Sahil Sethi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Sahil Sethi has a total of 30 co-authors. Here are the top co-authors by number of shared publications:
- Anthony Christiano
Department of Orthopaedic Surgery and Rehabilitation Medicine University of Chicago Chicago Illinois USA
1 shared publication Brooke R. D’aRcy
Department of Molecular Genetics and Microbiology, Duke University School of Medicine, Durham, NC 27710, USA
1 shared publicationCamila M. Musso
Department of Molecular Genetics and Microbiology, Duke University School of Medicine, Durham, NC 27710, USA
1 shared publicationDarlington Nwaudo
Department of Orthopaedic Surgery and Rehabilitation Medicine, University of Chicago Medicine, Chicago, Illinois, USA
1 shared publication- Debra L. Silver
Department of Molecular Genetics and Microbiology, Duke University School of Medicine, Durham, North Carolina, United States of America
1 shared publication Devin K. Hubbard
UNC/NCSU Joint Department of Biomedical Engineering
1 shared publication
Latest Publications
- 1 July 2026
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 13 (04) , 1-11
Beyond accuracy: a systematic scoping review of algorithmic bias measurement and mitigation in peer-reviewed publications in medical imaging AI development and evaluation for classification
PurposeThe rapid development of medical imaging artificial intelligence (AI) models has raised concerns over ensuring that the outputs of these models, when deployed in clinical environments, provide unbiased classifications across different patient populations or operational factors. This scoping review aimed to investigate how the current literature evaluates algorithmic bias in medical imaging AI models using quantitative fairness metrics.ApproachFollowing the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR), PubMed and Embase were systematically searched for peer-reviewed literature published between January 2015 and May 2025. Studies that evaluated classification medical imaging AI models for algorithmic bias and reported at least one quantitative measure of fairness were included. Articles that investigated fairness in nonimaging AI models used qualitative bias mitigation methods, examined technical bias (e.g., systematic measurement or preprocessing), or were conference proceedings were excluded.ResultsAfter removing duplicates, 2557 studies were included for title and abstract screening. Following initial screening, 143 articles were retrieved for full-text review. Of these, eight studies were eligible for inclusion. Half of the articles focused on bias detection, whereas the other half investigated mitigation techniques. There was considerable heterogeneity in the bias metrics used across the studies. Nearly all studies (7/8) used a chest X-ray radiography classifier for evaluation and examined algorithmic bias with respect to gender, age, and race.ConclusionsThe relative paucity of existing literature and variability in fairness evaluation methods underscore the need for more standardized frameworks to evaluate and ensure fairness in imaging AI systems.
- 1 July 2026
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 13 (04) , 1-11
Beyond accuracy: a systematic scoping review of algorithmic bias measurement and mitigation in peer-reviewed publications in medical imaging AI development and evaluation for classification
PurposeThe rapid development of medical imaging artificial intelligence (AI) models has raised concerns over ensuring that the outputs of these models, when deployed in clinical environments, provide unbiased classifications across different patient populations or operational factors. This scoping review aimed to investigate how the current literature evaluates algorithmic bias in medical imaging AI models using quantitative fairness metrics.ApproachFollowing the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR), PubMed and Embase were systematically searched for peer-reviewed literature published between January 2015 and May 2025. Studies that evaluated classification medical imaging AI models for algorithmic bias and reported at least one quantitative measure of fairness were included. Articles that investigated fairness in nonimaging AI models used qualitative bias mitigation methods, examined technical bias (e.g., systematic measurement or preprocessing), or were conference proceedings were excluded.ResultsAfter removing duplicates, 2557 studies were included for title and abstract screening. Following initial screening, 143 articles were retrieved for full-text review. Of these, eight studies were eligible for inclusion. Half of the articles focused on bias detection, whereas the other half investigated mitigation techniques. There was considerable heterogeneity in the bias metrics used across the studies. Nearly all studies (7/8) used a chest X-ray radiography classifier for evaluation and examined algorithmic bias with respect to gender, age, and race.ConclusionsThe relative paucity of existing literature and variability in fairness evaluation methods underscore the need for more standardized frameworks to evaluate and ensure fairness in imaging AI systems.
- 27 May 2026
- Published by SAGE Publications in The American Journal of Sports Medicine
- Vol. 54 (9) , 2244-2252
AI-Assisted MRI Interpretation in Diagnosing Bankart and Reverse Bankart Lesions
Background: Accurate detection of glenoid labral tears remains challenging; however, it is essential for guiding treatment and return-to-play decisions in athletes. Magnetic resonance arthrography (MRA) improves diagnostic performance over noncontrast magnetic resonance imaging (MRI) but is invasive and more costly. Recent studies have shown that artificial intelligence (AI) models can accurately detect Bankart lesions on MRI and MRA, but whether AI assistance can improve clinician interpretation remains unknown. Purpose: To evaluate whether AI assistance improves clinician sensitivity and diagnostic confidence in detecting Bankart and reverse Bankart lesions on MRI and MRA. Study Design: Cohort study; Level of evidence, 3. Methods: A retrospective reader study was conducted using 586 shoulder examinations (335 MRI, 251 MRA) from 546 patients in the publicly available ScopeMRI data set, with arthroscopic findings as the reference standard. Custom AI models were trained to provide binary tear predictions for Bankart and reverse Bankart lesions. Four orthopaedic clinicians (2 fellowship-trained shoulder surgeons and 2 residents) independently reviewed all 117 test cases twice in a paired design: once unaided and once with AI assistance after a 60-day washout. The primary outcome was the change in sensitivity from unaided to aided interpretation, pooled across MRI and MRA, and evaluated separately for Bankart and reverse Bankart lesions. Secondary outcomes included changes in accuracy, specificity, and confidence. Results: For Bankart lesions, the mean sensitivity across clinicians increased from 38% unaided to 78.3% with AI assistance ( P < .001), while accuracy increased from 77.1% to 84% ( P = .002). For reverse Bankart lesions, sensitivity improved from 31.2% to 50% ( P = .049). Specificity remained high across conditions. Across both pathologies, AI assistance corrected 51.1% of initially missed labral tears. Diagnostic confidence increased with AI assistance by 0.80 for noncontrast MRIs ( P < .001) and 0.54 for MRAs ( P < .001) on a 10-point scale. Conclusion: AI assistance significantly improved clinician sensitivity and diagnostic confidence for detecting Bankart and reverse Bankart lesions on MRI and MRA, with minimal impact on specificity.
- 27 May 2026
- Published by SAGE Publications in The American Journal of Sports Medicine
- Vol. 54 (9) , 2244-2252
AI-Assisted MRI Interpretation in Diagnosing Bankart and Reverse Bankart Lesions
Background: Accurate detection of glenoid labral tears remains challenging; however, it is essential for guiding treatment and return-to-play decisions in athletes. Magnetic resonance arthrography (MRA) improves diagnostic performance over noncontrast magnetic resonance imaging (MRI) but is invasive and more costly. Recent studies have shown that artificial intelligence (AI) models can accurately detect Bankart lesions on MRI and MRA, but whether AI assistance can improve clinician interpretation remains unknown. Purpose: To evaluate whether AI assistance improves clinician sensitivity and diagnostic confidence in detecting Bankart and reverse Bankart lesions on MRI and MRA. Study Design: Cohort study; Level of evidence, 3. Methods: A retrospective reader study was conducted using 586 shoulder examinations (335 MRI, 251 MRA) from 546 patients in the publicly available ScopeMRI data set, with arthroscopic findings as the reference standard. Custom AI models were trained to provide binary tear predictions for Bankart and reverse Bankart lesions. Four orthopaedic clinicians (2 fellowship-trained shoulder surgeons and 2 residents) independently reviewed all 117 test cases twice in a paired design: once unaided and once with AI assistance after a 60-day washout. The primary outcome was the change in sensitivity from unaided to aided interpretation, pooled across MRI and MRA, and evaluated separately for Bankart and reverse Bankart lesions. Secondary outcomes included changes in accuracy, specificity, and confidence. Results: For Bankart lesions, the mean sensitivity across clinicians increased from 38% unaided to 78.3% with AI assistance ( P < .001), while accuracy increased from 77.1% to 84% ( P = .002). For reverse Bankart lesions, sensitivity improved from 31.2% to 50% ( P = .049). Specificity remained high across conditions. Across both pathologies, AI assistance corrected 51.1% of initially missed labral tears. Diagnostic confidence increased with AI assistance by 0.80 for noncontrast MRIs ( P < .001) and 0.54 for MRAs ( P < .001) on a 10-point scale. Conclusion: AI assistance significantly improved clinician sensitivity and diagnostic confidence for detecting Bankart and reverse Bankart lesions on MRI and MRA, with minimal impact on specificity.
- Open Access
Development and Preclinical Evaluation of a CPAP-Compatible EtCO2 Sampler
BACKGROUND: Capnography is a vital tool in EMS because changes to end-tidal CO 2 (ETCO 2 ) generally precede blood oxygen desaturation, and waveform morphology can be used to assess bronchial patency. Many of capnography’s indications overlap with those of continuous positive airway pressure (CPAP) therapy. However, there is currently no validated method to measure capnography during CPAP administration, and combining the two technologies in the prehospital setting lacks investigation in peer-reviewed literature. METHODS: We developed a novel exhalatory CO 2 sampler designed to seamlessly integrate with existing prehospital CPAP systems, manufactured via stereolithography. Using a Human Patient Simulator, we compared this novel sampler in vitro with a traditional nasal-oral cannula (NC), examining the effects of CPAP administration on ETCO 2 , capnogram shape, and respiration rate (RR). RESULTS: The CCA maintained an air-tight seal during CPAP administration and showed no significant differences in ETCO 2 values or RR compared to the NC. However, CPAP administration significantly reduced ETCO 2 values and altered capnogram morphology for both samplers. These changes indicate that CPAP itself affects capnography readings, necessitating caution in interpreting these measurements. CONCLUSIONS: The novel CCA is a viable tool for integrating capnography with CPAP in prehospital settings, but the effects of CPAP on capnography readings must be further studied. Future research should focus on clinical trials with human subjects to validate these findings and improve the interpretation of capnography data during CPAP therapy.
Open AccessDevelopment and Preclinical Evaluation of a CPAP-Compatible EtCO2 Sampler
BACKGROUND: Capnography is a vital tool in EMS because changes to end-tidal CO 2 (ETCO 2 ) generally precede blood oxygen desaturation, and waveform morphology can be used to assess bronchial patency. Many of capnography’s indications overlap with those of continuous positive airway pressure (CPAP) therapy. However, there is currently no validated method to measure capnography during CPAP administration, and combining the two technologies in the prehospital setting lacks investigation in peer-reviewed literature. METHODS: We developed a novel exhalatory CO 2 sampler designed to seamlessly integrate with existing prehospital CPAP systems, manufactured via stereolithography. Using a Human Patient Simulator, we compared this novel sampler in vitro with a traditional nasal-oral cannula (NC), examining the effects of CPAP administration on ETCO 2 , capnogram shape, and respiration rate (RR). RESULTS: The CCA maintained an air-tight seal during CPAP administration and showed no significant differences in ETCO 2 values or RR compared to the NC. However, CPAP administration significantly reduced ETCO 2 values and altered capnogram morphology for both samplers. These changes indicate that CPAP itself affects capnography readings, necessitating caution in interpreting these measurements. CONCLUSIONS: The novel CCA is a viable tool for integrating capnography with CPAP in prehospital settings, but the effects of CPAP on capnography readings must be further studied. Future research should focus on clinical trials with human subjects to validate these findings and improve the interpretation of capnography data during CPAP therapy.