EUVIMEDEUVIMED entdecken
Crossref

Application of Explainable Artificial Intelligence in Detecting Foot Characteristics: An Exploratory Study Focused on Sex Differences

Taro Kasai, Ryutaro Takeda, Toshiko Iidaka, Kenta Makabe, Yasunori Omata, Noriko Yoshimura, Sakae Tanaka, Takumi Matsumoto

Journal of Orthopaedic Foot and Ankle Science · 27.07.2026 · 1 (2) · 66-75

DOI: 10.64079/jofas.2025-0020 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Orthopaedic Foot and Ankle Science, erschienen am 27.07.2026. Verfasst von Taro Kasai und 7 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Foot morphologies related to specific populations and diseases have been reported; however, their identification remains limited by human cognition. This study aimed to evaluate the utility of explainable artificial intelligence (XAI) in detecting foot characteristics, using sex differences as an exploratory study.

Methods

A dataset of 2,004 radiographs of bilateral feet (657 male, 1,347 female) from a nationwide cohort was collected. After selecting 2,734 foot images derived from 1,367 radiographs of bilateral feet that met our AI criteria, 2,186 foot images (80%) were allocated for training, 274 (10%) for validation, and 274 (10%) for testing datasets. The VGG16 deep learning algorithm was applied to develop an AI program for sex identification. Gradient-weighted class activation mapping (Grad-CAM), a type of XAI, visualized the AI program’s regions of interest. Radiographic parameters relevant to these regions were exploratorily analyzed using 100 randomly selected images for each sex.

Results

The AI program demonstrated high performance in sex classification from foot images, correctly classifying 266 of 274 images and achieving an accuracy of 0.97. Grad-CAM identified the navicular and talar head as key regions in males, and the distal tibial shaft and the second and third metatarsophalangeal joints in females. Exploratory comparative analysis of radiographic parameters in these regions highlighted significant sex-based differences: males tended to have a lower central height of the navicular; females had wider bases and narrower shafts in the second and third proximal phalanges, less flat-shaped second and third metatarsal heads, and a narrower distal tibial shaft width relative to the malleolar width.

Conclusions

XAI revealed sex-specific foot morphologies. This methodology may provide a framework for future studies exploring population- or disease-specific foot morphologies.Level of Evidence: Level III

Zitieren

APA

Kasai, T., Takeda, R., Iidaka, T., Makabe, K., Omata, Y., Yoshimura, N., Tanaka, S., & Matsumoto, T. (2026). Application of Explainable Artificial Intelligence in Detecting Foot Characteristics: An Exploratory Study Focused on Sex Differences. Journal of Orthopaedic Foot and Ankle Science, 1(2), 66-75. https://doi.org/10.64079/jofas.2025-0020

Vancouver

Kasai T, Takeda R, Iidaka T, Makabe K, Omata Y, Yoshimura N, et al. Application of Explainable Artificial Intelligence in Detecting Foot Characteristics: An Exploratory Study Focused on Sex Differences. Journal of Orthopaedic Foot and Ankle Science. 2026;1(2):66-75. doi:10.64079/jofas.2025-0020

BibTeX
@article{Kasai2026,
  author = {Kasai, Taro and Takeda, Ryutaro and Iidaka, Toshiko and Makabe, Kenta and Omata, Yasunori and Yoshimura, Noriko and Tanaka, Sakae and Matsumoto, Takumi},
  title = {Application of Explainable Artificial Intelligence in Detecting Foot Characteristics: An Exploratory Study Focused on Sex Differences},
  journal = {Journal of Orthopaedic Foot and Ankle Science},
  year = {2026},
  volume = {1},
  number = {2},
  pages = {66--75},
  doi = {10.64079/jofas.2025-0020}
}
Kostenlos starten →