Automated Diabetic Foot Ulcer Detection Using Hybrid Short Term Memory Architecture

Diagnosis of diabetic foot ulcer in early stage is vital to prevent complications viz., amputations and infections. Conventional image processing methods lack to provide automatic and accurate detection of Diabetic Foot Ulcers (DFUs) in earlier stages. Deep learning based method serves as a better solution for diabetic foot ulcer detection by deeply analyzing complex features. This paper proposes a novel model that integrates the strengths of Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) for comprehensive analysis and detection of DFUs in the NIH dataset. The combination of the hybrid approach enables the extraction of both spatial and temporal features from DFU images, improving the model's accuracy by learning appropriate patterns for precise diagnosis. The extracted features are collected and given as input to Recurrent Neural Networks (RNNs) to capture temporal dependencies and sequential patterns across successive image frames. Experimental analysis reveals that the proposed method achieves an accuracy of 94.4%, F1 Score of 94.76%, a precision of 96.61 % and a specificity of 93.47%. From the analysis, it is depicted that the hybrid deep learning approach reduces the burden of health care systems that rely on manual feature extraction and it also efficiently learned the subtle features to provide better results.