A Screening Computational Tool for Detection of Diabetic Neuropathy and Non-Neuropathy in Type-2 Diabetes Subjects

Foot problems are common complication in people with diabetes mellitus. An early detection of diabetic neuropathy and non-neuropathy in type 2 diabetes is a crucial test to monitor the nerve damage, sensory loss, pain, numbness and other symptoms in the foot. But available detection techniques are depend on experimental results and highly subjective. Therefore in this study we present artificial neural network (ANN) computational tool for classification of diabetic neuropathy and non-neuropathy from type-2 diabetes subjects by evaluating two parameters (standing foot pressures distribution parameter-Power ratio, foot sole hardness) under the foot sole of different areas. Hardness (H) using shore meter and standing plantar pressure distribution parameter-power ratio (PR) using portable PedoPowerGraph were obtained in the areas of right foot of 170 type-2 diabetes subjects (110 neuropathic patients and 60 non-neuropathic subjects). A sample data (PR and H value) of 40 diabetic subjects (23 were neuropathic and 17 were non-neuropathic) were computed and fed to back propagation feed forward artificial neural network (ANN) for automatic classification of diabetic neuropathy at an early stage, so as to prevent foot ulcer formation and amputation. Our results show that the foot area 2 and 7 are the risk zone for detection of neuropathy with the overall classification accuracy of 94.5%. The average computational time taken by the ANN was 0.75 seconds. The sensitivity and specificity of classifier were found to be 95.6% and 94.1%. Therefore this paper suggests that the proposed classifier is accurate and time efficient compared to other existing systems. However classification accuracy can be further increased with large number of sample data. Hence such computational screening tool can help the clinicians to cross check their diagnosis.