<![CDATA[Diabetic foot ulcers affect 15-25% of diabetic patients globally, causing significant morbidity, healthcare costs, and amputation risks. Early detection is crucial for preventing severe complications, yet limited access to specialized healthcare services, especially in remote areas, creates substantial barriers to timely wound management. This study developed a mobile-based self-detection system using deep learning technology to enable early diabetic wound identification and improve healthcare accessibility for patients in resource-limited settings. A Convolutional Neural Network with MobileNetV2 architecture was trained on over 5,000 diabetic wound images categorized into five classes: healthy skin and Wagner grades 1-4 ulcers. The cross-platform mobile application features AI-based wound detection, diagnosis history tracking, educational content, and integrated telemedicine consultation. Clinical validation compared AI predictions with three certified wound care specialists across 500 cases, while usability testing involved 100 diabetic patients.The model achieved 92.4% accuracy with 94.2% sensitivity and 91.7% specificity. For high-grade ulcers, performance improved to 96.8% sensitivity with 1.8-second processing time per image. Clinical validation showed substantial agreement with specialists (Cohen's kappa = 0.89). The application scored 82.5/100 on System Usability Scale with 88.5% user satisfaction and 91.2% willing to recommend it. This mobile-based deep learning system demonstrates high accuracy and clinical reliability for diabetic wound self-detection, successfully bridging healthcare gaps for underserved populations and showing significant potential for early intervention and healthcare cost reduction.]]>
