This review presents and discusses the ways in which artificial intelligence (AI) tools currently intervene, or could potentially intervene in the future, to enhance the diverse tasks involved in the radiotherapy workflow. The radiotherapy framework is presented on two different levels for the personalization of the treatment, distinct in tasks and methodologies. The first level is the clinically well-established anatomy-based workflow, known as adaptive radiation therapy (ART). The second level is referred to as biology-driven workflow, explored in research literature and recently appearing in some preliminary clinical trials for personalized radiation treatments. A twofold role for AI is defined according to these two different levels. In the anatomy-based workflow, the role of AI is to streamline and improve the tasks in terms of time and variability reductions compared to conventional methodologies. The biology-driven workflow instead fully relies on AI, which introduces decision-making tools opening uncharted frontiers which were in the past deemed challenging to explore. These methodologies are referred to as radiomics and dosiomics, handling imaging and dosimetric information, or to multiomics, when complemented by clinical and biological parameters (i.e., biomarkers). The review explicitly highlights the methodologies that are currently incorporated into clinical practice or still in research, with the aim of presenting the AI’s growing role in personalized radiotherapy.
{"title":"“Under the hood”: artificial intelligence in personalized radiotherapy","authors":"C. Gianoli, Elisabetta De Bernardi, Katia Parodi","doi":"10.1093/bjro/tzae017","DOIUrl":"https://doi.org/10.1093/bjro/tzae017","url":null,"abstract":"\u0000 This review presents and discusses the ways in which artificial intelligence (AI) tools currently intervene, or could potentially intervene in the future, to enhance the diverse tasks involved in the radiotherapy workflow. The radiotherapy framework is presented on two different levels for the personalization of the treatment, distinct in tasks and methodologies. The first level is the clinically well-established anatomy-based workflow, known as adaptive radiation therapy (ART). The second level is referred to as biology-driven workflow, explored in research literature and recently appearing in some preliminary clinical trials for personalized radiation treatments. A twofold role for AI is defined according to these two different levels. In the anatomy-based workflow, the role of AI is to streamline and improve the tasks in terms of time and variability reductions compared to conventional methodologies. The biology-driven workflow instead fully relies on AI, which introduces decision-making tools opening uncharted frontiers which were in the past deemed challenging to explore. These methodologies are referred to as radiomics and dosiomics, handling imaging and dosimetric information, or to multiomics, when complemented by clinical and biological parameters (i.e., biomarkers). The review explicitly highlights the methodologies that are currently incorporated into clinical practice or still in research, with the aim of presenting the AI’s growing role in personalized radiotherapy.","PeriodicalId":516126,"journal":{"name":"BJR|Open","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141642812","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}
Di Jin, Zicong li, Zhikang Deng, Jiayu Nan, Pei Huang, Bingliang Zeng, Bing Fan
To efficiently use medical resources and offer optimal personalized treatment for individuals with Omicron infection, it's vital to predict the disease's outcome early on. This research developed three machine learning models to foresee the results for Omicron-infected patients. Data from 253 Omicron-infected patients, including their CT scans, clinical details, and relevant laboratory values, were studied. The patients were categorized into two groups based on their disease progression: favorable prognosis and unfavorable prognosis. Patients manifesting respiratory failure, acute liver or kidney impairment, or fatalities were placed in the “poor” group. Those lacking such symptoms were allocated to the “good” group. The participants were randomly split into training set (202) and validation set (51) with an 8:2 ratio. Radiomics features were produced using image processing, focused segmentation, feature extraction, and selection, leading to the establishment of a radiomics model. A univariate logistic regression method identified potential clinical factors contributing to a clinical model's development. Eventually, the fused feature set, integrating radiomics features and clinical indicators, was used for the combined model. The model's prediction performance was assessed using the area under the receiver operating characteristic curve (AUC). The model's clinical usefulness was evaluated by generating calibration and decision curves. Compared to other classification models, the combined model showcased the best classification performance. It achieved an AUC of 0.848 and accuracy of 0.763 in the training set, and 0.797 and 0.750 in the validation set, respectively. This study employed machine learning model to accurately predict the prognosis of Omicron-infected patients. (1) Topic innovation: At present, there is a lack of research on the use of CT images to construct machine learning models to predict the prognosis of patients with Omicjon infection. This study intends to establish clinical, radiomics and combined models to provide more possibilities for the identification of the two. (2) Platform innovation: The feature extraction and screening and the establishment of omics model in this study will be completed in the intelligent scientific research platform, which can reduce the error caused by human error, simplify the operation steps and save the time of data processing time.
{"title":"Combined with the semantic features of CT and selected clinical variables, a machine learning model for accurately predicting the prognosis of omicron was established","authors":"Di Jin, Zicong li, Zhikang Deng, Jiayu Nan, Pei Huang, Bingliang Zeng, Bing Fan","doi":"10.1093/bjro/tzae013","DOIUrl":"https://doi.org/10.1093/bjro/tzae013","url":null,"abstract":"\u0000 \u0000 \u0000 To efficiently use medical resources and offer optimal personalized treatment for individuals with Omicron infection, it's vital to predict the disease's outcome early on. This research developed three machine learning models to foresee the results for Omicron-infected patients.\u0000 \u0000 \u0000 \u0000 Data from 253 Omicron-infected patients, including their CT scans, clinical details, and relevant laboratory values, were studied. The patients were categorized into two groups based on their disease progression: favorable prognosis and unfavorable prognosis. Patients manifesting respiratory failure, acute liver or kidney impairment, or fatalities were placed in the “poor” group. Those lacking such symptoms were allocated to the “good” group. The participants were randomly split into training set (202) and validation set (51) with an 8:2 ratio. Radiomics features were produced using image processing, focused segmentation, feature extraction, and selection, leading to the establishment of a radiomics model. A univariate logistic regression method identified potential clinical factors contributing to a clinical model's development. Eventually, the fused feature set, integrating radiomics features and clinical indicators, was used for the combined model. The model's prediction performance was assessed using the area under the receiver operating characteristic curve (AUC). The model's clinical usefulness was evaluated by generating calibration and decision curves.\u0000 \u0000 \u0000 \u0000 Compared to other classification models, the combined model showcased the best classification performance. It achieved an AUC of 0.848 and accuracy of 0.763 in the training set, and 0.797 and 0.750 in the validation set, respectively.\u0000 \u0000 \u0000 \u0000 This study employed machine learning model to accurately predict the prognosis of Omicron-infected patients.\u0000 \u0000 \u0000 \u0000 (1) Topic innovation: At present, there is a lack of research on the use of CT images to construct machine learning models to predict the prognosis of patients with Omicjon infection. This study intends to establish clinical, radiomics and combined models to provide more possibilities for the identification of the two. (2) Platform innovation: The feature extraction and screening and the establishment of omics model in this study will be completed in the intelligent scientific research platform, which can reduce the error caused by human error, simplify the operation steps and save the time of data processing time.\u0000","PeriodicalId":516126,"journal":{"name":"BJR|Open","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141383314","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}
Christopher Thomas, Isabel Dregely, I. Oksuz, Teresa Guerrero Urbano, T. Greener, Andrew P King, Sally F Barrington
Toxicity-driven adaptive radiotherapy (RT) is enhanced by the superior soft tissue contrast of magnetic resonance (MR) imaging compared with conventional computed tomography (CT). However, in an MR-only RT pathway synthetic CTs (sCT) are required for dose calculation. This study evaluates 3 sCT approaches for accurate rectal toxicity prediction in prostate RT. Thirty-six patients had MR (T2-weighted acquisition optimised for anatomical delineation, and T1-Dixon) with same day standard-of-care planning CT for prostate RT. Multiple sCT were created per patient using bulk density (BD), tissue stratification (TS, from T1-Dixon) and deep-learning (DL) artificial intelligence (AI) (from T2-weighted) approaches for dose distribution calculation and creation of rectal dose volume histograms (DVH) and dose surface maps (DSM) to assess grade-2 (G2) rectal bleeding risk. Maximum absolute errors using sCT for DVH-based G2 rectal bleeding risk (risk range 1.6% to 6.1%) were 0.6% (BD), 0.3% (TS) and 0.1% (DL). DSM-derived risk prediction errors followed a similar pattern. DL sCT has voxel-wise density generated from T2-weighted MR and improved accuracy for both risk-prediction methods. DL improves dosimetric and predicted risk calculation accuracy. Both TS and DL methods are clinically suitable for sCT generation in toxicity-guided RT, however DL offers increased accuracy and offers efficiencies by removing the need for T1-Dixon MR. This study demonstrates novel insights regarding the effect of sCT on predictive toxicity metrics, demonstrating clear accuracy improvement with increased sCT resolution. Accuracy of toxicity calculation in MR-only RT should be assessed for all treatment sites where dose to critical structures will guide adaptive-RT strategies.
{"title":"Effect of synthetic CT on dose-derived toxicity predictors for MR-only prostate radiotherapy","authors":"Christopher Thomas, Isabel Dregely, I. Oksuz, Teresa Guerrero Urbano, T. Greener, Andrew P King, Sally F Barrington","doi":"10.1093/bjro/tzae014","DOIUrl":"https://doi.org/10.1093/bjro/tzae014","url":null,"abstract":"\u0000 \u0000 \u0000 Toxicity-driven adaptive radiotherapy (RT) is enhanced by the superior soft tissue contrast of magnetic resonance (MR) imaging compared with conventional computed tomography (CT). However, in an MR-only RT pathway synthetic CTs (sCT) are required for dose calculation. This study evaluates 3 sCT approaches for accurate rectal toxicity prediction in prostate RT.\u0000 \u0000 \u0000 \u0000 Thirty-six patients had MR (T2-weighted acquisition optimised for anatomical delineation, and T1-Dixon) with same day standard-of-care planning CT for prostate RT. Multiple sCT were created per patient using bulk density (BD), tissue stratification (TS, from T1-Dixon) and deep-learning (DL) artificial intelligence (AI) (from T2-weighted) approaches for dose distribution calculation and creation of rectal dose volume histograms (DVH) and dose surface maps (DSM) to assess grade-2 (G2) rectal bleeding risk.\u0000 \u0000 \u0000 \u0000 Maximum absolute errors using sCT for DVH-based G2 rectal bleeding risk (risk range 1.6% to 6.1%) were 0.6% (BD), 0.3% (TS) and 0.1% (DL). DSM-derived risk prediction errors followed a similar pattern. DL sCT has voxel-wise density generated from T2-weighted MR and improved accuracy for both risk-prediction methods.\u0000 \u0000 \u0000 \u0000 DL improves dosimetric and predicted risk calculation accuracy. Both TS and DL methods are clinically suitable for sCT generation in toxicity-guided RT, however DL offers increased accuracy and offers efficiencies by removing the need for T1-Dixon MR.\u0000 \u0000 \u0000 \u0000 This study demonstrates novel insights regarding the effect of sCT on predictive toxicity metrics, demonstrating clear accuracy improvement with increased sCT resolution. Accuracy of toxicity calculation in MR-only RT should be assessed for all treatment sites where dose to critical structures will guide adaptive-RT strategies.\u0000","PeriodicalId":516126,"journal":{"name":"BJR|Open","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141269609","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}
{"title":"Celebrating five years of BJR|Open","authors":"Katja Pinker, Habib Zaidi","doi":"10.1093/bjro/tzae009","DOIUrl":"https://doi.org/10.1093/bjro/tzae009","url":null,"abstract":"","PeriodicalId":516126,"journal":{"name":"BJR|Open","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-05-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140993783","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}
Rikke Bachmann, Gozde Gunes, Stine Hangaard, Andreas Nexmann, P. Lisouski, Mikael Boesen, Michael Lundemann, Scott G Baginski
The aim of this study was to evaluate the diagnostic performance of non-specialist readers with and without the use of an AI support tool to detect traumatic fractures on radiographs of the appendicular skeleton. The design was a retrospective, fully-crossed multi-reader, multi-case study on a balanced dataset of patients (≥2 years of age) with an AI tool as a diagnostic intervention. Fifteen readers assessed 340 radiographic exams, with and without the AI tool in two different sessions and the time spent was automatically recorded. Reference standard was established by three consultant radiologists. Sensitivity, specificity, and false positives per patient were calculated. Patient-wise sensitivity increased from 72% to 80% (p < 0.05) and patient-wise specificity increased from 81% to 85% (p < 0.05) in exams aided by the AI tool compared to the unaided exams. The increase in sensitivity resulted in a relative reduction of missed fractures of 29%. The average rate of false positives per patient decreased from 0.16 to 0.14, corresponding to a relative reduction of 21%. There was no significant difference in average reading time spent per exam. The largest gain in fracture detection performance, with AI support, across all readers, was on non-obvious fractures with a significant increase in sensitivity of 11 percentage points (60% to 71%). The diagnostic performance for detection of traumatic fractures on radiographs of the appendicular skeleton improved among non-specialist readers tested AI fracture detection support tool showed an overall reader improvement in sensitivity and specificity when supported by an AI tool. Improvement was seen in both sensitivity and specificity and without negatively affecting the interpretation time. The division and analysis of obvious and non-obvious fractures are novel in AI reader comparison studies like this.
{"title":"Improving Traumatic Fracture Detection on Radiographs with Artificial Intelligence Support: A Multi-Reader Study","authors":"Rikke Bachmann, Gozde Gunes, Stine Hangaard, Andreas Nexmann, P. Lisouski, Mikael Boesen, Michael Lundemann, Scott G Baginski","doi":"10.1093/bjro/tzae011","DOIUrl":"https://doi.org/10.1093/bjro/tzae011","url":null,"abstract":"\u0000 \u0000 \u0000 The aim of this study was to evaluate the diagnostic performance of non-specialist readers with and without the use of an AI support tool to detect traumatic fractures on radiographs of the appendicular skeleton.\u0000 \u0000 \u0000 \u0000 The design was a retrospective, fully-crossed multi-reader, multi-case study on a balanced dataset of patients (≥2 years of age) with an AI tool as a diagnostic intervention. Fifteen readers assessed 340 radiographic exams, with and without the AI tool in two different sessions and the time spent was automatically recorded. Reference standard was established by three consultant radiologists. Sensitivity, specificity, and false positives per patient were calculated.\u0000 \u0000 \u0000 \u0000 Patient-wise sensitivity increased from 72% to 80% (p < 0.05) and patient-wise specificity increased from 81% to 85% (p < 0.05) in exams aided by the AI tool compared to the unaided exams. The increase in sensitivity resulted in a relative reduction of missed fractures of 29%. The average rate of false positives per patient decreased from 0.16 to 0.14, corresponding to a relative reduction of 21%. There was no significant difference in average reading time spent per exam. The largest gain in fracture detection performance, with AI support, across all readers, was on non-obvious fractures with a significant increase in sensitivity of 11 percentage points (60% to 71%).\u0000 \u0000 \u0000 \u0000 The diagnostic performance for detection of traumatic fractures on radiographs of the appendicular skeleton improved among non-specialist readers tested AI fracture detection support tool showed an overall reader improvement in sensitivity and specificity when supported by an AI tool. Improvement was seen in both sensitivity and specificity and without negatively affecting the interpretation time.\u0000 \u0000 \u0000 \u0000 The division and analysis of obvious and non-obvious fractures are novel in AI reader comparison studies like this.\u0000","PeriodicalId":516126,"journal":{"name":"BJR|Open","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-04-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140654997","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}
Maud E P Rijkx, Esther M Heuts, J. Houwers, J. Hommes, Andrzej A Piatkowski, T. V. van Nijnatten
Autologous fat transfer (AFT) is an upcoming technique for total breast reconstruction. Consequently, radiological imaging of women with an AFT reconstructed breast will increase in the coming years, yet radiological experience and evidence after AFT is limited. The surgical procedure of AFT and follow-up with imaging modalities including mammography (MG), ultrasound (US), and magnetic resonance imaging (MRI) in patients with a total breast reconstruction with AFT are summarized to illustrate the radiological normal and suspicious findings for malignancy. Imaging after a total breast reconstruction with AFT appears to be based mostly on benign imaging findings with an overall low biopsy rate. As higher volumes are injected in this technique, the risk for the onset of fat necrosis increases. Imaging findings most often are related to fat necrosis after AFT. On MG, fat necrosis can mostly be seen as oil cysts. Breast seromas after total breast reconstruction with AFT is an unfavourable outcome and may require special treatment. Fat deposition in the pectoral muscle is a previously unknown, but benign entity. Although fat necrosis is a benign entity, it can mimic breast cancer (recurrence). In symptomatic women after total breast reconstruction with AFT, MG and US can be considered as first diagnostic modalities. Breast MRI can be used as a problem-solving tool during later stage. Future studies should investigate the most optimal follow-up strategy, including different imaging modalities, in patients treated with AFT for total breast reconstruction.
{"title":"Imaging findings after a total reconstructed breast with autologous fat transfer (AFT): what the radiologist needs to know","authors":"Maud E P Rijkx, Esther M Heuts, J. Houwers, J. Hommes, Andrzej A Piatkowski, T. V. van Nijnatten","doi":"10.1093/bjro/tzae010","DOIUrl":"https://doi.org/10.1093/bjro/tzae010","url":null,"abstract":"\u0000 \u0000 \u0000 Autologous fat transfer (AFT) is an upcoming technique for total breast reconstruction. Consequently, radiological imaging of women with an AFT reconstructed breast will increase in the coming years, yet radiological experience and evidence after AFT is limited.\u0000 \u0000 \u0000 \u0000 The surgical procedure of AFT and follow-up with imaging modalities including mammography (MG), ultrasound (US), and magnetic resonance imaging (MRI) in patients with a total breast reconstruction with AFT are summarized to illustrate the radiological normal and suspicious findings for malignancy.\u0000 \u0000 \u0000 \u0000 Imaging after a total breast reconstruction with AFT appears to be based mostly on benign imaging findings with an overall low biopsy rate. As higher volumes are injected in this technique, the risk for the onset of fat necrosis increases. Imaging findings most often are related to fat necrosis after AFT. On MG, fat necrosis can mostly be seen as oil cysts. Breast seromas after total breast reconstruction with AFT is an unfavourable outcome and may require special treatment. Fat deposition in the pectoral muscle is a previously unknown, but benign entity. Although fat necrosis is a benign entity, it can mimic breast cancer (recurrence).\u0000 \u0000 \u0000 \u0000 In symptomatic women after total breast reconstruction with AFT, MG and US can be considered as first diagnostic modalities. Breast MRI can be used as a problem-solving tool during later stage. Future studies should investigate the most optimal follow-up strategy, including different imaging modalities, in patients treated with AFT for total breast reconstruction.\u0000","PeriodicalId":516126,"journal":{"name":"BJR|Open","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140659673","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}
Francesca De Luca, Thröstur Finnbogason, Ola Kvist
MRI is an emerging imaging modality to assess skeletal maturity. This study aimed to chart the learning curves of paediatric radiologists when using an unfamiliar MRI grading system of skeletal maturity and to assess the clinical feasibility of implementing said system. 958 healthy paediatric volunteers were prospectively included in a dual-facility study. Each subject underwent a conventional MRI scan at 1.5 T. To perform the image reading, the participants were grouped into five subsets (subsets 1 to 5) of equal size (n∼192) in chronological order for scan acquisition. Two paediatric radiologists (R1–2) with different levels of MRI experience, both of whom were previously unfamiliar with the study’s MRI grading system, independently evaluated the subsets to assess skeletal maturity in five different growth plate locations. Congruent cases at blinded reading established the consensus reading. For discrepant cases, the consensus reading was obtained through an unblinded reading by a third paediatric radiologist (R3), also unfamiliar with the MRI grading system. Further, R1 performed a second blinded image reading for all included subjects with a memory wash-out of 180 days. Weighted Cohen’s kappa was used to assess interreader reliability (R1 vs consensus; R2 vs consensus) at non-cumulative and cumulative time points, as well as interreader (R1 vs R2) and intrareader (R1 vs R1) reliability at non-cumulative time points. Mean weighted Cohen’s kappa values for each pair of blinded readers compared to consensus reading (interreader reliability, R1–2 vs consensus) were ≥0.85, showing a strong to almost perfect interreader agreement at both non-cumulative and cumulative time points and in all growth plate locations. Weighted Cohen’s kappa values for interreader (R1 vs R2) and intrareader reliability (R1 vs R1) were ≥0.72 at non-cumulative time points, with values ≥ 0.82 at subset 5. Paediatric radiologists’ clinical confidence when introduced to a new MRI grading system for skeletal maturity was high from the outset of their learning curve, despite the radiologists’ varying levels of work experience with MRI assessment. The MRI grading system for skeletal maturity investigated in this study is a robust clinical method when used by paediatric radiologists and can be used in clinical practice. Radiologists with fellowship training in paediatric radiology experienced no learning curve progress when introduced to a new MRI grading system for skeletal maturity and achieved desirable agreement from the first time point of the learning curve. The robustness of the investigated MRI grading system was not affected by the earlier different levels of MRI experience among the readers.
磁共振成像是一种新兴的评估骨骼成熟度的成像模式。本研究旨在绘制儿科放射医师在使用陌生的骨骼成熟度核磁共振成像分级系统时的学习曲线图，并评估实施该系统的临床可行性。 在一项双机构研究中，958 名健康的儿科志愿者参与了前瞻性研究。每位受试者都接受了 1.5 T 的常规磁共振成像扫描。为了进行图像读取，受试者按扫描时间顺序被分成五个人数相等的子组（子组 1 至 5）（n∼192）。两名具有不同磁共振成像经验的儿科放射科医生（R1-2）独立评估子集，评估五个不同生长板位置的骨骼成熟度。在盲读时，一致的病例确定为共识读数。对于不一致的病例，则由同样不熟悉磁共振成像分级系统的第三位儿科放射科医生（R3）进行非盲读，以获得共识读数。此外，R1 对所有纳入的受试者进行了第二次盲法图像判读，并进行了 180 天的记忆冲洗。加权科恩卡帕用于评估非累积和累积时间点的读片者间可靠性（R1 vs 共识；R2 vs 共识），以及非累积时间点的读片者间（R1 vs R2）和读片者内（R1 vs R1）可靠性。 与共识读数（读数间可靠性，R1-2 vs 共识读数）相比，每对盲人读数的平均加权科恩卡帕值均≥0.85，表明在非累积和累积时间点以及所有生长板位置，读数间的一致性很强，几乎达到完美。在非累积时间点，读片者之间（R1 vs R2）和读片者内部（R1 vs R1）的加权科恩卡帕值均≥0.72，在子集 5 中的值≥0.82。 尽管放射科医生在核磁共振成像评估方面的工作经验各不相同，但儿科放射科医生在学习新的骨骼成熟度核磁共振成像分级系统之初就有很高的临床信心。本研究调查的骨骼成熟度核磁共振成像分级系统在儿科放射医师使用时是一种可靠的临床方法，可用于临床实践。 接受过儿科放射学研究培训的放射科医生在学习新的骨骼成熟度核磁共振成像分级系统时没有经历学习曲线的变化，并在学习曲线的第一个时间点就达到了理想的一致性。所研究的磁共振成像分级系统的稳健性并没有受到早期不同磁共振成像经验水平的读者的影响。
{"title":"Specialist learning curves and clinical feasibility of introducing a new MRI grading system for skeletal maturity","authors":"Francesca De Luca, Thröstur Finnbogason, Ola Kvist","doi":"10.1093/bjro/tzae008","DOIUrl":"https://doi.org/10.1093/bjro/tzae008","url":null,"abstract":"\u0000 \u0000 \u0000 MRI is an emerging imaging modality to assess skeletal maturity. This study aimed to chart the learning curves of paediatric radiologists when using an unfamiliar MRI grading system of skeletal maturity and to assess the clinical feasibility of implementing said system.\u0000 \u0000 \u0000 \u0000 958 healthy paediatric volunteers were prospectively included in a dual-facility study. Each subject underwent a conventional MRI scan at 1.5 T. To perform the image reading, the participants were grouped into five subsets (subsets 1 to 5) of equal size (n∼192) in chronological order for scan acquisition. Two paediatric radiologists (R1–2) with different levels of MRI experience, both of whom were previously unfamiliar with the study’s MRI grading system, independently evaluated the subsets to assess skeletal maturity in five different growth plate locations. Congruent cases at blinded reading established the consensus reading. For discrepant cases, the consensus reading was obtained through an unblinded reading by a third paediatric radiologist (R3), also unfamiliar with the MRI grading system. Further, R1 performed a second blinded image reading for all included subjects with a memory wash-out of 180 days. Weighted Cohen’s kappa was used to assess interreader reliability (R1 vs consensus; R2 vs consensus) at non-cumulative and cumulative time points, as well as interreader (R1 vs R2) and intrareader (R1 vs R1) reliability at non-cumulative time points.\u0000 \u0000 \u0000 \u0000 Mean weighted Cohen’s kappa values for each pair of blinded readers compared to consensus reading (interreader reliability, R1–2 vs consensus) were ≥0.85, showing a strong to almost perfect interreader agreement at both non-cumulative and cumulative time points and in all growth plate locations. Weighted Cohen’s kappa values for interreader (R1 vs R2) and intrareader reliability (R1 vs R1) were ≥0.72 at non-cumulative time points, with values ≥ 0.82 at subset 5.\u0000 \u0000 \u0000 \u0000 Paediatric radiologists’ clinical confidence when introduced to a new MRI grading system for skeletal maturity was high from the outset of their learning curve, despite the radiologists’ varying levels of work experience with MRI assessment. The MRI grading system for skeletal maturity investigated in this study is a robust clinical method when used by paediatric radiologists and can be used in clinical practice.\u0000 \u0000 \u0000 \u0000 Radiologists with fellowship training in paediatric radiology experienced no learning curve progress when introduced to a new MRI grading system for skeletal maturity and achieved desirable agreement from the first time point of the learning curve. The robustness of the investigated MRI grading system was not affected by the earlier different levels of MRI experience among the readers.\u0000","PeriodicalId":516126,"journal":{"name":"BJR|Open","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140718502","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}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：