Anomaly Detection in Morphologies of Human Legs for the Implementation of Adaptive Leg Morphotypes and Medical Compression Stockings  
Author Timea Banfalvi

 
Co-Author(s) Dac Hieu Nguyen; Kim Duc Tran; Guillaume Tartare; Pascal Bruniaux; Xianyi Zeng; Kim Phuc Tran

 
Abstract This study introduces a deep learning-based method combining 3D feature detection with anomaly detection to identify and categorize morphological variations in human legs using unlabeled data. By clustering 3D leg scans into distinct morphotypes, the method enhances the design and sizing of medical compression stockings, crucial for treating chronic conditions such as chronic venous insufficiency and lymphedema, which significantly alter leg morphology. The proposed approach utilizes a Deep Convolutional K-Means (DCKM) model alongside the Elliptic Envelope algorithm to detect outliers, which improves data quality and refines model training without relying on labeled data. This approach holds promise for optimizing compression therapy by enabling more effective and personalized compression garments.

 
Keywords 3D morphology · Medical compression stockings · Deep Convolutional K-Mean · leg shape clustering · leg morphotypes
   
    Article #:  DSBFI25-48
 
Proceedings of 3rd ISSAT International Conference on Data Science in Business, Finance and Industry
January 6-8, 2025 - Da Nang, Vietnam