Radiomics improves thyroid nodule cancer prediction with uncertainty estimates

Radiomics-Based Thyroid Nodule Classification with Label-Conditional Conformal Prediction

Computer Vision and Pattern Recognition

Summary

It can be hard and inconsistent for doctors to decide if thyroid nodules seen in ultrasounds might be cancerous, leading to many unnecessary biopsies. The authors propose a method that uses computer analysis of ultrasound images to identify features that help classify nodules as benign or malignant. Their approach also provides uncertainty estimates, so uncertain cases are flagged for further clinical review instead of forcing a yes-or-no answer. This aims to reduce unnecessary biopsies while keeping doctors involved in difficult cases.

What this means in practice

  • For radiologists: Support clinical decisions by flagging uncertain thyroid nodules for review while automatically classifying clearer cases using ultrasound-based features.
  • For medical image software developers: Incorporate label-conditional conformal prediction with radiomics features to build thyroid nodule classification tools that provide calibrated uncertainty estimates.

Tested on one dataset.

Authors

Laura Testa, Pierpaolo Brutti, Umberto Ferraro Petrillo

Abstract

Thyroid nodule malignancy assessment relies on ultrasound imaging, yet current risk stratification systems suffer from substantial interobserver variability, contributing to a high rate of unnecessary biopsies. We propose a radiomics-based framework combining quantitative feature extraction via PyRadiomics with a Generalized Additive Model (GAM) classifier, selected among six candidates for its favorable trade-off between discriminative performance and interpretability. To complement point predictions with uncertainty estimates, Cross-Conformal Prediction with label-conditional guarantees is applied, flagging ambiguous cases for clinical review rather than forcing a binary decision. Experiments are conducted on TN5000, a publicly available ultrasound dataset of thyroid nodules. After Spearman correlation filtering, 36 features are retained and the GAM achieves an AUC of $0.89$, a sensitivity of $0.83$, and a specificity of $0.82$ on the validation set. At a significance level $ε=0.09$, selected to ensure empirical coverage guarantees on both classes, 63\% of nodules on the test set receive a certain (i.e., single-label) prediction, with conditional precision of $0.94$ for certain malignant and conditional Negative Predictive Value (NPV) of $0.73$ for certain benign cases, while the remaining 37\% are referred to the clinician.