This　study　aims　to　address　the　challenge　of　low　diagnostic　accuracy　of　deep　learning　(DL)　models　due　to　the　limited　availability　of　magnetic　resonance　imaging　(MRI)　data　of　the　foot　and　ankle　region.　For　this　purpose,　plantar　fasciitis　(PF)　was　selected　as　the　focus　of　the　study-a　common　foot　and　ankle　condition-and　the　feasibility　and　effectiveness　of　DL　models　trained　on　small-sample　datasets　for　assisting　PF　diagnosis　were　evaluated.　Firstly,　MRI　images　of　PF　patients　were　collected　and　augmented　using　transformation-based　data　augmentation　techniques.　Secondly,　a　DL　model　based　on　Faster　Regional　Convolutional　Neural　Network　(Faster　R-CNN)　was　constructed　and　trained　using　a　combination　of　data　augmentation　and　transfer　learning　strategies,　with　performance　evaluated　through　five-fold　cross-validation.　Finally,　the　generalizability　of　the　DL　model　was　evaluated.　The　experiments　demonstrate　that,　PF　lesions　can　be　efficiently　identified　using　DL　techniques,　especially　through　integrating　the　Faster　R-CNN　model,　data　augmentation　methods,　and　transfer　learning　strategies,　even　when　training　data　are　limited.　The　proposed　method　provides　a　good　prospect　for　assisting　PF　diagnosis.　©　2025　School　of　Science,　DUTH.　All　rights　reserved.