Scholarlink

Background <p>Diabetic foot ulcer (DFU) is a serious complication of diabetes, primarily caused by poor foot care and inadequate disease management. If left untreated, DFUs can lead to severe infections and even amputation, significantly affecting patient quality of life and survival. Existing DFU detection methods often suffer from low accuracy, high computational complexity, and increased false positive rates, highlighting the need for more reliable and efficient diagnostic solutions.</p> Objective <p>This study aims to develop a robust and efficient deep learning (DL)-based framework, termed Dynamic Slimmable Network with Pyramid Attention Network (DSN-PyAN), for accurate and computationally efficient DFU detection and classification.</p> Methods <p>The proposed DSN-PyAN framework integrates multiple advanced techniques to enhance performance. The method removes noise from input images without erasing crucial edge information by using a robust generalized <i>t</i> distribution-based Kalman filter (RGt-DKF). Moreover, the continuous short-time Fourier wavelet transform (CSTFWF) delineates informative features by integrating spatial and frequency-domain information. Anatomical structures are guided by the Anatomy-Aware Hover Transformer for ROI (AAHT-ROI) to accomplish accurate segmentation and DFU region localization. The DSN-PyAN classifier automatically modifies network width to attain optimal computation and simultaneously better classification performance. The Human Memory Optimization Algorithm (HMOA) tunes model parameters to facilitate better decision-making.</p> Results <p>Experimental evaluation on a DFU dataset demonstrates that the proposed DSN-PyAN model achieves superior performance, with an accuracy of 99.8% and a recall of 99.7%. The model outperforms existing methods in terms of accuracy, computational efficiency, and reduction of false positives.</p> Conclusion <p>The DSN-PyAN framework integrates adaptive DL, anatomy-aware segmentation, and optimization-driven classification, marking it as a fascinating and promising tool for clinical DFU diagnosis. The modes aim to support early detection of DFUs, improve treatment strategies, and mitigate complications caused by DFUs.</p>

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

Diabetic foot ulcer detection using a Dynamic Slimmable Network with Pyramid Attention Network-based Human Memory Optimization Algorithm

  • 期刊论文

  • 发表日期：

    2026年4月1日

查看全文
  • G. Karthi,
  • K. Suresh Kumar,
  • Law Kumar Singh,
  • Prabha Murugesan

摘要

Background

Diabetic foot ulcer (DFU) is a serious complication of diabetes, primarily caused by poor foot care and inadequate disease management. If left untreated, DFUs can lead to severe infections and even amputation, significantly affecting patient quality of life and survival. Existing DFU detection methods often suffer from low accuracy, high computational complexity, and increased false positive rates, highlighting the need for more reliable and efficient diagnostic solutions.

Objective

This study aims to develop a robust and efficient deep learning (DL)-based framework, termed Dynamic Slimmable Network with Pyramid Attention Network (DSN-PyAN), for accurate and computationally efficient DFU detection and classification.

Methods

The proposed DSN-PyAN framework integrates multiple advanced techniques to enhance performance. The method removes noise from input images without erasing crucial edge information by using a robust generalized t distribution-based Kalman filter (RGt-DKF). Moreover, the continuous short-time Fourier wavelet transform (CSTFWF) delineates informative features by integrating spatial and frequency-domain information. Anatomical structures are guided by the Anatomy-Aware Hover Transformer for ROI (AAHT-ROI) to accomplish accurate segmentation and DFU region localization. The DSN-PyAN classifier automatically modifies network width to attain optimal computation and simultaneously better classification performance. The Human Memory Optimization Algorithm (HMOA) tunes model parameters to facilitate better decision-making.

Results

Experimental evaluation on a DFU dataset demonstrates that the proposed DSN-PyAN model achieves superior performance, with an accuracy of 99.8% and a recall of 99.7%. The model outperforms existing methods in terms of accuracy, computational efficiency, and reduction of false positives.

Conclusion

The DSN-PyAN framework integrates adaptive DL, anatomy-aware segmentation, and optimization-driven classification, marking it as a fascinating and promising tool for clinical DFU diagnosis. The modes aim to support early detection of DFUs, improve treatment strategies, and mitigate complications caused by DFUs.