Automatic Scoring of Diabetic Foot Ulcers through Deep CNN Based Feature Extraction with Low Rank Matrix Factorization

Severity stage classification of diabetic foot ulcers is a vital way to improve the decision support on treatment planning and diagnostic performance. Nevertheless, the precise manual identification of ulcer severity stage is challenging since vision may vary upon the consultant, tedious and time-consuming. Accordingly, automatic severity stage classification is of significant prominence. Yet a comprehensive computer-aided Wagner scale based severity stage classification system for diabetic foot ulcers is not available in the literature. In this paper, we propose to use a convolutional neural network engineered from DenseNet-201 based architecture as the feature extractor paradigm followed by a global average pooling (GAP) layer to predict six-class severity stages of diabetic foot ulcers. Furthermore, we use a practice of optimizing processing time and memory consumption while conserving the accuracy of the classification model through feature extraction with SVD. The proposed architecture could achieve an accuracy of over 96%. Moreover, we evaluated how different pre-trained CNN state of the art architectures (DenseNet, ResNet, Xception, InceptionV3, InceptionResNetV2, and VGG) which can be used for the task of transfer learning.