Radiomics-Based Prediction of Patient Demographic Characteristics on Chest Radiographs: Looking Beyond Deep Learning for Risk of Bias

Using two public CXR datasets, radiomics ML models showed moderate-to-good performance for predicting patients’ age, sex, and self-reported race. This encoding of demographic characteristics by radiomics ML models raises concern for biases similar to those previously identified for deep-learning models, requiring radiologist vigilance and further study.