Scholarlink

Diabetes mellitus (DM) causes several complications, including diabetic foot ulceration (DFU), which can lead to lower-limb amputation. Non-invasive thermographic imaging using infrared cameras has shown promise in detecting early thermal anomalies in diabetic feet. This paper proposes a deep learning-based approach using ResNet 50 v2 and AlexNet to classify thermographic images based on the Thermal Change Index (TCI). We focus on temperature variations across multiple foot angiosomes, including the Lateral and Medial Plantar Arteries (LPA, MPA) and Medial and Lateral Calcaneal Arteries (MCA, LCA). Extensive image enhancement techniques and novel CNN architectures resulted in a 93.2% classification accuracy across five TCI-based classes. Our results indicate significant performance improvements over traditional machine learning methods. These findings could enable real-time remote monitoring for early detection of DFU in diabetic patients for good health and well-being.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep Learning Techniques for Diabetic Foot Thermogram Image Classification Based on Thermal Change Index

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • D. Jeyandhan,
  • P. Nagaraj,
  • Josephine Selle Jeyanathan

摘要

Diabetes mellitus (DM) causes several complications, including diabetic foot ulceration (DFU), which can lead to lower-limb amputation. Non-invasive thermographic imaging using infrared cameras has shown promise in detecting early thermal anomalies in diabetic feet. This paper proposes a deep learning-based approach using ResNet 50 v2 and AlexNet to classify thermographic images based on the Thermal Change Index (TCI). We focus on temperature variations across multiple foot angiosomes, including the Lateral and Medial Plantar Arteries (LPA, MPA) and Medial and Lateral Calcaneal Arteries (MCA, LCA). Extensive image enhancement techniques and novel CNN architectures resulted in a 93.2% classification accuracy across five TCI-based classes. Our results indicate significant performance improvements over traditional machine learning methods. These findings could enable real-time remote monitoring for early detection of DFU in diabetic patients for good health and well-being.