Olsen MD , Carol L. Shields MD , Jerry A. Shields MD , Lauren A. Dalvin MD","doi":"10.1016/j.xops.2024.100584","DOIUrl":"10.1016/j.xops.2024.100584","url":null,"abstract":"<div><h3>Purpose</h3><p>To develop and validate machine learning (ML) models to predict choroidal nevus transformation to melanoma based on multimodal imaging at initial presentation.</p></div><div><h3>Design</h3><p>Retrospective multicenter study.</p></div><div><h3>Participants</h3><p>Patients diagnosed with choroidal nevus on the Ocular Oncology Service at Wills Eye Hospital (2007–2017) or Mayo Clinic Rochester (2015–2023).</p></div><div><h3>Methods</h3><p>Multimodal imaging was obtained, including fundus photography, fundus autofluorescence, spectral domain OCT, and B-scan ultrasonography. Machine learning models were created (XGBoost, LGBM, Random Forest, Extra Tree) and optimized for area under receiver operating characteristic curve (AUROC). The Wills Eye Hospital cohort was used for training and testing (80% training–20% testing) with fivefold cross validation. The Mayo Clinic cohort provided external validation. Model performance was characterized by AUROC and area under precision–recall curve (AUPRC). Models were interrogated using SHapley Additive exPlanations (SHAP) to identify the features most predictive of conversion from nevus to melanoma. Differences in AUROC and AUPRC between models were tested using 10 000 bootstrap samples with replacement and results.</p></div><div><h3>Main Outcome Measures</h3><p>Area under receiver operating curve and AUPRC for each ML model.</p></div><div><h3>Results</h3><p>There were 2870 nevi included in the study, with conversion to melanoma confirmed in 128 cases. Simple AI Nevus Transformation System (SAINTS; XGBoost) was the top-performing model in the test cohort [pooled AUROC 0.864 (95% confidence interval (CI): 0.864–0.865), pooled AUPRC 0.244 (95% CI: 0.243–0.246)] and in the external validation cohort [pooled AUROC 0.931 (95% CI: 0.930–0.931), pooled AUPRC 0.533 (95% CI: 0.531–0.535)]. Other models also had good discriminative performance: LGBM (test set pooled AUROC 0.831, validation set pooled AUROC 0.815), Random Forest (test set pooled AUROC 0.812, validation set pooled AUROC 0.866), and Extra Tree (test set pooled AUROC 0.826, validation set pooled AUROC 0.915). A model including only nevi with at least 5 years of follow-up demonstrated the best performance in AUPRC (test: pooled 0.592 (95% CI: 0.590–0.594); validation: pooled 0.656 [95% CI: 0.655–0.657]). The top 5 features in SAINTS by SHAP values were: tumor thickness, largest tumor basal diameter, tumor shape, distance to optic nerve, and subretinal fluid extent.</p></div><div><h3>Conclusions</h3><p>We demonstrate accuracy and generalizability of a ML model for predicting choroidal nevus transformation to melanoma based on multimodal imaging.</p></div><div><h3>Financial Disclosures</h3><p>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</p></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"5 1","pages":"Article 100584"},"PeriodicalIF":3.2,"publicationDate":"2024-07-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S2666914524001209/pdfft?md5=611e184a6e4ec4c46ad3dd1688c15182&pid=1-s2.0-S2666914524001209-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141844736","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Reference Database for a Novel Binocular Visual Function Perimeter: A Randomized Clinical Trial 新型双眼视觉功能周界参考数据库：随机临床试验

IF 3.2 Q1 OPHTHALMOLOGY

Ophthalmology science

Pub Date : 2024-07-20 DOI: 10.1016/j.xops.2024.100583

Vincent Michael Patella OD , Nevin W. El-Nimri OD, PhD , John G. Flanagan PhD , Mary K. Durbin PhD , Timothy Bossie OD , Derek Y. Ho MD, PhD , Mayra Tafreshi MBA , Michael A. Chaglasian OD , David Kasanoff OD , Satoshi Inoue MSc , Sasan Moghimi MD , Takashi Nishida MD, PhD , Murray Fingeret OD , Robert N. Weinreb MD

Purpose

To construct a comprehensive reference database (RDB) for a novel binocular automated perimeter.

Design

A four-site prospective randomized clinical trial.

Subjects and Controls

Three hundred fifty-six healthy subjects without ocular conditions that might affect visual function were categorized into 7 age groups.

Methods

Subjects underwent comprehensive ocular examination of both eyes before enrollment. Using the TEMPO/IMOvifa automated perimeter (Topcon Healthcare/CREWT Medical Systems), each subject completed 4 binocular threshold visual field (VF) tests during a single visit: First, practice 24-2 and 10-2 tests were obtained from both eyes. Next, study 24-2 and 10-2 tests were obtained from both eyes. Test order of each sequence was randomized, and the tests were conducted under standard automated perimetry testing conditions: Goldmann stimulus size III, 3183 cd/m² maximum stimulus intensity, and background intensity of 10 cd/m², using AIZE-Rapid test strategy. Standard VF reliability indices were assessed. For each subject, 24-2 and 10-2 test results from 1 randomly selected eye were analyzed.

Main Outcome Measures

Perimetric threshold sensitivity and reference limits for each test analysis parameter.

Results

The ages of the study cohort were widely distributed, with a mean age (standard deviation [SD]) of 52.3 (18.5) years. Sex assignment was 44.0% male and 56.0% female. The majority of subjects self-identified as White (67.4%), followed by Black or African American (13.5%) and Asian (8.7%), with 14.6% self-identified as Hispanic or Latino ethnicity. Mean sensitivity (SD) was 29.1 (1.3) decibels (dB) for the 24-2 and 32.4 (1.0) dB for the 10-2 test. For the 24-2 and 10-2, mean sensitivity (SD) age-related changes averaged −0.06 (0.01) dB and −0.05 (0.01) dB per year, respectively. The normal range of pointwise threshold sensitivity increased with eccentricity and showed asymmetry around the mean, particularly notable in the 24-2 test. Mean (SD) binocular test duration was 3.18 (0.38) minutes (1 minute 35 seconds per eye) for the 24-2 test and 3.58 (0.43) minutes (1 minute 47 seconds per eye) for the 10-2 test.

Conclusions

An RDB for the TEMPO/IMOvifa perimeter was established, highlighting the significance of considering both age and stimulus eccentricity in interpreting threshold VF test results.

Financial Disclosure(s)

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

目的为新型双眼自动周界构建一个综合参考数据库（RDB）。方法受试者在入组前接受双眼综合眼科检查。使用 TEMPO/IMOvifa 自动周波仪（Topcon Healthcare/CREWT Medical Systems），每位受试者在一次就诊中完成 4 次双眼阈值视野（VF）测试：首先，从双眼获得练习 24-2 和 10-2 测试结果。然后，进行双眼 24-2 和 10-2 研究测试。每个序列的测试顺序都是随机的，测试在标准自动视力测试条件下进行：戈德曼刺激大小为 III，最大刺激强度为 3183 cd/m2，背景强度为 10 cd/m2，采用 AIZE-Rapid 测试策略。对标准 VF 可靠性指数进行了评估。对每个受试者随机选取的 1 只眼睛的 24-2 和 10-2 测试结果进行分析。结果研究对象的年龄分布广泛，平均年龄（标准差 [SD]）为 52.3 (18.5)岁。男性占 44.0%，女性占 56.0%。大多数受试者自认为是白人（67.4%），其次是黑人或非裔美国人（13.5%）和亚裔（8.7%），14.6%自认为是西班牙裔或拉丁裔。24-2 测试的平均灵敏度（标清）为 29.1 (1.3) 分贝，10-2 测试的平均灵敏度（标清）为 32.4 (1.0) 分贝。在 24-2 和 10-2 测试中，与年龄有关的平均灵敏度（标度）变化分别为每年-0.06 (0.01) 分贝和-0.05 (0.01) 分贝。点阈灵敏度的正常范围随偏心率的增加而增大，并在平均值周围表现出不对称，这在 24-2 测试中尤为明显。24-2 测试的平均（标清）双眼测试时间为 3.18 (0.38) 分钟（每只眼 1 分 35 秒），10-2 测试的平均（标清）双眼测试时间为 3.58 (0.43) 分钟（每只眼 1 分 47 秒）。结论建立了TEMPO/IMOvifa周界的RDB，强调了在解释阈值VF测试结果时同时考虑年龄和刺激偏心率的重要性。

{"title":"Reference Database for a Novel Binocular Visual Function Perimeter: A Randomized Clinical Trial","authors":"Vincent Michael Patella OD , Nevin W. El-Nimri OD, PhD , John G. Flanagan PhD , Mary K. Durbin PhD , Timothy Bossie OD , Derek Y. Ho MD, PhD , Mayra Tafreshi MBA , Michael A. Chaglasian OD , David Kasanoff OD , Satoshi Inoue MSc , Sasan Moghimi MD , Takashi Nishida MD, PhD , Murray Fingeret OD , Robert N. Weinreb MD","doi":"10.1016/j.xops.2024.100583","DOIUrl":"10.1016/j.xops.2024.100583","url":null,"abstract":"<div><h3>Purpose</h3><p>To construct a comprehensive reference database (RDB) for a novel binocular automated perimeter.</p></div><div><h3>Design</h3><p>A four-site prospective randomized clinical trial.</p></div><div><h3>Subjects and Controls</h3><p>Three hundred fifty-six healthy subjects without ocular conditions that might affect visual function were categorized into 7 age groups.</p></div><div><h3>Methods</h3><p>Subjects underwent comprehensive ocular examination of both eyes before enrollment. Using the TEMPO/IMOvifa automated perimeter (Topcon Healthcare/CREWT Medical Systems), each subject completed 4 binocular threshold visual field (VF) tests during a single visit: First, practice 24-2 and 10-2 tests were obtained from both eyes. Next, study 24-2 and 10-2 tests were obtained from both eyes. Test order of each sequence was randomized, and the tests were conducted under standard automated perimetry testing conditions: Goldmann stimulus size III, 3183 cd/m<sup>2</sup> maximum stimulus intensity, and background intensity of 10 cd/m<sup>2</sup>, using AIZE-Rapid test strategy. Standard VF reliability indices were assessed. For each subject, 24-2 and 10-2 test results from 1 randomly selected eye were analyzed.</p></div><div><h3>Main Outcome Measures</h3><p>Perimetric threshold sensitivity and reference limits for each test analysis parameter.</p></div><div><h3>Results</h3><p>The ages of the study cohort were widely distributed, with a mean age (standard deviation [SD]) of 52.3 (18.5) years. Sex assignment was 44.0% male and 56.0% female. The majority of subjects self-identified as White (67.4%), followed by Black or African American (13.5%) and Asian (8.7%), with 14.6% self-identified as Hispanic or Latino ethnicity. Mean sensitivity (SD) was 29.1 (1.3) decibels (dB) for the 24-2 and 32.4 (1.0) dB for the 10-2 test. For the 24-2 and 10-2, mean sensitivity (SD) age-related changes averaged −0.06 (0.01) dB and −0.05 (0.01) dB per year, respectively. The normal range of pointwise threshold sensitivity increased with eccentricity and showed asymmetry around the mean, particularly notable in the 24-2 test. Mean (SD) binocular test duration was 3.18 (0.38) minutes (1 minute 35 seconds per eye) for the 24-2 test and 3.58 (0.43) minutes (1 minute 47 seconds per eye) for the 10-2 test.</p></div><div><h3>Conclusions</h3><p>An RDB for the TEMPO/IMOvifa perimeter was established, highlighting the significance of considering both age and stimulus eccentricity in interpreting threshold VF test results.</p></div><div><h3>Financial Disclosure(s)</h3><p>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</p></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"4 6","pages":"Article 100583"},"PeriodicalIF":3.2,"publicationDate":"2024-07-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S2666914524001192/pdfft?md5=ec0b7377fc17660ad95c7721e62a1e84&pid=1-s2.0-S2666914524001192-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141838949","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Censoring the Floor Effect in Long-Term Stargardt Disease Microperimetry Data Produces a Faster Rate of Decline 在长期斯塔加特病显微视力数据中删去底线效应，可加快视力下降速度

IF 3.2 Q1 OPHTHALMOLOGY

Ophthalmology science

Pub Date : 2024-07-20 DOI: 10.1016/j.xops.2024.100581

Jason Charng PhD , Jennifer A. Thompson PhD , Rachael C. Heath Jeffery MD , Amy Kalantary MD , Tina M. Lamey PhD , Terri L. McLaren BSc , Fred K. Chen PhD

Purpose

To evaluate progression rate estimation in long-term Stargardt disease microperimetry data by accounting for floor effect.

Design

Cohort study.

Subjects

Thirty-seven subjects (23 females, 14 males) with biallelic ABCA4 pathogenic or likely pathogenic variants and more than >2 years of longitudinal microperimetry data.

Methods

Cross-sectional and longitudinal microperimetry data (Grid A: 18° diameter, Grid B: 6° diameter; Macular Integrity Assessment microperimeter, dynamic range 0–36 decibels [dB]) was extracted from patients with biallelic mutation in the adenosine triphosphate-binding cassette subfamily A member 4 (ABCA4) gene. For each eye, mean sensitivity (MS) and responding point sensitivity (RPS) rates were extracted. Floor censored sensitivity (FCS) progression rate, which accounts for the floor effect at each locus by terminating calculation when scotoma was observed in 2 consecutive visits, was also calculated. In a subset of eyes with ≥1 scotomatous locus at baseline (Grid A), sensitivity progression of loci around the scotoma (edge of scotoma sensitivity [ESS]) was examined against other progression parameters. Paired t test compared progression rate parameters across the same eyes.

Main Outcome Measures

Microperimetry grid parameters at baseline and progression rates.

Results

A total of 37 subjects with biallelic ABCA4 mutations and >2 years of longitudinal microperimetry data were included in the study. In Grid A, at baseline, the average MS and RPS were 16.5 ± 7.9 and 19.1 ± 5.7 dB, respectively. Similar MS (18.4 ± 7.6 dB) and RPS (20.0 ± 5.5 dB) values were found at baseline for Grid B. In Grid A, overall, MS, RPS, and FCS progression rates were −0.57 ± 1.05, −0.74 ± 1.24, and −1.26 ± 1.65 (all dB/year), respectively. Floor censored sensitivity progression rate was significantly greater than the MS or RPS progression rates. Similar findings were observed in Grid B (MS −1.22 ± 1.42, RPS −1.44 ± 1.44, FCS −2.16 ± 2.24, all dB/year), with paired t test again demonstrated that FCS had a significantly faster rate of decline than MS or RPS. In patients with progression data in both grids, MS, RPS, and FCS progression rates were significantly faster in the smaller Grid B. In 24 eyes with scotoma at baseline, fastest rate of decline was ESS combined with FCS compared with other progression parameters.

Conclusions

Incorporation of FCS can reduce confound of floor effect in perimetry analysis and can in turn detect a faster rate of decline.

Financial Disclosure(s)

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

设计队列研究对象37名受试者（23名女性，14名男性）患有双叶ABCA4致病变异或可能致病变异，并有超过>2年的纵向微观视力数据。方法从三磷酸腺苷结合盒A亚家族成员4（ABCA4）基因双倍拷贝突变患者中提取横向和纵向微观视力数据（A栅：18°直径，B栅：6°直径；黄斑完整性评估微压计，动态范围0-36分贝[dB]）。提取了每只眼睛的平均灵敏度 (MS) 和反应点灵敏度 (RPS) 率。此外，还计算了底限删减灵敏度（FCS）进展率，该进展率考虑了每个位点的底限效应，即在连续两次观察中观察到黑影时终止计算。在基线（网格 A）上有≥1 个焦斑的眼球子集中，对照其他进展参数检查焦斑周围的灵敏度进展（焦斑边缘灵敏度 [ESS]）。主要结果测量基线时的微观视力栅参数和进展率。结果研究共纳入 37 名双侧 ABCA4 突变的受试者和 >2年的纵向微观视力数据。在网格 A 中，基线平均 MS 和 RPS 分别为 16.5 ± 7.9 和 19.1 ± 5.7 dB。在网格 A 中，总体 MS、RPS 和 FCS 的进展率分别为 -0.57±1.05、-0.74±1.24 和 -1.26±1.65（均为 dB/年）。地板删减灵敏度进展率明显高于 MS 或 RPS 进展率。在 B 网格中也观察到类似的结果（MS -1.22 ± 1.42，RPS -1.44 ± 1.44，FCS -2.16 ± 2.24，均为 dB/年），配对 t 检验再次表明 FCS 的下降速度明显快于 MS 或 RPS。在基线存在视网膜朦胧的 24 只眼睛中，与其他视力下降参数相比，ESS 结合 FCS 的视力下降速度最快。结论在近视分析中纳入 FCS 可以减少底线效应的干扰，进而检测出更快的视力下降速度。

{"title":"Censoring the Floor Effect in Long-Term Stargardt Disease Microperimetry Data Produces a Faster Rate of Decline","authors":"Jason Charng PhD , Jennifer A. Thompson PhD , Rachael C. Heath Jeffery MD , Amy Kalantary MD , Tina M. Lamey PhD , Terri L. McLaren BSc , Fred K. Chen PhD","doi":"10.1016/j.xops.2024.100581","DOIUrl":"10.1016/j.xops.2024.100581","url":null,"abstract":"<div><h3>Purpose</h3><p>To evaluate progression rate estimation in long-term Stargardt disease microperimetry data by accounting for floor effect.</p></div><div><h3>Design</h3><p>Cohort study.</p></div><div><h3>Subjects</h3><p>Thirty-seven subjects (23 females, 14 males) with biallelic ABCA4 pathogenic or likely pathogenic variants and more than >2 years of longitudinal microperimetry data.</p></div><div><h3>Methods</h3><p>Cross-sectional and longitudinal microperimetry data (Grid A: 18° diameter, Grid B: 6° diameter; Macular Integrity Assessment microperimeter, dynamic range 0–36 decibels [dB]) was extracted from patients with biallelic mutation in the adenosine triphosphate-binding cassette subfamily A member 4 (<em>ABCA4</em>) gene. For each eye, mean sensitivity (MS) and responding point sensitivity (RPS) rates were extracted. Floor censored sensitivity (FCS) progression rate, which accounts for the floor effect at each locus by terminating calculation when scotoma was observed in 2 consecutive visits, was also calculated. In a subset of eyes with ≥1 scotomatous locus at baseline (Grid A), sensitivity progression of loci around the scotoma (edge of scotoma sensitivity [ESS]) was examined against other progression parameters. Paired <em>t</em> test compared progression rate parameters across the same eyes.</p></div><div><h3>Main Outcome Measures</h3><p>Microperimetry grid parameters at baseline and progression rates.</p></div><div><h3>Results</h3><p>A total of 37 subjects with biallelic <em>ABCA4</em> mutations and >2 years of longitudinal microperimetry data were included in the study. In Grid A, at baseline, the average MS and RPS were 16.5 ± 7.9 and 19.1 ± 5.7 dB, respectively. Similar MS (18.4 ± 7.6 dB) and RPS (20.0 ± 5.5 dB) values were found at baseline for Grid B. In Grid A, overall, MS, RPS, and FCS progression rates were −0.57 ± 1.05, −0.74 ± 1.24, and −1.26 ± 1.65 (all dB/year), respectively. Floor censored sensitivity progression rate was significantly greater than the MS or RPS progression rates. Similar findings were observed in Grid B (MS −1.22 ± 1.42, RPS −1.44 ± 1.44, FCS −2.16 ± 2.24, all dB/year), with paired <em>t</em> test again demonstrated that FCS had a significantly faster rate of decline than MS or RPS. In patients with progression data in both grids, MS, RPS, and FCS progression rates were significantly faster in the smaller Grid B. In 24 eyes with scotoma at baseline, fastest rate of decline was ESS combined with FCS compared with other progression parameters.</p></div><div><h3>Conclusions</h3><p>Incorporation of FCS can reduce confound of floor effect in perimetry analysis and can in turn detect a faster rate of decline.</p></div><div><h3>Financial Disclosure(s)</h3><p>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</p></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"4 6","pages":"Article 100581"},"PeriodicalIF":3.2,"publicationDate":"2024-07-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S2666914524001179/pdfft?md5=5a10ead90090ebe64a026d61cb8212f2&pid=1-s2.0-S2666914524001179-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141840548","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Quantitative Volumetric Analysis of Retinal Ischemia with an Oxygen Diffusion Model and OCT Angiography 利用氧扩散模型和光学相干断层扫描血管造影术对视网膜缺血进行体积定量分析

IF 3.2 Q1 OPHTHALMOLOGY

Ophthalmology science

Pub Date : 2024-07-19 DOI: 10.1016/j.xops.2024.100579

Pengxiao Zang PhD , Tristan T. Hormel PhD , Thomas S. Hwang MD , Yali Jia PhD

Purpose

Retinal ischemia is a major feature of diabetic retinopathy (DR). Traditional nonperfused areas measured by OCT angiography (OCTA) measure blood supply but not ischemia. We propose a novel 3-dimensional (3D) quantitative method to derive ischemia measurements from OCTA data.

Design

Cross-sectional study.

Participants

We acquired 223 macular OCTA volumes from 33 healthy eyes, 33 diabetic eyes without retinopathy, 7 eyes with nonreferable DR, 17 eyes with referable but nonvision-threatening DR, and 133 eyes with vision-threatening DR.

Methods

Each eye was scanned using a spectral-domain OCTA system (Avanti RTVue-XR, Visionix/Optovue, Inc) with 1.6-mm scan depth in a 3 × 3-mm region (640 × 304 × 304 voxels) centered on the fovea. For each scanned OCTA volume, a custom algorithm removed flow projection artifacts. We then enhanced, binarized, and skeletonized the vasculature in each OCTA volume and generated a 3D oxygen tension map using a zero-order kinetics oxygen diffusion model. Each volume was scaled to the average retina thickness in healthy controls after foveal registration and flattening of the Bruch's membrane. Finally, we extracted 3D ischemia maps by comparison with a reference map established from scans of healthy eyes using the same processing. To assess the ability of the ischemia maps to grade DR severity, we constructed receiver operating characteristic curves for diagnosing diabetes, referable DR, and vision-threatening DR.

Main Outcome Measures

Spearman correlation coefficient and area under receiver operating characteristic curve (AUC) were used to quantify the ability of the ischemia maps to DR.

Results

The ischemia maps showed that the ischemic tissues were at or near pathologically nonperfused areas, but not the normally nonvascular tissue, such as the foveal avascular zone. We found multiple novel metrics, including inferred 3D-oxygen tension, ischemia index, and ischemic volume ratio, were strongly correlated with DR severity. The AUCs of ischemia index measured were 0.94 for diabetes, 0.89 for DR, 0.88 for referable DR, and 0.85 for vision-threatening DR.

Conclusions

A quantitative method to infer 3D oxygen tension and ischemia using OCTA in diabetic eyes can identify ischemic tissue that are more specific to pathologic changes in DR.

Financial Disclosures

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

目的视网膜缺血是糖尿病视网膜病变（DR）的一个主要特征。传统的 OCT 血管造影（OCTA）测量的非灌注区可测量血液供应，但不能测量缺血。我们从 33 只健康眼、33 只无视网膜病变的糖尿病眼、7 只不可转诊的 DR 眼、17 只可转诊但不威胁视力的 DR 眼和 133 只威胁视力的 DR 眼中获取了 223 个黄斑 OCTA 容量。方法使用光谱域 OCTA 系统（Avanti RTVue-XR，Visionix/Optovue, Inc）扫描每只眼睛，扫描深度为 1.6 毫米，扫描区域为 3 × 3 毫米（640 × 304 × 304 像素），以眼窝为中心。对于每个扫描的 OCTA 容积，我们都采用定制算法去除血流投影伪影。然后，我们对每个 OCTA 容积中的血管进行了增强、二值化和骨架化处理，并使用零阶动力学氧扩散模型生成了三维氧张力图。在进行眼窝配准和布氏膜平整后，将每个体积按健康对照组视网膜平均厚度进行缩放。最后，我们通过与使用相同处理方法从健康眼睛扫描中建立的参考图进行比较，提取出三维缺血图。为了评估缺血图对DR严重程度进行分级的能力，我们构建了接收者操作特征曲线，用于诊断糖尿病、可转诊的DR和危及视力的DR。结果缺血图显示缺血组织位于或靠近病理上无灌注的区域，但不包括正常的无血管组织，如眼窝无血管区。我们发现推断的三维氧张力、缺血指数和缺血体积比等多个新指标与 DR 的严重程度密切相关。结论在糖尿病眼中使用 OCTA 定量推断三维氧张力和缺血的方法可以识别缺血组织，这些缺血组织对 DR 的病理变化更具特异性。

{"title":"Quantitative Volumetric Analysis of Retinal Ischemia with an Oxygen Diffusion Model and OCT Angiography","authors":"Pengxiao Zang PhD , Tristan T. Hormel PhD , Thomas S. Hwang MD , Yali Jia PhD","doi":"10.1016/j.xops.2024.100579","DOIUrl":"10.1016/j.xops.2024.100579","url":null,"abstract":"<div><h3>Purpose</h3><p>Retinal ischemia is a major feature of diabetic retinopathy (DR). Traditional nonperfused areas measured by OCT angiography (OCTA) measure blood supply but not ischemia. We propose a novel 3-dimensional (3D) quantitative method to derive ischemia measurements from OCTA data.</p></div><div><h3>Design</h3><p>Cross-sectional study.</p></div><div><h3>Participants</h3><p>We acquired 223 macular OCTA volumes from 33 healthy eyes, 33 diabetic eyes without retinopathy, 7 eyes with nonreferable DR, 17 eyes with referable but nonvision-threatening DR, and 133 eyes with vision-threatening DR.</p></div><div><h3>Methods</h3><p>Each eye was scanned using a spectral-domain OCTA system (Avanti RTVue-XR, Visionix/Optovue, Inc) with 1.6-mm scan depth in a 3 × 3-mm region (640 × 304 × 304 voxels) centered on the fovea. For each scanned OCTA volume, a custom algorithm removed flow projection artifacts. We then enhanced, binarized, and skeletonized the vasculature in each OCTA volume and generated a 3D oxygen tension map using a zero-order kinetics oxygen diffusion model. Each volume was scaled to the average retina thickness in healthy controls after foveal registration and flattening of the Bruch's membrane. Finally, we extracted 3D ischemia maps by comparison with a reference map established from scans of healthy eyes using the same processing. To assess the ability of the ischemia maps to grade DR severity, we constructed receiver operating characteristic curves for diagnosing diabetes, referable DR, and vision-threatening DR.</p></div><div><h3>Main Outcome Measures</h3><p>Spearman correlation coefficient and area under receiver operating characteristic curve (AUC) were used to quantify the ability of the ischemia maps to DR.</p></div><div><h3>Results</h3><p>The ischemia maps showed that the ischemic tissues were at or near pathologically nonperfused areas, but not the normally nonvascular tissue, such as the foveal avascular zone. We found multiple novel metrics, including inferred 3D-oxygen tension, ischemia index, and ischemic volume ratio, were strongly correlated with DR severity. The AUCs of ischemia index measured were 0.94 for diabetes, 0.89 for DR, 0.88 for referable DR, and 0.85 for vision-threatening DR.</p></div><div><h3>Conclusions</h3><p>A quantitative method to infer 3D oxygen tension and ischemia using OCTA in diabetic eyes can identify ischemic tissue that are more specific to pathologic changes in DR.</p></div><div><h3>Financial Disclosures</h3><p>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</p></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"4 6","pages":"Article 100579"},"PeriodicalIF":3.2,"publicationDate":"2024-07-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S2666914524001155/pdfft?md5=b7ecfcf304d2f9f069a370d571d1ba6c&pid=1-s2.0-S2666914524001155-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141845803","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

The Association between the Pulsatile Choroidal Volume Change and Ocular Rigidity 脉动脉络膜体积变化与眼球僵直之间的关联

IF 3.2 Q1 OPHTHALMOLOGY

Ophthalmology science

Pub Date : 2024-07-18 DOI: 10.1016/j.xops.2024.100576

Diane N. Sayah OD, PhD , Denise Descovich MD , Santiago Costantino PhD , Mark R. Lesk MD, MSc

Purpose

To assess the relationship between the pulsatile choroidal volume change (ΔV) and ocular rigidity (OR), an important biomechanical property of the eye.

Design

This is a prospective cross-sectional study.

Subjects

Two hundred seventeen participants (235 eyes) were included in this study. Of those, 18 eyes (18 participants) had exudative retinal disease, and 217 eyes (199 participants) had open-angle glaucoma (39.2%), suspect discs (12.4%), ocular hypertension (14.3%), or healthy eyes (34.1%).

Methods

Pulsatile choroidal volume change was measured using dynamic OCT, which detects the change in choroidal thickness during the cardiac cycle. Ocular rigidity was measured using an invasive procedure as well as using a validated optical method. Correlations between ΔV and OR were assessed in subjects with healthy eyes, eyes with glaucoma, or eyes with exudative retinal disease.

Main Outcome Measures

Ocular rigidity and pulsatile ocular volume change.

Results

In 18 eyes where OR was obtained invasively and ΔV was obtained noninvasively, a significant correlation was found between ΔV and OR (r_s = −0.664, P = 0.003). Similarly, a strong inverse correlation was found between the noninvasive measurements of both ΔV and OR (r_s = −0.748, P < 0.001) in a large cohort and maintained its significance across diagnostic groups (a more compliant eye is associated with greater ΔV). No correlation was found between ΔV and age, blood pressure, intraocular pressure, axial length, or diagnosis (P ≥ 0.05). Mean ΔV was 7.3 ± 3.4 μL for all groups combined with a range of 3.0 to 20.8 μL.

Conclusions

These results suggest an association between the biomechanics of the corneoscleral shell and pulsatile ocular blood flow, which may indicate that a more rigid eye exerts more resistance to pulsatile choroidal expansion. This highlights the dynamic nature of both blood flow and biomechanics in the eye, as well as how they may interact, leading to a greater understanding of the pathophysiology of ocular disease.

Financial Disclosures

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

目的评估脉动脉络膜体积变化（ΔV）与眼球刚度（OR）之间的关系，眼球刚度是眼球的重要生物力学特性。其中，18 只眼睛（18 名参与者）患有渗出性视网膜疾病，217 只眼睛（199 名参与者）患有开角型青光眼（39.2%）、可疑椎间盘（12.4%）、眼压过高（14.3%）或健康眼睛（34.1%）。方法使用动态 OCT 测量脉动脉络膜体积变化，该技术可检测心动周期中脉络膜厚度的变化。眼球僵硬度采用侵入性程序和有效光学方法进行测量。在健康眼、青光眼眼或渗出性视网膜疾病眼的受试者中评估了ΔV 和 OR 之间的相关性。结果在 18 只用有创方法测量 OR 而用无创方法测量ΔV 的眼睛中，发现ΔV 和 OR 之间存在显著相关性（rs = -0.664，P = 0.003）。同样，在一个大型队列中，ΔV 和 OR 的无创测量值之间也发现了很强的反相关性（rs = -0.748，P = 0.001），并且在不同诊断组别中都保持着显著性（顺应性更强的眼睛与更大的ΔV 相关）。ΔV与年龄、血压、眼压、眼轴长度或诊断之间没有相关性（P≥0.05）。结论：这些结果表明角巩膜壳的生物力学与眼部搏动性血流之间存在关联，这可能表明眼球越僵硬，对搏动性脉络膜扩张的阻力越大。这凸显了眼部血流和生物力学的动态性质，以及它们之间的相互作用，从而加深了对眼部疾病病理生理学的理解。

{"title":"The Association between the Pulsatile Choroidal Volume Change and Ocular Rigidity","authors":"Diane N. Sayah OD, PhD , Denise Descovich MD , Santiago Costantino PhD , Mark R. Lesk MD, MSc","doi":"10.1016/j.xops.2024.100576","DOIUrl":"10.1016/j.xops.2024.100576","url":null,"abstract":"<div><h3>Purpose</h3><p>To assess the relationship between the pulsatile choroidal volume change (ΔV) and ocular rigidity (OR), an important biomechanical property of the eye.</p></div><div><h3>Design</h3><p>This is a prospective cross-sectional study.</p></div><div><h3>Subjects</h3><p>Two hundred seventeen participants (235 eyes) were included in this study. Of those, 18 eyes (18 participants) had exudative retinal disease, and 217 eyes (199 participants) had open-angle glaucoma (39.2%), suspect discs (12.4%), ocular hypertension (14.3%), or healthy eyes (34.1%).</p></div><div><h3>Methods</h3><p>Pulsatile choroidal volume change was measured using dynamic OCT, which detects the change in choroidal thickness during the cardiac cycle. Ocular rigidity was measured using an invasive procedure as well as using a validated optical method. Correlations between ΔV and OR were assessed in subjects with healthy eyes, eyes with glaucoma, or eyes with exudative retinal disease.</p></div><div><h3>Main Outcome Measures</h3><p>Ocular rigidity and pulsatile ocular volume change.</p></div><div><h3>Results</h3><p>In 18 eyes where OR was obtained invasively and ΔV was obtained noninvasively, a significant correlation was found between ΔV and OR (r<sub>s</sub> = −0.664, <em>P</em> = 0.003). Similarly, a strong inverse correlation was found between the noninvasive measurements of both ΔV and OR (r<sub>s</sub> = −0.748, <em>P</em> < 0.001) in a large cohort and maintained its significance across diagnostic groups (a more compliant eye is associated with greater ΔV). No correlation was found between ΔV and age, blood pressure, intraocular pressure, axial length, or diagnosis (<em>P</em> ≥ 0.05). Mean ΔV was 7.3 ± 3.4 μL for all groups combined with a range of 3.0 to 20.8 μL.</p></div><div><h3>Conclusions</h3><p>These results suggest an association between the biomechanics of the corneoscleral shell and pulsatile ocular blood flow, which may indicate that a more rigid eye exerts more resistance to pulsatile choroidal expansion. This highlights the dynamic nature of both blood flow and biomechanics in the eye, as well as how they may interact, leading to a greater understanding of the pathophysiology of ocular disease.</p></div><div><h3>Financial Disclosures</h3><p>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</p></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"4 6","pages":"Article 100576"},"PeriodicalIF":3.2,"publicationDate":"2024-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S266691452400112X/pdfft?md5=ddd401d8b394fedad1dc8122bed5d1a8&pid=1-s2.0-S266691452400112X-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142020826","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Improving the Identification of Diabetic Retinopathy and Related Conditions in the Electronic Health Record Using Natural Language Processing Methods 利用自然语言处理方法改进电子健康记录中糖尿病视网膜病变及相关病症的识别工作

IF 3.2 Q1 OPHTHALMOLOGY

Ophthalmology science

Pub Date : 2024-07-18 DOI: 10.1016/j.xops.2024.100578

Keith Harrigian MS , Diep Tran MSc , Tina Tang MD , Anthony Gonzales OD , Paul Nagy PhD , Hadi Kharrazi MD, PhD , Mark Dredze PhD , Cindy X. Cai MD, MS

Purpose

To compare the performance of 3 phenotyping methods in identifying diabetic retinopathy (DR) and related clinical conditions.

Design

Three phenotyping methods were used to identify clinical conditions including unspecified DR, nonproliferative DR (NPDR) (mild, moderate, severe), consolidated NPDR (unspecified DR or any NPDR), proliferative DR, diabetic macular edema (DME), vitreous hemorrhage, retinal detachment (RD) (tractional RD or combined tractional and rhegmatogenous RD), and neovascular glaucoma (NVG). The first method used only International Classification of Diseases, 10th Revision (ICD-10) diagnosis codes (ICD-10 Lookup System). The next 2 methods used a Bidirectional Encoder Representations from Transformers with a dense Multilayer Perceptron output layer natural language processing (NLP) framework. The NLP framework was applied either to free-text of provider notes (Text-Only NLP System) or both free-text and ICD-10 diagnosis codes (Text-and-International Classification of Diseases [ICD] NLP System).

Subjects

Adults ≥18 years with diabetes mellitus seen at the Wilmer Eye Institute.

Methods

We compared the performance of the 3 phenotyping methods in identifying the DR related conditions with gold standard chart review. We also compared the estimated disease prevalence using each method.

Main Outcome Measures

Performance of each method was reported as the macro F1 score. The agreement between the methods was calculated using the kappa statistic. Prevalence estimates were also calculated for each method.

Results

A total of 91 097 patients and 692 486 office visits were included in the study. Compared with the gold standard, the Text-and-ICD NLP System had the highest F1 score for most clinical conditions (range 0.39–0.64). The agreement between the ICD-10 Lookup System and Text-Only NLP System varied (kappa of 0.21–0.81). The prevalence of DR and related conditions ranged from 1.1% for NVG to 17.9% for DME (using the Text-and-ICD NLP System).

Conclusions

The prevalence of DR and related conditions varied significantly depending on the methodology of identifying cases. The best performing phenotyping method was the Text-and-ICD NLP System that used information in both diagnosis codes as well as free-text notes.

Financial Disclosures

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

目的比较三种表型方法在识别糖尿病视网膜病变（DR）及相关临床症状方面的性能。设计使用三种表型方法来识别临床病症，包括未指定的 DR、非增殖性 DR (NPDR)（轻度、中度、重度）、合并的 NPDR（未指定的 DR 或任何 NPDR）、增殖性 DR、糖尿病黄斑水肿 (DME)、玻璃体出血、视网膜脱离 (RD)（牵引性 RD 或牵引性和流变性联合 RD）和新生血管性青光眼 (NVG)。第一种方法仅使用国际疾病分类第十版（ICD-10）诊断代码（ICD-10 查询系统）。接下来的两种方法使用了变压器双向编码器表示法和密集多层感知器输出层自然语言处理（NLP）框架。方法我们比较了 3 种表型分析方法与金标准病历审查在识别糖尿病相关疾病方面的性能。我们还比较了使用每种方法估计的疾病患病率。主要结果测量每种方法的性能以宏观 F1 分数报告。使用卡帕统计量计算各种方法之间的一致性。研究共纳入了 91 097 名患者和 692 486 次诊疗。与金标准相比，文本和 ICD NLP 系统在大多数临床情况下的 F1 分数最高（范围为 0.39-0.64）。ICD-10 查询系统和纯文本 NLP 系统之间的一致性各不相同（kappa 为 0.21-0.81）。DR及相关疾病的患病率从NVG的1.1%到DME的17.9%不等（使用文本和ICD NLP系统）。表现最好的表型鉴定方法是Text-and-ICD NLP系统，该系统使用了诊断代码和自由文本注释中的信息。

{"title":"Improving the Identification of Diabetic Retinopathy and Related Conditions in the Electronic Health Record Using Natural Language Processing Methods","authors":"Keith Harrigian MS , Diep Tran MSc , Tina Tang MD , Anthony Gonzales OD , Paul Nagy PhD , Hadi Kharrazi MD, PhD , Mark Dredze PhD , Cindy X. Cai MD, MS","doi":"10.1016/j.xops.2024.100578","DOIUrl":"10.1016/j.xops.2024.100578","url":null,"abstract":"<div><h3>Purpose</h3><p>To compare the performance of 3 phenotyping methods in identifying diabetic retinopathy (DR) and related clinical conditions.</p></div><div><h3>Design</h3><p>Three phenotyping methods were used to identify clinical conditions including unspecified DR, nonproliferative DR (NPDR) (mild, moderate, severe), consolidated NPDR (unspecified DR or any NPDR), proliferative DR, diabetic macular edema (DME), vitreous hemorrhage, retinal detachment (RD) (tractional RD or combined tractional and rhegmatogenous RD), and neovascular glaucoma (NVG). The first method used only International Classification of Diseases, 10th Revision (ICD-10) diagnosis codes (<em>ICD-10 Lookup System</em>). The next 2 methods used a Bidirectional Encoder Representations from Transformers with a dense Multilayer Perceptron output layer natural language processing (NLP) framework. The NLP framework was applied either to free-text of provider notes (<em>Text-Only NLP System</em>) or both free-text and ICD-10 diagnosis codes (<em>Text-and-International Classification of Diseases</em> [<em>ICD</em>] <em>NLP System</em>).</p></div><div><h3>Subjects</h3><p>Adults ≥18 years with diabetes mellitus seen at the Wilmer Eye Institute.</p></div><div><h3>Methods</h3><p>We compared the performance of the 3 phenotyping methods in identifying the DR related conditions with gold standard chart review. We also compared the estimated disease prevalence using each method.</p></div><div><h3>Main Outcome Measures</h3><p>Performance of each method was reported as the macro F1 score. The agreement between the methods was calculated using the kappa statistic. Prevalence estimates were also calculated for each method.</p></div><div><h3>Results</h3><p>A total of 91 097 patients and 692 486 office visits were included in the study. Compared with the gold standard, the <em>Text-and-ICD NLP System</em> had the highest F1 score for most clinical conditions (range 0.39–0.64). The agreement between the <em>ICD-10 Lookup System</em> and <em>Text-Only NLP System</em> varied (kappa of 0.21–0.81). The prevalence of DR and related conditions ranged from 1.1% for NVG to 17.9% for DME (using the <em>Text-and-ICD NLP System</em>).</p></div><div><h3>Conclusions</h3><p>The prevalence of DR and related conditions varied significantly depending on the methodology of identifying cases. The best performing phenotyping method was the <em>Text-and-ICD NLP System</em> that used information in both diagnosis codes as well as free-text notes.</p></div><div><h3>Financial Disclosures</h3><p>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</p></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"4 6","pages":"Article 100578"},"PeriodicalIF":3.2,"publicationDate":"2024-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S2666914524001143/pdfft?md5=aee0aca9014224fef1aa919db24f5c88&pid=1-s2.0-S2666914524001143-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141844268","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Sex Differences in Inflammation-Related Biomarkers Detected with OCT in Patients with Diabetic Macular Edema 用光学相干断层扫描技术检测糖尿病黄斑水肿患者炎症相关生物标志物的性别差异

IF 3.2 Q1 OPHTHALMOLOGY

Ophthalmology science

Pub Date : 2024-07-18 DOI: 10.1016/j.xops.2024.100580

Xinyi Chen MD , Wendy Yang BS , Ashley Fong MS , Noor Chahal BS , Abu T. Taha BS , Jeremy D. Keenan MD, MPH , Jay M. Stewart MD

Purpose

To investigate sex-based differences in inflammation-related biomarkers on spectral-domain OCT.

Design

Cross-sectional study.

Participants

Patients with diabetic macular edema (DME) between February 1, 2019, and March 31, 2023, without intravitreal anti-VEGF injection within the previous 6 months.

Methods

We reviewed each patient’s medical record for age, biological sex, race and ethnicity, most recent glycated hemoglobin A1c (HbA1c) level, visual acuity (VA), and central macular thickness (CMT). OCT biomarkers that have been found in literature to be associated with inflammation, including disorganization of retinal inner layers (DRIL), retinal hyperreflective retinal foci (HRFs), hyperreflective choroidal foci (HCFs), subfoveal neuroretinal detachment (SND), and perturbation in retinal nerve fiber layer thickness, ganglion cell layer thickness, and inner nuclear layer (INL) thickness were evaluated by graders masked to the clinical characteristics of the patients. We performed multivariable regression analyses with the OCT biomarkers as the outcome variables and sex, age, HbA1c, and CMT as independent variables.

Main Outcome Measures

OCT inflammation-related biomarkers, as listed above.

Results

Female patients were, on average, 2 years older than male patients (P = 0.041). There were no significant differences in race and ethnicity, HbA1c, VA, or CMT between male and female patients. After controlling for age, HbA1c, and CMT, we found male sex to be associated with more HRF (incidence rate ratio [IRR] = 1.19; 95% confidence interval [CI] = 1.10–1.29), more HCF (odds ratio = 2.01; 95% CI = 1.12–3.64), and thicker INL (7 μm thicker in males; 95% CI = 2–12). Sex was not a significant predictor for either DRIL or SND in the multivariable regression models. Patients with higher HbA1c were more likely to have more HRF (IRR = 1.02 per 1 point increase; 95% CI = 1.00–1.04) after controlling for other factors.

Conclusions

Male sex was correlated with more inflammation-related biomarkers on OCT including more HRF, more HCF, and thicker INL, after accounting for age, glycemic control, and amount of DME. Further studies are needed to evaluate the potential implications of these sex-based differences for individualized treatment.

Financial Disclosures

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

目的研究光谱域 OCT 上炎症相关生物标志物的性别差异。方法我们查看了每位患者的病历，包括年龄、生理性别、种族和民族、最近的糖化血红蛋白 A1c (HbA1c) 水平、视力 (VA)、视力敏锐度 (VA)、视力敏锐度 (VA) 和视力敏锐度 (VA)。方法我们查看了每位患者的病历，包括年龄、生理性别、种族和民族、最近的糖化血红蛋白 A1c (HbA1c) 水平、视力 (VA) 和黄斑中心厚度 (CMT)。文献中发现的与炎症相关的 OCT 生物标记物包括视网膜内层紊乱（DRIL）、视网膜高反射灶（HRF）、脉络膜高反射灶（HCF）、眼底神经视网膜脱离（SND）以及视网膜神经纤维层厚度、神经节细胞层厚度和内核层（INL）厚度的扰动，这些标记物均由对患者临床特征进行掩蔽的分级人员进行评估。我们以 OCT 生物标志物为结果变量，以性别、年龄、HbA1c 和 CMT 为自变量，进行了多变量回归分析。结果女性患者比男性患者平均年龄大 2 岁（P = 0.041）。男女患者在种族和民族、HbA1c、VA 或 CMT 方面没有明显差异。在控制了年龄、HbA1c 和 CMT 后，我们发现男性与更多的 HRF（发病率比 [IRR] = 1.19；95% 置信区间 [CI] = 1.10-1.29）、更多的 HCF（几率比 = 2.01；95% 置信区间 = 1.12-3.64）和更厚的 INL（男性更厚 7 μm；95% 置信区间 = 2-12）相关。在多变量回归模型中，性别不是 DRIL 或 SND 的重要预测因素。结论在考虑年龄、血糖控制和 DME 的数量后，男性性别与 OCT 上更多的炎症相关生物标记物相关，包括更多的 HRF、更多的 HCF 和更厚的 INL。需要进一步的研究来评估这些性别差异对个体化治疗的潜在影响。

{"title":"Sex Differences in Inflammation-Related Biomarkers Detected with OCT in Patients with Diabetic Macular Edema","authors":"Xinyi Chen MD , Wendy Yang BS , Ashley Fong MS , Noor Chahal BS , Abu T. Taha BS , Jeremy D. Keenan MD, MPH , Jay M. Stewart MD","doi":"10.1016/j.xops.2024.100580","DOIUrl":"10.1016/j.xops.2024.100580","url":null,"abstract":"<div><h3>Purpose</h3><p>To investigate sex-based differences in inflammation-related biomarkers on spectral-domain OCT.</p></div><div><h3>Design</h3><p>Cross-sectional study.</p></div><div><h3>Participants</h3><p>Patients with diabetic macular edema (DME) between February 1, 2019, and March 31, 2023, without intravitreal anti-VEGF injection within the previous 6 months.</p></div><div><h3>Methods</h3><p>We reviewed each patient’s medical record for age, biological sex, race and ethnicity, most recent glycated hemoglobin A1c (HbA1c) level, visual acuity (VA), and central macular thickness (CMT). OCT biomarkers that have been found in literature to be associated with inflammation, including disorganization of retinal inner layers (DRIL), retinal hyperreflective retinal foci (HRFs), hyperreflective choroidal foci (HCFs), subfoveal neuroretinal detachment (SND), and perturbation in retinal nerve fiber layer thickness, ganglion cell layer thickness, and inner nuclear layer (INL) thickness were evaluated by graders masked to the clinical characteristics of the patients. We performed multivariable regression analyses with the OCT biomarkers as the outcome variables and sex, age, HbA1c, and CMT as independent variables.</p></div><div><h3>Main Outcome Measures</h3><p>OCT inflammation-related biomarkers, as listed above.</p></div><div><h3>Results</h3><p>Female patients were, on average, 2 years older than male patients (<em>P</em> = 0.041). There were no significant differences in race and ethnicity, HbA1c, VA, or CMT between male and female patients. After controlling for age, HbA1c, and CMT, we found male sex to be associated with more HRF (incidence rate ratio [IRR] = 1.19; 95% confidence interval [CI] = 1.10–1.29), more HCF (odds ratio = 2.01; 95% CI = 1.12–3.64), and thicker INL (7 μm thicker in males; 95% CI = 2–12). Sex was not a significant predictor for either DRIL or SND in the multivariable regression models. Patients with higher HbA1c were more likely to have more HRF (IRR = 1.02 per 1 point increase; 95% CI = 1.00–1.04) after controlling for other factors.</p></div><div><h3>Conclusions</h3><p>Male sex was correlated with more inflammation-related biomarkers on OCT including more HRF, more HCF, and thicker INL, after accounting for age, glycemic control, and amount of DME. Further studies are needed to evaluate the potential implications of these sex-based differences for individualized treatment.</p></div><div><h3>Financial Disclosures</h3><p>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</p></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"4 6","pages":"Article 100580"},"PeriodicalIF":3.2,"publicationDate":"2024-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S2666914524001167/pdfft?md5=fd1a7303565f3c755d6452f95c65ef8d&pid=1-s2.0-S2666914524001167-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141848952","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Posterior Eye Shape in Myopia 近视眼的眼球后部形状

IF 3.2 Q1 OPHTHALMOLOGY

Ophthalmology science

Pub Date : 2024-07-06 DOI: 10.1016/j.xops.2024.100575

Jost B. Jonas MD , Songhomitra Panda-Jonas MD , Zhe Pan MD , Jie Xu MD , Ya Xing Wang MD

Purpose

To explore prevalence and associated factors of abnormalities of the posterior eye shape in dependence of axial length.

Design

Population-based study.

Participants

Of the participants (n = 3468) of the Beijing Eye Study, we included all eyes with an axial length of ≥25 mm, and a randomized sample of eyes with an axial length of <25 mm.

Methods

Using 30°-wide, serial horizontal, and fovea-centered radial, OCT images, we examined location and depth of the most posterior point of the retinal pigment epithelium/Bruch’s membrane line (PP-RPE/BML).

Main Outcome Measures

Prevalence and depth of an extrafoveal PP-RPE/BML.

Results

The study included 366 eyes (314 individuals). On the radial OCT scans, the PP-RPE/BML was located in the foveola in 190 (51.9%) eyes, in 121 (33.1%) eyes in the 6 o’clock part of the vertical meridian (distance to foveola: 1.73 ± 0.70 mm), and in 54 (14.8%) eyes in the 12 o’clock part of the vertical meridian (fovea distance: 2.01 ± 0.66 mm). On the horizontal OCT scans, the PP-RPE/BML was located in the foveola in 304 (83.1%) eyes, between foveola and optic disc in 36 (9.8%) eyes (fovea distance: 1.59 ± 0.76 mm), and temporal to the foveola in 26 (7.1%) eyes (fovea distance: 1.20 ± 0.60 mm). Higher prevalence of an extrafoveal PP-RPE/BML correlated with longer axial length (odds ratio [OR]: 1.55; 95% confidence interval [CI]: 1.28, 1.89), higher corneal astigmatism (OR: 1.78; 95% CI: 1.14, 2.79), and female sex (OR: 2.74; 95% CI: 1.30, 5.77). The curvature of the RPE/BML at the posterior pole was similar to the RPE/BML curvature outside of the posterior pole in 309 (84.4%) eyes, and it was steeper (i.e., smaller curvature radius) in 57 (15.6%) eyes. In these eyes, axial length was longer (24.41 ± 1.78 mm versus 27.74 ± 1.88 mm; P < 0.001).

Conclusions

With longer axial length, the foveola is more often located outside of the geometrical posterior pole. It may be of importance for biometric axial length measurements. An extrafoveal location of the PP-RPE/BML may be due to an axial elongation-associated, meridionally asymmetric enlargement of Bruch’s membrane in the fundus midperiphery.

Financial Disclosure(s)

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

目的探讨眼球后部形状异常的发生率和相关因素与眼轴长度的关系。参与者在北京眼科研究的参与者（n = 3468）中，我们纳入了所有眼轴长度≥25 mm的眼睛，以及眼轴长度为<25 mm的随机样本。方法利用 30° 宽、连续水平和以眼窝为中心的径向 OCT 图像，我们检查了视网膜色素上皮/布氏膜线（PP-RPE/BML）最后方点的位置和深度。结果研究共纳入 366 只眼睛（314 人）。在径向 OCT 扫描中，190 只眼睛（51.9%）的 PP-RPE/BML 位于眼窝，121 只眼睛（33.1%）的 PP-RPE/BML 位于垂直子午线的 6 点钟部分（到眼窝的距离：1.73 ± 0.70 毫米），54 只眼睛（14.8%）的 PP-RPE/BML 位于垂直子午线的 12 点钟部分（到眼窝的距离：2.01 ± 0.66 毫米）。在水平 OCT 扫描中，304 只（83.1%）眼睛的 PP-RPE/BML 位于眼窝内，36 只（9.8%）眼睛的 PP-RPE/BML 位于眼窝和视盘之间（眼窝距离：1.59 ± 0.76 毫米），26 只（7.1%）眼睛的 PP-RPE/BML 位于眼窝的颞侧（眼窝距离：1.20 ± 0.60 毫米）。视网膜外 PP-RPE/BML 的高发生率与较长的轴长（几率比 [OR]：1.55；95% 置信区间 [CI]：1.28，1.89）、较高的角膜散光（OR：1.78；95% CI：1.14，2.79）和女性性别（OR：2.74；95% CI：1.30，5.77）相关。在 309 只（84.4%）眼睛中，后极部的 RPE/BML 曲度与后极部以外的 RPE/BML 曲度相似，而在 57 只（15.6%）眼睛中，后极部的 RPE/BML 曲度较陡（即曲率半径较小）。这些眼睛的眼轴长度较长（24.41 ± 1.78 mm 对 27.74 ± 1.88 mm；P < 0.001）。这可能对生物测定轴长具有重要意义。PP-RPE/BML的眼窝外位置可能是由于与轴向伸长相关的、眼底中周布鲁氏膜的子午线不对称扩大所致。

{"title":"Posterior Eye Shape in Myopia","authors":"Jost B. Jonas MD , Songhomitra Panda-Jonas MD , Zhe Pan MD , Jie Xu MD , Ya Xing Wang MD","doi":"10.1016/j.xops.2024.100575","DOIUrl":"10.1016/j.xops.2024.100575","url":null,"abstract":"<div><h3>Purpose</h3><p>To explore prevalence and associated factors of abnormalities of the posterior eye shape in dependence of axial length.</p></div><div><h3>Design</h3><p>Population-based study.</p></div><div><h3>Participants</h3><p>Of the participants (n = 3468) of the Beijing Eye Study, we included all eyes with an axial length of ≥25 mm, and a randomized sample of eyes with an axial length of <25 mm.</p></div><div><h3>Methods</h3><p>Using 30°-wide, serial horizontal, and fovea-centered radial, OCT images, we examined location and depth of the most posterior point of the retinal pigment epithelium/Bruch’s membrane line (PP-RPE/BML).</p></div><div><h3>Main Outcome Measures</h3><p>Prevalence and depth of an extrafoveal PP-RPE/BML.</p></div><div><h3>Results</h3><p>The study included 366 eyes (314 individuals). On the radial OCT scans, the PP-RPE/BML was located in the foveola in 190 (51.9%) eyes, in 121 (33.1%) eyes in the 6 o’clock part of the vertical meridian (distance to foveola: 1.73 ± 0.70 mm), and in 54 (14.8%) eyes in the 12 o’clock part of the vertical meridian (fovea distance: 2.01 ± 0.66 mm). On the horizontal OCT scans, the PP-RPE/BML was located in the foveola in 304 (83.1%) eyes, between foveola and optic disc in 36 (9.8%) eyes (fovea distance: 1.59 ± 0.76 mm), and temporal to the foveola in 26 (7.1%) eyes (fovea distance: 1.20 ± 0.60 mm). Higher prevalence of an extrafoveal PP-RPE/BML correlated with longer axial length (odds ratio [OR]: 1.55; 95% confidence interval [CI]: 1.28, 1.89), higher corneal astigmatism (OR: 1.78; 95% CI: 1.14, 2.79), and female sex (OR: 2.74; 95% CI: 1.30, 5.77). The curvature of the RPE/BML at the posterior pole was similar to the RPE/BML curvature outside of the posterior pole in 309 (84.4%) eyes, and it was steeper (i.e., smaller curvature radius) in 57 (15.6%) eyes. In these eyes, axial length was longer (24.41 ± 1.78 mm versus 27.74 ± 1.88 mm; <em>P</em> < 0.001).</p></div><div><h3>Conclusions</h3><p>With longer axial length, the foveola is more often located outside of the geometrical posterior pole. It may be of importance for biometric axial length measurements. An extrafoveal location of the PP-RPE/BML may be due to an axial elongation-associated, meridionally asymmetric enlargement of Bruch’s membrane in the fundus midperiphery.</p></div><div><h3>Financial Disclosure(s)</h3><p>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</p></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"4 6","pages":"Article 100575"},"PeriodicalIF":3.2,"publicationDate":"2024-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S2666914524001118/pdfft?md5=42f31daebc61edd63d9f5d3810c4a00e&pid=1-s2.0-S2666914524001118-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141715410","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Quantifying Changes on OCT in Eyes Receiving Treatment for Neovascular Age-Related Macular Degeneration 量化接受新生血管性老年黄斑变性治疗的眼睛在光学视网膜断层扫描上的变化

IF 3.2 Q1 OPHTHALMOLOGY

Ophthalmology science

Pub Date : 2024-06-28 DOI: 10.1016/j.xops.2024.100570

Gabriella Moraes MD, MSc , Robbert Struyven MD , Siegfried K. Wagner BMBCh, FRCOphth , Timing Liu BA , David Chong MBBChir , Abdallah Abbas iBSc, MBBS , Reena Chopra BSc , Praveen J. Patel MD, FRCOphth , Konstantinos Balaskas MD , Tiarnan D.L. Keenan BM BCh, PhD , Pearse A. Keane MD, FRCOphth

Purpose

Application of artificial intelligence (AI) to macular OCT scans to segment and quantify volumetric change in anatomical and pathological features during intravitreal treatment for neovascular age-related macular degeneration (AMD).

Design

Retrospective analysis of OCT images from the Moorfields Eye Hospital AMD Database.

Participants

A total of 2115 eyes from 1801 patients starting anti-VEGF treatment between June 1, 2012, and June 30, 2017.

Methods

The Moorfields Eye Hospital neovascular AMD database was queried for first and second eyes receiving anti-VEGF treatment and had an OCT scan at baseline and 12 months. Follow-up scans were input into the AI system and volumes of OCT variables were studied at different time points and compared with baseline volume groups. Cross-sectional comparisons between time points were conducted using Mann–Whitney U test.

Main Outcome Measures

Volume outputs of the following variables were studied: intraretinal fluid, subretinal fluid, pigment epithelial detachment (PED), subretinal hyperreflective material (SHRM), hyperreflective foci, neurosensory retina, and retinal pigment epithelium.

Results

Mean volumes of analyzed features decreased significantly from baseline to both 4 and 12 months, in both first-treated and second-treated eyes. Pathological features that reflect exudation, including pure fluid components (intraretinal fluid and subretinal fluid) and those with fluid and fibrovascular tissue (PED and SHRM), displayed similar responses to treatment over 12 months. Mean PED and SHRM volumes showed less pronounced but also substantial decreases over the first 2 months, reaching a plateau postloading phase, and minimal change to 12 months. Both neurosensory retina and retinal pigment epithelium volumes showed gradual reductions over time, and were not as substantial as exudative features.

Conclusions

We report the results of a quantitative analysis of change in retinal segmented features over time, enabled by an AI segmentation system. Cross-sectional analysis at multiple time points demonstrated significant associations between baseline OCT-derived segmented features and the volume of biomarkers at follow-up. Demonstrating how certain OCT biomarkers progress with treatment and the impact of pretreatment retinal morphology on different structural volumes may provide novel insights into disease mechanisms and aid the personalization of care. Data will be made public for future studies.

Financial Disclosure(s)

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

目的将人工智能（AI）应用于黄斑OCT扫描，以分割和量化玻璃体内治疗新生血管性年龄相关性黄斑变性（AMD）过程中解剖和病理特征的体积变化.设计对Moorfields眼科医院AMD数据库中的OCT图像进行回顾性分析.方法查询Moorfields眼科医院新生血管性AMD数据库中接受抗血管内皮生长因子治疗的第一只和第二只眼睛，并在基线和12个月时进行OCT扫描。将随访扫描结果输入 AI 系统，研究不同时间点的 OCT 变量体积，并与基线体积组进行比较。主要结果测量研究了以下变量的体积输出：视网膜内积液、视网膜下积液、色素上皮脱落（PED）、视网膜下超反光物质（SHRM）、超反光灶、神经感觉视网膜和视网膜色素上皮。结果分析特征的平均体积从基线到4个月和12个月都显著下降，在第一次治疗和第二次治疗的眼睛中都是如此。反映渗出的病理特征，包括纯液体成分（视网膜内液和视网膜下液）和含有液体和纤维血管组织的病理特征（PED 和 SHRM），在 12 个月的治疗中表现出相似的反应。PED 和 SHRM 的平均体积在最初 2 个月的下降并不明显，但也很显著，在加载阶段后达到平稳，12 个月后变化很小。随着时间的推移，神经感觉视网膜和视网膜色素上皮的体积也逐渐减少，但减少幅度不如渗出性特征大。多个时间点的横断面分析表明，基线 OCT 导出的分段特征与随访时生物标志物的体积之间存在显著关联。展示某些OCT生物标记物如何随治疗而进展，以及治疗前视网膜形态对不同结构体积的影响，可为疾病机制提供新的见解，并有助于个性化治疗。数据将在未来的研究中公开。财务信息披露：专有或商业信息披露请参见本文末尾的脚注和披露。

{"title":"Quantifying Changes on OCT in Eyes Receiving Treatment for Neovascular Age-Related Macular Degeneration","authors":"Gabriella Moraes MD, MSc , Robbert Struyven MD , Siegfried K. Wagner BMBCh, FRCOphth , Timing Liu BA , David Chong MBBChir , Abdallah Abbas iBSc, MBBS , Reena Chopra BSc , Praveen J. Patel MD, FRCOphth , Konstantinos Balaskas MD , Tiarnan D.L. Keenan BM BCh, PhD , Pearse A. Keane MD, FRCOphth","doi":"10.1016/j.xops.2024.100570","DOIUrl":"10.1016/j.xops.2024.100570","url":null,"abstract":"<div><h3>Purpose</h3><p>Application of artificial intelligence (AI) to macular OCT scans to segment and quantify volumetric change in anatomical and pathological features during intravitreal treatment for neovascular age-related macular degeneration (AMD).</p></div><div><h3>Design</h3><p>Retrospective analysis of OCT images from the Moorfields Eye Hospital AMD Database.</p></div><div><h3>Participants</h3><p>A total of 2115 eyes from 1801 patients starting anti-VEGF treatment between June 1, 2012, and June 30, 2017.</p></div><div><h3>Methods</h3><p>The Moorfields Eye Hospital neovascular AMD database was queried for first and second eyes receiving anti-VEGF treatment and had an OCT scan at baseline and 12 months. Follow-up scans were input into the AI system and volumes of OCT variables were studied at different time points and compared with baseline volume groups. Cross-sectional comparisons between time points were conducted using Mann–Whitney <em>U</em> test.</p></div><div><h3>Main Outcome Measures</h3><p>Volume outputs of the following variables were studied: intraretinal fluid, subretinal fluid, pigment epithelial detachment (PED), subretinal hyperreflective material (SHRM), hyperreflective foci, neurosensory retina, and retinal pigment epithelium.</p></div><div><h3>Results</h3><p>Mean volumes of analyzed features decreased significantly from baseline to both 4 and 12 months, in both first-treated and second-treated eyes. Pathological features that reflect exudation, including pure fluid components (intraretinal fluid and subretinal fluid) and those with fluid and fibrovascular tissue (PED and SHRM), displayed similar responses to treatment over 12 months. Mean PED and SHRM volumes showed less pronounced but also substantial decreases over the first 2 months, reaching a plateau postloading phase, and minimal change to 12 months. Both neurosensory retina and retinal pigment epithelium volumes showed gradual reductions over time, and were not as substantial as exudative features.</p></div><div><h3>Conclusions</h3><p>We report the results of a quantitative analysis of change in retinal segmented features over time, enabled by an AI segmentation system. Cross-sectional analysis at multiple time points demonstrated significant associations between baseline OCT-derived segmented features and the volume of biomarkers at follow-up. Demonstrating how certain OCT biomarkers progress with treatment and the impact of pretreatment retinal morphology on different structural volumes may provide novel insights into disease mechanisms and aid the personalization of care. Data will be made public for future studies.</p></div><div><h3>Financial Disclosure(s)</h3><p>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</p></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"4 6","pages":"Article 100570"},"PeriodicalIF":3.2,"publicationDate":"2024-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S2666914524001064/pdfft?md5=ba7935cea71518e47f805f746ad59e08&pid=1-s2.0-S2666914524001064-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141985193","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Advancing Glaucoma Diagnosis: Employing Confidence-Calibrated Label Smoothing Loss for Model Calibration 推进青光眼诊断：采用置信度校准标签平滑损失进行模型校准

IF 3.2 Q1 OPHTHALMOLOGY

Ophthalmology science

Pub Date : 2024-06-22 DOI: 10.1016/j.xops.2024.100555

Midhula Vijayan PhD, Deepthi Keshav Prasad PhD, Venkatakrishnan Srinivasan MTech

Objective

The aim of our research is to enhance the calibration of machine learning models for glaucoma classification through a specialized loss function named Confidence-Calibrated Label Smoothing (CC-LS) loss. This approach is specifically designed to refine model calibration without compromising accuracy by integrating label smoothing and confidence penalty techniques, tailored to the specifics of glaucoma detection.

Design

This study focuses on the development and evaluation of a calibrated deep learning model.

Participants

The study employs fundus images from both external datasets—the Online Retinal Fundus Image Database for Glaucoma Analysis and Research (482 normal, 168 glaucoma) and the Retinal Fundus Glaucoma Challenge (720 normal, 80 glaucoma)—and an extensive internal dataset (4639 images per category), aiming to bolster the model's generalizability. The model's clinical performance is validated using a comprehensive test set (47 913 normal, 1629 glaucoma) from the internal dataset.

Methods

The CC-LS loss function seamlessly integrates label smoothing, which tempers extreme predictions to avoid overfitting, with confidence-based penalties. These penalties deter the model from expressing undue confidence in incorrect classifications. Our study aims at training models using the CC-LS and comparing their performance with those trained using conventional loss functions.

Main Outcome Measures

The model's precision is evaluated using metrics like the Brier score, sensitivity, specificity, and the false positive rate, alongside qualitative heatmap analyses for a holistic accuracy assessment.

Results

Preliminary findings reveal that models employing the CC-LS mechanism exhibit superior calibration metrics, as evidenced by a Brier score of 0.098, along with notable accuracy measures: sensitivity of 81%, specificity of 80%, and weighted accuracy of 80%. Importantly, these enhancements in calibration are achieved without sacrificing classification accuracy.

Conclusions

The CC-LS loss function presents a significant advancement in the pursuit of deploying machine learning models for glaucoma diagnosis. By improving calibration, the CC-LS ensures that clinicians can interpret and trust the predictive probabilities, making artificial intelligence-driven diagnostic tools more clinically viable. From a clinical standpoint, this heightened trust and interpretability can potentially lead to more timely and appropriate interventions, thereby optimizing patient outcomes and safety.

Financial Disclosure(s)

Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

目标我们的研究旨在通过一种名为 "置信度校准标签平滑（CC-LS）损失 "的专门损失函数，加强青光眼分类机器学习模型的校准。这种方法通过整合标签平滑和置信度惩罚技术，专门针对青光眼检测的具体情况，在不影响准确性的前提下完善模型校准。设计本研究重点关注校准深度学习模型的开发和评估。参与者该研究采用了来自外部数据集--用于青光眼分析和研究的在线视网膜眼底图像数据库（482 张正常图像，168 张青光眼图像）和视网膜眼底青光眼挑战赛（720 张正常图像，80 张青光眼图像）--以及广泛的内部数据集（每个类别 4639 张图像）的眼底图像，旨在增强模型的通用性。该模型的临床性能通过内部数据集的综合测试集（47 913 张正常图像、1629 张青光眼图像）进行了验证。方法CC-LS 损失函数将标签平滑与基于置信度的惩罚无缝整合在一起，标签平滑可以缓和极端预测以避免过度拟合。这些惩罚措施可防止模型对不正确的分类表现出过度的信心。我们的研究旨在使用 CC-LS 训练模型，并将它们的性能与使用传统损失函数训练的模型进行比较。主要结果测量使用 Brier 分数、灵敏度、特异性和假阳性率等指标评估模型的精确度，同时进行定性热图分析，以全面评估精确度。结果初步研究结果表明，采用 CC-LS 机制的模型显示出更优越的校准指标，具体表现为布赖尔评分为 0.098，以及显著的准确性指标：灵敏度为 81%，特异性为 80%，加权准确性为 80%。重要的是，这些校准方面的改进是在不牺牲分类准确性的前提下实现的。结论CC-LS 损失函数在为青光眼诊断部署机器学习模型方面取得了重大进展。通过改进校准，CC-LS 可确保临床医生能够解释并信任预测概率，从而使人工智能驱动的诊断工具在临床上更加可行。从临床角度来看，这种信任度和可解释性的提高有可能带来更及时、更适当的干预，从而优化患者的治疗效果和安全性。

{"title":"Advancing Glaucoma Diagnosis: Employing Confidence-Calibrated Label Smoothing Loss for Model Calibration","authors":"Midhula Vijayan PhD, Deepthi Keshav Prasad PhD, Venkatakrishnan Srinivasan MTech","doi":"10.1016/j.xops.2024.100555","DOIUrl":"10.1016/j.xops.2024.100555","url":null,"abstract":"<div><h3>Objective</h3><p>The aim of our research is to enhance the calibration of machine learning models for glaucoma classification through a specialized loss function named Confidence-Calibrated Label Smoothing (CC-LS) loss. This approach is specifically designed to refine model calibration without compromising accuracy by integrating label smoothing and confidence penalty techniques, tailored to the specifics of glaucoma detection.</p></div><div><h3>Design</h3><p>This study focuses on the development and evaluation of a calibrated deep learning model.</p></div><div><h3>Participants</h3><p>The study employs fundus images from both external datasets—the Online Retinal Fundus Image Database for Glaucoma Analysis and Research (482 normal, 168 glaucoma) and the Retinal Fundus Glaucoma Challenge (720 normal, 80 glaucoma)—and an extensive internal dataset (4639 images per category), aiming to bolster the model's generalizability. The model's clinical performance is validated using a comprehensive test set (47 913 normal, 1629 glaucoma) from the internal dataset.</p></div><div><h3>Methods</h3><p>The CC-LS loss function seamlessly integrates label smoothing, which tempers extreme predictions to avoid overfitting, with confidence-based penalties. These penalties deter the model from expressing undue confidence in incorrect classifications. Our study aims at training models using the CC-LS and comparing their performance with those trained using conventional loss functions.</p></div><div><h3>Main Outcome Measures</h3><p>The model's precision is evaluated using metrics like the Brier score, sensitivity, specificity, and the false positive rate, alongside qualitative heatmap analyses for a holistic accuracy assessment.</p></div><div><h3>Results</h3><p>Preliminary findings reveal that models employing the CC-LS mechanism exhibit superior calibration metrics, as evidenced by a Brier score of 0.098, along with notable accuracy measures: sensitivity of 81%, specificity of 80%, and weighted accuracy of 80%. Importantly, these enhancements in calibration are achieved without sacrificing classification accuracy.</p></div><div><h3>Conclusions</h3><p>The CC-LS loss function presents a significant advancement in the pursuit of deploying machine learning models for glaucoma diagnosis. By improving calibration, the CC-LS ensures that clinicians can interpret and trust the predictive probabilities, making artificial intelligence-driven diagnostic tools more clinically viable. From a clinical standpoint, this heightened trust and interpretability can potentially lead to more timely and appropriate interventions, thereby optimizing patient outcomes and safety.</p></div><div><h3>Financial Disclosure(s)</h3><p>Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</p></div>","PeriodicalId":74363,"journal":{"name":"Ophthalmology science","volume":"4 6","pages":"Article 100555"},"PeriodicalIF":3.2,"publicationDate":"2024-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S2666914524000915/pdfft?md5=427dfb03f669c0bff9531a39938549c2&pid=1-s2.0-S2666914524000915-main.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142006395","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0