DFUNet: Convolutional Neural Networks for Diabetic Foot Ulcer Classification (1711.10448v2)

Abstract: Globally, in 2016, one out of eleven adults suffered from Diabetes Mellitus. Diabetic Foot Ulcers (DFU) are a major complication of this disease, which if not managed properly can lead to amputation. Current clinical approaches to DFU treatment rely on patient and clinician vigilance, which has significant limitations such as the high cost involved in the diagnosis, treatment and lengthy care of the DFU. We collected an extensive dataset of foot images, which contain DFU from different patients. In this paper, we have proposed the use of traditional computer vision features for detecting foot ulcers among diabetic patients, which represent a cost-effective, remote and convenient healthcare solution. Furthermore, we used Convolutional Neural Networks (CNNs) for the first time in DFU classification. We have proposed a novel convolutional neural network architecture, DFUNet, with better feature extraction to identify the feature differences between healthy skin and the DFU. Using 10-fold cross-validation, DFUNet achieved an AUC score of 0.962. This outperformed both the machine learning and deep learning classifiers we have tested. Here we present the development of a novel and highly sensitive DFUNet for objectively detecting the presence of DFUs. This novel approach has the potential to deliver a paradigm shift in diabetic foot care.

• The paper introduces DFUNet, a novel CNN architecture that classifies diabetic foot ulcers with a 0.961 AUC, outperforming traditional classifiers.
• The methodology leverages multi-level feature extraction via parallel convolutions and robust data augmentation to improve generalization.
• The study demonstrates DFUNet's potential for real-time telemedicine, reducing computational resources while enhancing diagnostic accuracy.

DFUNet: A Novel Approach to Diabetic Foot Ulcer Classification

Diabetic Foot Ulcers (DFUs) represent a significant complication of Diabetes Mellitus (DM), often leading to severe consequences, including lower limb amputation if left untreated. Traditional clinical methods for DFU management require vigilance and costly interventions, making automated, efficient, and remote solutions imperative. This paper introduces DFUNet, a convolutional neural network (CNN) architecture designed to classify DFUs with high sensitivity and specificity, leveraging a significant dataset of foot images collected from patients experiencing DFUs.

Methodological Innovations and Contributions

DFUNet addresses the primary challenge of DFU identification through computer vision techniques by assessing visual features of foot ulcers compared to normal skin. This approach is novel, being the first application of CNNs for DFU classification, demonstrating strong performance metrics across various machine learning models, with an AUC of 0.961 using a 10-fold cross-validation technique. This result noticeably surpasses alternative classifiers traditionally employed in medical imaging tasks.

The methodology undertakes comprehensive data augmentation to improve the deep learning model's generalization abilities. Pre-trained CNN architectures like AlexNet and GoogLeNet were tested alongside DFUNet, the latter being specifically tailored to recognize DFUs efficiently. DFUNet utilizes a blend of traditional CNN layers initially, followed by parallel convolutions enabling multi-level feature extraction, effectively differentiating between healthy skin and DFUs.

The proposed DFUNet showed noteworthy processing efficiency compared to established deep learning architectures, not just matching but outperforming their classification accuracy, indicating its potential for application in real-time clinical assessments. The research underlines the importance of fine-tuning CNN architectures for specific binary classifications with fewer layers compared to more complex networks like GoogLeNet, thus reducing the required computational resources.

Implications and Future Directions

The implications of DFUNet in diabetic foot care are substantial, offering a pathway toward automated, cost-effective, and accurate DFU assessments outside conventional healthcare settings. Deployment of models like DFUNet in telemedicine applications could transform diabetic patient monitoring, delivering preventive care efficiently and potentially halting progression to severe complications.

Future advancements could explore the extension of DFUNet to multi-class classifications, including various DFU pathologies, enhancing the system's diagnostic capabilities aligned with clinical grading scales. Moreover, exploring DFUNet’s generalizability for various skin lesions beyond DFUs, such as wound classifications or infections like chickenpox, could open broader applications in dermatological assessments. Additionally, integrating DFUNet into mobile applications could democratize access to vital diagnostic tools where expert medical care is limited, particularly in underserved regions globally.

In conclusion, DFUNet establishes a strong foundation for innovative diabetic foot care solutions, demonstrating the efficacy of tailored CNNs in medical image classifications. Its development heralds an era of intelligent, automated healthcare systems, potentially changing the landscape of diabetic management, prevention, and patient education. These research advancements lay the groundwork for more inclusive and robust health technologies, driving progress toward accessible and efficient healthcare solutions worldwide.