Improving Trustworthiness of AI Disease Severity Rating in Medical Imaging with Ordinal Conformal Prediction Sets

2022/07/05 by Charles Lu, Anastasios N. Angelopoulos, Lu, Charles +4 · 11 citations
Computer Science · Engineering · Medicine · #Artificial Intelligence in Healthcare and Education #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Medical Imaging and Analysis #Radiomics and Machine Learning in Medical Imaging #cs.LG #eess.IV #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.2207.02238

arxiv created 2022/07/05 · openalex publication_date 2022/07/05 · arxiv updated 2022/07/07 · openalex created_date 2023/01/02 · openalex updated_date 2026/07/28

Abstract

The regulatory approval and broad clinical deployment of medical AI have been hampered by the perception that deep learning models fail in unpredictable and possibly catastrophic ways. A lack of statistically rigorous uncertainty quantification is a significant factor undermining trust in AI results. Recent developments in distribution-free uncertainty quantification present practical solutions for these issues by providing reliability guarantees for black-box models on arbitrary data distributions as formally valid finite-sample prediction intervals. Our work applies these new uncertainty quantification methods -- specifically conformal prediction -- to a deep-learning model for grading the severity of spinal stenosis in lumbar spine MRI. We demonstrate a technique for forming ordinal prediction sets that are guaranteed to contain the correct stenosis severity within a user-defined probability (confidence interval). On a dataset of 409 MRI exams processed by the deep-learning model, the conformal method provides tight coverage with small prediction set sizes. Furthermore, we explore the potential clinical applicability of flagging cases with high uncertainty predictions (large prediction sets) by quantifying an increase in the prevalence of significant imaging abnormalities (e.g. motion artifacts, metallic artifacts, and tumors) that could degrade confidence in predictive performance when compared to a random sample of cases.

Cited by

• Cost-Sensitive Conformal Training with Provably Controllable Learning Bounds
• Provably Minimum-Length Conformal Prediction Sets for Ordinal Classification
• Analyzing Uncertainty of LLM-as-a-Judge: Interval Evaluations with Conformal Prediction
• Efficient Conformal Prediction for Regression Models under Label Noise
• Conformal Prediction and Trustworthy AI
• Pitfalls of Conformal Predictions for Medical Image Classification
• Trustworthy Few-Shot Transfer of Medical VLMs through Split Conformal Prediction
• Direct Prediction Set Minimization via Bilevel Conformal Classifier Training
• Risk-Sensitive Conformal Prediction for Catheter Placement Detection in Chest X-rays
• WQLCP: Weighted Adaptive Conformal Prediction for Robust Uncertainty Quantification Under Distribution Shifts
• Privacy-Preserving Conformal Prediction Under Local Differential Privacy

Related

• Toward Clinically Trustworthy Deep Learning: Applying Conformal Prediction to Intracranial Hemorrhage Detection
• A conformalized learning of a prediction set with applications to medical imaging classification
• A Critical Perspective on Finite Sample Conformal Prediction Theory in Medical Applications
• A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification
• Empirical Validation of Conformal Prediction for Trustworthy Skin Lesions Classification