Pub Date : 2025-01-01DOI: 10.1016/j.imu.2024.101606
Saleh Alrashed , Nasro Min-Allah
With the emergence of ever-improving quantum computers, technology is making its way to revolutionize many fields, and the medical sector is no exception. Recent efforts have explored applications of quantum computing in areas such as drug discovery, patient privacy, and information security. It is expected that, with improved and stable quantum computing technologies, the medical sector will benefit significantly in many areas, including efficient patient care, reduced clinical trial durations, enhanced imaging technologies, and post-quantum cryptography, to name a few.
In this work, we highlight recent advancements in the medical sector driven by quantum computing, encompassing computation, optimization, security, machine learning, data processing, simulation, and healthcare perspectives. We also discuss the limitations of current technologies, and the challenges associated with the quantum computing revolution.
{"title":"Quantum computing research in medical sciences","authors":"Saleh Alrashed , Nasro Min-Allah","doi":"10.1016/j.imu.2024.101606","DOIUrl":"10.1016/j.imu.2024.101606","url":null,"abstract":"<div><div>With the emergence of ever-improving quantum computers, technology is making its way to revolutionize many fields, and the medical sector is no exception. Recent efforts have explored applications of quantum computing in areas such as drug discovery, patient privacy, and information security. It is expected that, with improved and stable quantum computing technologies, the medical sector will benefit significantly in many areas, including efficient patient care, reduced clinical trial durations, enhanced imaging technologies, and post-quantum cryptography, to name a few.</div><div>In this work, we highlight recent advancements in the medical sector driven by quantum computing, encompassing computation, optimization, security, machine learning, data processing, simulation, and healthcare perspectives. We also discuss the limitations of current technologies, and the challenges associated with the quantum computing revolution.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"52 ","pages":"Article 101606"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143178835","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}
Pub Date : 2025-01-01DOI: 10.1016/j.imu.2024.101610
Manal Almalki
Background
The COVID-19 pandemic significantly altered health behaviors, particularly among adult students in Saudi Arabia. The increased use of walking-tracking apps and the challenges faced by individuals with chronic medical conditions have influenced overall quality of life (QOL).
Objective
To assess the influence of having a medical condition and the use of walking-tracking apps on QOL among adult students in Saudi Arabia.
Methods
An online questionnaire was utilized in June 2024 to measure QOL using the WHOQOL-BREF scale, which covers physical health, psychological well-being, social relationships, and environmental health. Participants were grouped based on their use of walking-tracking apps and the presence of a chronic medical condition. Statistical analysis included independent t-tests, Pearson correlations, and chi-square tests to determine significant associations (p < 0.05).
Results
The sample consisted of 412 participants. The chi-square test revealed a significant association between having a medical condition and using a walking-tracking app (p = 0.037), with individuals without medical conditions being more likely to use these apps. However, despite the high prevalence of app usage (65.3 %), no significant improvements in QOL were observed for app users across any of the QOL domains. Participants with medical conditions reported significantly higher QOL scores in all domains, particularly in psychological health (p < 0.001) and social relationships (p = 0.001). Positive correlations were observed for factors like meaningful life, concentration, and access to healthcare among those with medical conditions.
Conclusion
Students with chronic medical conditions reported higher QOL whereas the use of walking-tracking apps had limited direct impact on their QOL. Future studies should explore factors that play a critical role in enhancing QOL beyond physical health and technology usage, including social support and the Saudi healthcare system.
{"title":"The role of walking-tracking apps and chronic medical conditions for adult students’ quality of life: A cross-sectional study from Saudi Arabia","authors":"Manal Almalki","doi":"10.1016/j.imu.2024.101610","DOIUrl":"10.1016/j.imu.2024.101610","url":null,"abstract":"<div><h3>Background</h3><div>The COVID-19 pandemic significantly altered health behaviors, particularly among adult students in Saudi Arabia. The increased use of walking-tracking apps and the challenges faced by individuals with chronic medical conditions have influenced overall quality of life (QOL).</div></div><div><h3>Objective</h3><div>To assess the influence of having a medical condition and the use of walking-tracking apps on QOL among adult students in Saudi Arabia.</div></div><div><h3>Methods</h3><div>An online questionnaire was utilized in June 2024 to measure QOL using the WHOQOL-BREF scale, which covers physical health, psychological well-being, social relationships, and environmental health. Participants were grouped based on their use of walking-tracking apps and the presence of a chronic medical condition. Statistical analysis included independent t-tests, Pearson correlations, and chi-square tests to determine significant associations (p < 0.05).</div></div><div><h3>Results</h3><div>The sample consisted of 412 participants. The chi-square test revealed a significant association between having a medical condition and using a walking-tracking app (p = 0.037), with individuals without medical conditions being more likely to use these apps. However, despite the high prevalence of app usage (65.3 %), no significant improvements in QOL were observed for app users across any of the QOL domains. Participants with medical conditions reported significantly higher QOL scores in all domains, particularly in psychological health (p < 0.001) and social relationships (p = 0.001). Positive correlations were observed for factors like meaningful life, concentration, and access to healthcare among those with medical conditions.</div></div><div><h3>Conclusion</h3><div>Students with chronic medical conditions reported higher QOL whereas the use of walking-tracking apps had limited direct impact on their QOL. Future studies should explore factors that play a critical role in enhancing QOL beyond physical health and technology usage, including social support and the Saudi healthcare system.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"52 ","pages":"Article 101610"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143178369","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}
Pub Date : 2025-01-01DOI: 10.1016/j.imu.2024.101598
Vardhan Shorewala , Shivam Shorewala
{"title":"Corrigendum to “Early detection of coronary heart disease using ensemble techniques” [Inform Med Unlocked 26 (2021) 100655]","authors":"Vardhan Shorewala , Shivam Shorewala","doi":"10.1016/j.imu.2024.101598","DOIUrl":"10.1016/j.imu.2024.101598","url":null,"abstract":"","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"52 ","pages":"Article 101598"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143178836","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}
Pub Date : 2025-01-01DOI: 10.1016/j.imu.2025.101637
Nuria Lebeña , Arantza Casillas , Alicia Pérez
Background and Objective
Healthcare documentation processing is becoming more and more efficient and effective as a result of advances in machine learning and natural language processing (NLP). One challenge in clinical practice is the early detection of future patient potential diagnoses, which is crucial for preventive medicine. Estimating the potential future diagnoses, helps to speed up the management of Electronic Health Records (EHRs) and opens a path towards clinical prevention. It is a challenging task, as there are thousands of possible diseases, and, in general, there is limited data available to train systems due to privacy concerns.
The objective of his study is to infer future probable diagnoses given patients diagnosis history. In previous works, this task has been carried out using structured data, such as, ICD-coded diagnoses, overlooking unstructured textual information in EHRs. Unlike traditional methods, this study aims to enhance next-diagnosis prediction by integrating patient diagnosis information codified according to the International Classification of Diseases (ICD) with unstructured clinical text.
Methods:
We propose a multi-faceted model that integrates structured ICD-encoded patient histories with unstructured EHR text for future diagnosis prediction. Our approach consists of (1) a sequential model trained on structured diagnosis timelines, (2) a Clinical Longformer-based model trained on unstructured EHRs, and (3) an ensemble strategy to combine predictions from both components.
Results:
Our proposed ensemble strategy significantly outperforms current state-of-the-art approaches in predicting future diagnoses, achieving a Precision@5 of 72.34% and a Precision@20 of 77.49%. Additionally, it showed high robustness and reliability across different demographic groups and a varying scope of medical history.
Conclusion:
This research demonstrates that the integration of structured ICD diagnoses timelines with unstructured EHRs achieves improved results compared to just using structured diagnosis timelines. Notably, the proposed model also maintained high accuracy even with a short-term history of diagnoses.
{"title":"Large language models aided patient progression documentation according to the ICD standard","authors":"Nuria Lebeña , Arantza Casillas , Alicia Pérez","doi":"10.1016/j.imu.2025.101637","DOIUrl":"10.1016/j.imu.2025.101637","url":null,"abstract":"<div><h3>Background and Objective</h3><div>Healthcare documentation processing is becoming more and more efficient and effective as a result of advances in machine learning and natural language processing (NLP). One challenge in clinical practice is the early detection of future patient potential diagnoses, which is crucial for preventive medicine. Estimating the potential future diagnoses, helps to speed up the management of Electronic Health Records (EHRs) and opens a path towards clinical prevention. It is a challenging task, as there are thousands of possible diseases, and, in general, there is limited data available to train systems due to privacy concerns.</div><div>The objective of his study is to infer future probable diagnoses given patients diagnosis history. In previous works, this task has been carried out using structured data, such as, ICD-coded diagnoses, overlooking unstructured textual information in EHRs. Unlike traditional methods, this study aims to enhance next-diagnosis prediction by integrating patient diagnosis information codified according to the International Classification of Diseases (ICD) with unstructured clinical text.</div></div><div><h3>Methods:</h3><div>We propose a multi-faceted model that integrates structured ICD-encoded patient histories with unstructured EHR text for future diagnosis prediction. Our approach consists of (1) a sequential model trained on structured diagnosis timelines, (2) a Clinical Longformer-based model trained on unstructured EHRs, and (3) an ensemble strategy to combine predictions from both components.</div></div><div><h3>Results:</h3><div>Our proposed ensemble strategy significantly outperforms current state-of-the-art approaches in predicting future diagnoses, achieving a Precision@5 of 72.34% and a Precision@20 of 77.49%. Additionally, it showed high robustness and reliability across different demographic groups and a varying scope of medical history.</div></div><div><h3>Conclusion:</h3><div>This research demonstrates that the integration of structured ICD diagnoses timelines with unstructured EHRs achieves improved results compared to just using structured diagnosis timelines. Notably, the proposed model also maintained high accuracy even with a short-term history of diagnoses.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"55 ","pages":"Article 101637"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143739798","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}
Postmenopausal osteoporosis (PMOP) is the most prevalent metabolic bone disease among women, characterized by significant bone density loss and increased fracture risk. With a genetic component, a systematic review was conducted on the association between genetic polymorphisms and PMOP risk.
Methods
A comprehensive review of PubMed literature examined genetic polymorphisms linked to PMOP risk. The primary outcome was to identify the most frequently studied genes linked to PMOP. The secondary outcome was to perform a meta-analysis on the top genetic markers to assess their overall association with PMOP risk.
Results
Six genes, accounting for 55.08 % of all studies, were strongly associated with PMOP. Of these, the VDR gene was featured in 35 articles (18.72 % of studies), TNFRSF11B in 23 (12.30 %), ESR1 in 18 (9.63 %), COL1A1 in 12 (6.42 %), MTHFR in 8 (4.27 %), and TGFb1 in 7 (3.74 %). Meta-analysis showed five markers significantly associated with PMOP: SNP rs1544410 (ORG: 0.74 (0.59, 0.92)), SNP rs11568820 (ORG: 1.40 (1.03, 1.91)), and SNP rs2228570 (ORT: 1.39 (1.12, 1.73)) in the VDR gene; and PvuII variant (ORP: 0.80 (0.67, 0.96)) in the ESR1 gene.
Conclusion
This review strengthens the importance of conducting a robust, multi-ethnic, large cohort study with functional analysis to corroborate the findings of the six key genes associated with PMOP. Replicating these findings in larger and more diverse datasets is crucial to validate their biological relevance and potential clinical application.
{"title":"Examining the association between genetic polymorphisms and osteoporosis among post-menopausal women: a systematic review","authors":"Zainab Alhalwachi , Mira Mousa , Salsabeel Juneidi , Gabriela Restrepo-Rodas , Spyridon Karras , Habiba Alsafar , Fatme Al Anouti","doi":"10.1016/j.imu.2025.101652","DOIUrl":"10.1016/j.imu.2025.101652","url":null,"abstract":"<div><h3>Purpose</h3><div>Postmenopausal osteoporosis (PMOP) is the most prevalent metabolic bone disease among women, characterized by significant bone density loss and increased fracture risk. With a genetic component, a systematic review was conducted on the association between genetic polymorphisms and PMOP risk.</div></div><div><h3>Methods</h3><div>A comprehensive review of PubMed literature examined genetic polymorphisms linked to PMOP risk. The primary outcome was to identify the most frequently studied genes linked to PMOP. The secondary outcome was to perform a meta-analysis on the top genetic markers to assess their overall association with PMOP risk.</div></div><div><h3>Results</h3><div>Six genes, accounting for 55.08 % of all studies, were strongly associated with PMOP. Of these, the <em>VDR</em> gene was featured in 35 articles (18.72 % of studies), TNFRSF11B in 23 (12.30 %), <em>ESR1</em> in 18 (9.63 %), <em>COL1A1</em> in 12 (6.42 %), <em>MTHFR</em> in 8 (4.27 %), and TGFb1 in 7 (3.74 %). Meta-analysis showed five markers significantly associated with PMOP: SNP rs1544410 (OR<sub>G</sub>: 0.74 (0.59, 0.92)), SNP rs11568820 (OR<sub>G</sub>: 1.40 (1.03, 1.91)), and SNP rs2228570 (OR<sub>T</sub>: 1.39 (1.12, 1.73)) in the <em>VDR</em> gene; and PvuII variant (OR<sub>P</sub>: 0.80 (0.67, 0.96)) in the <em>ESR1</em> gene.</div></div><div><h3>Conclusion</h3><div>This review strengthens the importance of conducting a robust, multi-ethnic, large cohort study with functional analysis to corroborate the findings of the six key genes associated with PMOP. Replicating these findings in larger and more diverse datasets is crucial to validate their biological relevance and potential clinical application.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"56 ","pages":"Article 101652"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144070359","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Renal impairment poses a risk across all ages. With the global nephrologist shortage, the rising public health concerns over kidney failure, and advancements in technology, there is a growing need for an AI system capable of identifying kidney anomalies automatically. Chronic kidney disease is marked by a gradual failure in kidney function due to various factors, such as stones, cysts, and tumors. Chronic kidney disease often presents without noticeable symptoms initially, leading to cases remaining untreated until advanced stages. Tumors, which are dense tissue masses, can directly harm organs, including glands and spinal cells. Kidney stone disease, or urolithiasis, occurs when many solids accumulate in the urinary tract, leading to stone formation. This research paper leveraged a deep learning approach to address the worldwide shortage of urologists by facilitating the detection of kidney diseases. A novel deep learning technique is proposed using Darknet53 for the classification of kidney diseases using a large dataset gathered from five resources. The total number of images is 27,145 scans of the entire abdomen and urogram, focusing on common kidney conditions, including stones, cysts, and tumors. The data was grouped into four classes: normal, cyst, tumor, and stone. The proposed technique involves the use of 16 deep-learning models to obtain enhanced performance based on accuracy, recall, specificity, and precision, offering new potential for detecting kidney abnormalities. Model performance was evaluated, achieving 99.69 %, 0.31 %, 99.66 %, 99.88 %, 99.77 %, 0.12 %, 99.71 %, 99.60 %, and 99.17 % for accuracy, error, recall, specificity, precision, false positive rate, F1_score, Matthews Correlation Coefficient, and Kappa, respectively. Our simulation results using the Fuzzy Decision by Opinion Score Method indicated that the Darknet53 generated the best results for detecting kidney abnormalities.
{"title":"Multi-model deep learning approach for the classification of kidney diseases using medical images","authors":"Waleed Obaid , Abir Hussain , Tamer Rabie , Dhafar Hamed Abd , Wathiq Mansoor","doi":"10.1016/j.imu.2025.101663","DOIUrl":"10.1016/j.imu.2025.101663","url":null,"abstract":"<div><div>Renal impairment poses a risk across all ages. With the global nephrologist shortage, the rising public health concerns over kidney failure, and advancements in technology, there is a growing need for an AI system capable of identifying kidney anomalies automatically. Chronic kidney disease is marked by a gradual failure in kidney function due to various factors, such as stones, cysts, and tumors. Chronic kidney disease often presents without noticeable symptoms initially, leading to cases remaining untreated until advanced stages. Tumors, which are dense tissue masses, can directly harm organs, including glands and spinal cells. Kidney stone disease, or urolithiasis, occurs when many solids accumulate in the urinary tract, leading to stone formation. This research paper leveraged a deep learning approach to address the worldwide shortage of urologists by facilitating the detection of kidney diseases. A novel deep learning technique is proposed using Darknet53 for the classification of kidney diseases using a large dataset gathered from five resources. The total number of images is 27,145 scans of the entire abdomen and urogram, focusing on common kidney conditions, including stones, cysts, and tumors. The data was grouped into four classes: normal, cyst, tumor, and stone. The proposed technique involves the use of 16 deep-learning models to obtain enhanced performance based on accuracy, recall, specificity, and precision, offering new potential for detecting kidney abnormalities. Model performance was evaluated, achieving 99.69 %, 0.31 %, 99.66 %, 99.88 %, 99.77 %, 0.12 %, 99.71 %, 99.60 %, and 99.17 % for accuracy, error, recall, specificity, precision, false positive rate, F1_score, Matthews Correlation Coefficient, and Kappa, respectively. Our simulation results using the Fuzzy Decision by Opinion Score Method indicated that the Darknet53 generated the best results for detecting kidney abnormalities.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"57 ","pages":"Article 101663"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144298316","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-01DOI: 10.1016/j.imu.2025.101676
Mohaimen Al-Zubaidy , Agnieszka Stankiewicz , Matthew Anderson , Jordan Reed , Veronica Corona , Rebecca Pope , Boguslaw Obara , Maged S. Habib , David H. Steel
Objective
This review aims to identify gaps and provide direction for future research examining the use of artificial intelligence (AI) and optical coherence tomography (OCT) in the investigation and management of diabetic macular oedema (DMO).
Methods
A comprehensive literature search was conducted using MEDLINE, EMBASE, the Cochrane Central Register of Controlled Trials (CENTRAL), the Cochrane Database, and the Web of Science. The search focused on AI applications in DMO diagnosis, grading, and outcome prediction, and adhered to a predefined protocol following the Cochrane Methodology for Scoping Reviews.
Results
Following screening 40 studies were included for review. The review highlighted significant advancements in the use of AI for DMO, particularly in diagnosis and biomarker detection. AI models demonstrated high accuracy in distinguishing DMO from other retinal conditions and in segmenting key DMO biomarkers.
Conclusion
The review concludes that future research should focus on developing robust prognostic and treatment prediction models, improving external validation and standardising performance metrics. Addressing these challenges is essential for optimising the integration of AI into DMO management, ultimately improving patient outcomes and reducing vision impairment.
Significance
This review underscores AI's potential to transform DMO management, a leading cause of vision impairment in diabetes. The identified gaps and future research directions offer valuable insights for researchers and practitioners, with the potential to significantly improve patient care and healthcare efficiency.
目的探讨人工智能（AI）和光学相干断层扫描（OCT）在糖尿病性黄斑水肿（DMO）调查和治疗中的应用，为今后的研究提供方向。方法采用MEDLINE、EMBASE、Cochrane Central Register of Controlled Trials （Central）、Cochrane Database和Web of Science进行综合文献检索。检索重点关注人工智能在DMO诊断、分级和结果预测中的应用，并遵循Cochrane范围评价方法学的预定义协议。结果筛选后纳入40项研究。该综述强调了人工智能用于DMO的重大进展，特别是在诊断和生物标志物检测方面。人工智能模型在区分DMO和其他视网膜疾病以及分割关键DMO生物标志物方面表现出很高的准确性。结论本综述认为，未来的研究应侧重于建立可靠的预后和治疗预测模型，改进外部验证和标准化绩效指标。解决这些挑战对于优化人工智能与DMO管理的整合，最终改善患者的治疗效果和减少视力损害至关重要。这篇综述强调了人工智能在改变糖尿病视力损害的主要原因DMO管理方面的潜力。确定的差距和未来的研究方向为研究人员和从业人员提供了有价值的见解，有可能显著改善患者护理和医疗保健效率。
{"title":"A scoping review of the use of artificial intelligence models in automated OCT analysis and prediction of treatment outcomes in diabetic macular oedema","authors":"Mohaimen Al-Zubaidy , Agnieszka Stankiewicz , Matthew Anderson , Jordan Reed , Veronica Corona , Rebecca Pope , Boguslaw Obara , Maged S. Habib , David H. Steel","doi":"10.1016/j.imu.2025.101676","DOIUrl":"10.1016/j.imu.2025.101676","url":null,"abstract":"<div><h3>Objective</h3><div>This review aims to identify gaps and provide direction for future research examining the use of artificial intelligence (AI) and optical coherence tomography (OCT) in the investigation and management of diabetic macular oedema (DMO)<strong>.</strong></div></div><div><h3>Methods</h3><div>A comprehensive literature search was conducted using MEDLINE, EMBASE, the Cochrane Central Register of Controlled Trials (CENTRAL), the Cochrane Database, and the Web of Science. The search focused on AI applications in DMO diagnosis, grading, and outcome prediction, and adhered to a predefined protocol following the Cochrane Methodology for Scoping Reviews.</div></div><div><h3>Results</h3><div>Following screening 40 studies were included for review. The review highlighted significant advancements in the use of AI for DMO, particularly in diagnosis and biomarker detection. AI models demonstrated high accuracy in distinguishing DMO from other retinal conditions and in segmenting key DMO biomarkers.</div></div><div><h3>Conclusion</h3><div>The review concludes that future research should focus on developing robust prognostic and treatment prediction models, improving external validation and standardising performance metrics. Addressing these challenges is essential for optimising the integration of AI into DMO management, ultimately improving patient outcomes and reducing vision impairment.</div></div><div><h3>Significance</h3><div>This review underscores AI's potential to transform DMO management, a leading cause of vision impairment in diabetes. The identified gaps and future research directions offer valuable insights for researchers and practitioners, with the potential to significantly improve patient care and healthcare efficiency.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"57 ","pages":"Article 101676"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144749341","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The application of artificial intelligence in oncology has been limited by its reliance on large, annotated datasets and the need for retraining models for domain-specific diagnostic tasks. Taking heed of these limitations, we investigated in-context learning as a pragmatic alternative to model retraining by allowing models to adapt to new diagnostic tasks using only a few labeled examples at inference, without the need for retraining. Using four vision-language models (VLMs) -- Paligemma, CLIP, ALIGN and GPT-4o, we evaluated the performance across three oncology datasets: MHIST, PatchCamelyon and HAM10000. To the best of our knowledge, this is the first study to compare the performance of multiple VLMs with in-context learning on different oncology classification tasks. Without any parameter updates, all models showed significant gains with few-shot prompting, with GPT-4o reaching an F1 score of 0.81 in binary classification and 0.60 in multi-class classification settings. While these results remain below the ceiling of fully fine-tuned systems, they highlight the potential of ICL to approximate task-specific behavior using only a handful of examples, reflecting how clinicians often reason from prior cases. Notably, open-source models like Paligemma and CLIP demonstrated competitive gains despite their smaller size, suggesting feasibility for deployment in computing constrained clinical environments. Overall, these findings highlight the potential of ICL as a practical solution in oncology, particularly for rare cancers and resource-limited contexts where fine-tuning is infeasible and annotated data is difficult to obtain.
{"title":"In-context learning for label-efficient cancer image classification in oncology","authors":"Mobina Shrestha , Bishwas Mandal , Vishal Mandal , Asis Shrestha , Amir Babu Shrestha","doi":"10.1016/j.imu.2025.101683","DOIUrl":"10.1016/j.imu.2025.101683","url":null,"abstract":"<div><div>The application of artificial intelligence in oncology has been limited by its reliance on large, annotated datasets and the need for retraining models for domain-specific diagnostic tasks. Taking heed of these limitations, we investigated in-context learning as a pragmatic alternative to model retraining by allowing models to adapt to new diagnostic tasks using only a few labeled examples at inference, without the need for retraining. Using four vision-language models (VLMs) -- Paligemma, CLIP, ALIGN and GPT-4o, we evaluated the performance across three oncology datasets: MHIST, PatchCamelyon and HAM10000. To the best of our knowledge, this is the first study to compare the performance of multiple VLMs with in-context learning on different oncology classification tasks. Without any parameter updates, all models showed significant gains with few-shot prompting, with GPT-4o reaching an F1 score of 0.81 in binary classification and 0.60 in multi-class classification settings. While these results remain below the ceiling of fully fine-tuned systems, they highlight the potential of ICL to approximate task-specific behavior using only a handful of examples, reflecting how clinicians often reason from prior cases. Notably, open-source models like Paligemma and CLIP demonstrated competitive gains despite their smaller size, suggesting feasibility for deployment in computing constrained clinical environments. Overall, these findings highlight the potential of ICL as a practical solution in oncology, particularly for rare cancers and resource-limited contexts where fine-tuning is infeasible and annotated data is difficult to obtain.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"58 ","pages":"Article 101683"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144887256","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-01DOI: 10.1016/j.imu.2025.101688
Sreebhadra Vallukappully , Ian van der Linde , Ashim Chakraborty
Diabetic Retinopathy (DR) is a progressive eye disease that affects those with long-term diabetes. It can lead to irreversible blindness if not detected and treated early. Early detection is challenging as changes to the retina are initially subtle. A number of computational models have been proposed to detect DR in fundus images, including in its early stages. Here, a novel transfer learning approach is proposed using the NASNet-Large and ResNet-50 convolutional neural networks. Image pre-processing steps are tested combinatorically. Class imbalance is addressed with oversampling and data augmentation to give trustworthy performance metrics. The models give impressive detection rates using a standard dataset containing expert-labelled DR fundus images (APTOS 2019), with the best performing models giving accuracy in classifying unseen images exceeding 0.96 (F1 score 0.97) for Early-stage DR detection (no DR vs mild and moderate), and over 0.91 accuracy (F1 score 0.91) for Multi-stage classification (no DR, mild, moderate, severe, and proliferative). This work highlights the potential of combining the transfer learning of state-of-the-art deep learning models with classical image processing for effective DR detection and classification.
{"title":"Early detection and classification of diabetic retinopathy by transfer learning of NASNet-large and ResNet-50 convolutional neural networks","authors":"Sreebhadra Vallukappully , Ian van der Linde , Ashim Chakraborty","doi":"10.1016/j.imu.2025.101688","DOIUrl":"10.1016/j.imu.2025.101688","url":null,"abstract":"<div><div>Diabetic Retinopathy (DR) is a progressive eye disease that affects those with long-term diabetes. It can lead to irreversible blindness if not detected and treated early. Early detection is challenging as changes to the retina are initially subtle. A number of computational models have been proposed to detect DR in fundus images, including in its early stages. Here, a novel transfer learning approach is proposed using the NASNet-Large and ResNet-50 convolutional neural networks. Image pre-processing steps are tested combinatorically. Class imbalance is addressed with oversampling and data augmentation to give trustworthy performance metrics. The models give impressive detection rates using a standard dataset containing expert-labelled DR fundus images (APTOS 2019), with the best performing models giving accuracy in classifying unseen images exceeding 0.96 (F1 score 0.97) for Early-stage DR detection (no DR vs mild and moderate), and over 0.91 accuracy (F1 score 0.91) for Multi-stage classification (no DR, mild, moderate, severe, and proliferative). This work highlights the potential of combining the transfer learning of state-of-the-art deep learning models with classical image processing for effective DR detection and classification.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"58 ","pages":"Article 101688"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145010545","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-01DOI: 10.1016/j.imu.2025.101691
Allan G. Duah , Roland V. Bumbuc , H. Ibrahim Korkmaz , Rory Wilding , Vivek M. Sheraton
Medical data, hospital patient-specific data, are highly sensitive to privacy and are essential for research in the biomedical field. Although there are many new approaches to creating databases that ensure data must be FAIR and GDPR compliant, these approaches require the intervention of secured data handlers. To address this gap, this study investigates and designs a standardized Federated Learning (FL) architecture for medical data. Specifically, we examine traditional and novel methods for preprocessing, handling, and utilizing such data in FL. We develop “FedDeepInsight”, a novel data transformation framework that enables tabular data augmentation and transformation into image data prior to neural network training and FL. Additionally, we analyze how the type of dataset influences the performance of federated learning algorithms and machine learning models in terms of accuracy and efficiency. Our results indicate that FedAvg is the most reliable aggregation algorithm, providing superior accuracy, stability, and convergence, and FedYogi is also viable with well-tuned hyperparameters. For privacy protection, we recommend Differential Privacy (DP) with calibrated noise multipliers and initial upper and lower bounds for stability. Ultimately, we emerge as a promising solution for secure, privacy-preserving federation learning in healthcare.
{"title":"FedDeepInsight—A privacy-first federated learning architecture for medical data","authors":"Allan G. Duah , Roland V. Bumbuc , H. Ibrahim Korkmaz , Rory Wilding , Vivek M. Sheraton","doi":"10.1016/j.imu.2025.101691","DOIUrl":"10.1016/j.imu.2025.101691","url":null,"abstract":"<div><div>Medical data, hospital patient-specific data, are highly sensitive to privacy and are essential for research in the biomedical field. Although there are many new approaches to creating databases that ensure data must be FAIR and GDPR compliant, these approaches require the intervention of secured data handlers. To address this gap, this study investigates and designs a standardized Federated Learning (FL) architecture for medical data. Specifically, we examine traditional and novel methods for preprocessing, handling, and utilizing such data in FL. We develop “FedDeepInsight”, a novel data transformation framework that enables tabular data augmentation and transformation into image data prior to neural network training and FL. Additionally, we analyze how the type of dataset influences the performance of federated learning algorithms and machine learning models in terms of accuracy and efficiency. Our results indicate that FedAvg is the most reliable aggregation algorithm, providing superior accuracy, stability, and convergence, and FedYogi is also viable with well-tuned hyperparameters. For privacy protection, we recommend Differential Privacy (DP) with calibrated noise multipliers and initial upper and lower bounds for stability. Ultimately, we emerge as a promising solution for secure, privacy-preserving federation learning in healthcare.</div></div>","PeriodicalId":13953,"journal":{"name":"Informatics in Medicine Unlocked","volume":"58 ","pages":"Article 101691"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145059881","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号