Explainable artificial intelligence for automated flatfoot detection in foot x-ray images

Flatfoot (pes planus) is a common musculoskeletal condition characterized by collapse of the medial longitudinal arch, which can cause pain, functional impairment, and progression to severe deformities if not detected early. Conventional assessment using weight-bearing radiographs is often time-consuming and subject to inter-observer variability. We present a cost-efficient explainable system for automated detection of pes planus from lateral foot X-ray images using deep learning-based feature representations. The dataset included 402 pes planus and 440 normal radiographs. The framework employs transfer learning with a modified VGG16 backbone and rigorous statistical feature selection, reducing the feature space from 128 to 11 discriminative features. Four machine learning classifiers, support vector machine (SVM), decision tree (DT), random forest (RF), and logistic regression (LR), were trained and validated, then tested on an unseen patient. Both DT and RF achieved strong and comparable performance. Analysis of receiver operating characteristic (ROC) showed that DT attained an area under the curve (AUC) of 0.97 (95% CI 0.93 to 0.99), while RF reached a slightly higher AUC of 0.99 (95% CI 0.96 to 1.00). The results indicate that the proposed method is effective, interpretable, and computationally efficient. Its potential clinical relevance is further supported by decision curve analysis and explainable artificial intelligence using local interpretable model-agnostic explanations, suggesting its utility as a supportive tool for flatfoot assessment.