AI-BASED EARLY DIABETIC FOOT RISK PREDICTION USING MULTIMODAL DATA (THERMAL + RGB + PATIENT HISTORY)

Abstract

Diabetic Foot Ulcer (DFU) is a critical
complication of diabetes that can lead to severe
infections, hospitalization, and amputation if not
detected early. Conventional diagnostic methods
rely heavily on clinical observation, which often
identifies the condition only after visible tissue
damage has occurred. This project proposes an AIBased
Early Diabetic Foot Risk Prediction System
that leverages multimodal data, including RGB
images, thermal images, and patient clinical history,
to enable early detection and preventive
intervention. RGB images provide visual insights
into skin abnormalities such as discoloration and
swelling, while thermal imaging captures
temperature variations associated with
inflammation and tissue stress. Additionally, patient
medical data such as age, HbA1c levels, duration of
diabetes, and prior ulcer history enhance the
predictive accuracy. Deep learning techniques,
particularly Convolutional Neural Networks
(CNNs), are employed for extracting meaningful
features from image data, while machine learning
algorithms analyze structured clinical data. A
multimodal fusion approach integrates all extracted
features to classify patients into low, medium, or
high-risk categories. The system is non-invasive,
cost-effective, and suitable for both clinical and
remote healthcare environments. By enabling early
identification of at-risk patients, the proposed
system supports timely intervention, reduces
complications, and improves patient outcomes. The
integration of artificial intelligence and multimodal
data demonstrates a significant advancement in
preventive healthcare and clinical decision support
systems.