Analysis of Diabetic Foot Ulcer Risk Factors, Causes, and Quantitative Measurement tools
Vaidya, Palavi   (University of Michigan-Ann Arbor)
Mentor: Najarian, Kayvan (University of Michigan)
Diabetic foot ulceration (DFU) is one of the most common reasons for lower extremity amputations. These ulcers cause infections and cause tissue necrosis, which leads to decreased wound healing. It is important to gain a comprehensive understanding of risk factors, causes, and current clinical quantitative measurement techniques for DFU wounds. All diabetic patients require an annual comprehensive foot examination to determine the risk factors for ulcers and amputations. There are several risk factors that increase the risk of DFU and affect the healing process. These factors include previous amputation, previous history of DFU, peripheral neuropathy, foot deformity, peripheral arterial disease (PAD), diabetic nephropathy specially ESRD, poor glycemic control, and cigarette smoking. Current treatment of foot ulcers consists of relief of pressure and protection of the ulcer, restoration of skin perfusion, treatment of infection, metabolic control and treatment of comorbidity, education, and prevention of recurrence. Etiology of DFU is multifactorial with various pathophysiological pathways that are activated due to Diabetes Mellitus. These pathways can be broken down into three categories: neuropathy, trauma, and vascular disease. Neuropathies can include deficits seen in motor, sensory or autonomic systems. Trauma from footwear, foot deformity, or high-pressure spots can lead to foot ulceration. Vascular effects can be seen from both micro and macro perspectives. Importance in early diagnosis for DFU is critical to faster healing rates. For this reason, an integrated system based on deep learning Convolutional Neural Networks (CNN) was created to perform wound segmentation and analysis. Using a combination of local and global textures, an input image is analyzed by the CNN to determine wound surface area, presence of infection, and healing progress. The accuracy and precision of this combined CNN model is assessed using images of DFUs taken from the Wound Care Clinic. Once validated, patient data for those who are being treated in the Wound Care Clinic at University of Michigan are fed to the model for analysis. Each patient will have about 7-8 data points for a total of 700-800 total data points. Measurements such as age, gender, wound images, arterial Doppler study results, wound infection status, healing date, lab work such as HbA1c, albumin/prealbumin, eGFR, ESR and CRP are provided to the learning model to create a more precise and accurate healing rate prediction. This current proposal is under IRB approval.