Toward Smarter Orthopedic Care: Classifying Plantar Footprints from RGB Images Using Vision Transformers and CNNs.

Area of Science:

• Orthopedics and Sports Medicine
• Medical Imaging Analysis
• Computational Biology

Background:

• Foot anatomical structure assessment is crucial for orthopedic intervention.
• Foot structure classification is essential for early diagnosis of musculoskeletal disorders affecting lower extremities.

Purpose of the Study:

• To develop an accurate method for classifying plantar footprints and hindfoot alignment in the sagittal plane.
• To evaluate the efficacy of state-of-the-art deep learning models for this classification task.

Main Methods:

• Creation of a custom dataset of 603 augmented RGB plantar images.
• Training and evaluation of six advanced deep learning models: swin_tiny_patch4_window7_224, convnextv2_tiny, deit3_base_patch16_224, xception41, inception-v4, and efficientnet_b0.

Main Results:

• The swin_tiny_patch4_window7_224 model demonstrated superior performance.
• Achieved an accuracy of 98.013% in classifying plantar footprints and hindfoot alignment.

Conclusions:

• The swin_tiny_patch4_window7_224 model shows significant potential as a reliable and cost-effective tool.
• This method can aid in clinical screening and diagnosis of foot-related orthopedic conditions.