Diagnosis of superficial ailments using infrared thermal imaging and CapsNet

Infrared thermal imaging has been widely recognized as a reliable tool for detecting superficial diseases by analyzing temperature variations. However, external factors can influence regional temperature, introducing potential errors in diagnosis. This study investigates the integration of a computer vision-based classification algorithm with thermal imaging to enhance disease classification and diagnosis. Specifically, the research explores the application of Capsule Networks (CapsNet) for diagnosing conditions with skin-manifesting symptoms, such as breast cancer, pressure ulcers, and sinusitis. The proposed approach involves processing thermal images as conventional images using a computer vision-based classification algorithm. CapsNet is employed as the primary deep learning model for disease classification. The performance of CapsNet is evaluated and compared against benchmark models to assess its effectiveness in diagnosing breast cancer, pressure ulcers, and sinusitis. Experimental results demonstrate that CapsNet consistently outperforms traditional models across all evaluated conditions. The model achieves an accuracy of 99.91 % in breast cancer detection, 92.12 % in pressure ulcer staging, and 99.10 % and 98.93 % in multiclass and binary sinusitis classification, respectively. The integration of thermal imaging with CapsNet presents a promising approach for diagnosing superficial diseases. The high accuracy rates across different medical conditions highlight the potential of this method in medical imaging applications. These findings suggest that thermal imaging, when combined with advanced deep learning techniques, should be more widely adopted in medical diagnostics.