Publication

Abstract : Pes planus, also known as flatfoot, is a condition in which the foot’s medial longitudinal arch is lowered or missing. If untreated, it can cause issues with posture and movement. Proper clinical management requires careful and timely diagnosis. This study compares different deep learning (DL) models and optimizers to find the best method for effective classification of flatfoot in diagnostic imaging. InceptionV3, Xception, ResNet101V2, and DenseNet201 were trained using Adam, RMSprop, and SGD optimizers to determine their robustness in classifying flatfoot from normal foot images. ResNet101V2 model paired with SGD optimizer obtained the maximum classification accuracy of 98.24%, closely followed by the DenseNet201 model with SGD optimizer at 98.08%. These results imply that the combined use of deep neural networks and optimization methods can be proven advantageous in diagnostic medical imaging. The findings demonstrate the potential of the DL-based diagnostic support systems in enhancing the reliability of Pes Planus classification and thereby help clinicians automate the screening process. The scalability of these models is also emphasized in the study, which makes it possible to apply them to other musculoskeletal imaging tasks. The promising results also illustrate the importance of incorporating AI-powered tools into the regular workflow of clinical environments to enhance efficiency and patient care.