Title: LDA-aided threshold to classify neuropathy and non-neuropathy in diabetic patients

Authors: R. Periyasamy; Deepak Joshi; Sonal Atreya; Sneh Anand

Addresses: Centre for Biomedical Engineering, Indian Institute of Technology, Delhi New Delhi, India; Biomedical Engineering Unit, All India Institute of Medical Sciences, New Delhi, India. ' Centre for Biomedical Engineering, Indian Institute of Technology, Delhi New Delhi, India; Biomedical Engineering Unit, All India Institute of Medical Sciences, New Delhi, India. ' Centre for Biomedical Engineering, Indian Institute of Technology, Delhi New Delhi, India; Biomedical Engineering Unit, All India Institute of Medical Sciences, New Delhi, India. ' Centre for Biomedical Engineering, Indian Institute of Technology, Delhi New Delhi, India; Biomedical Engineering Unit, All India Institute of Medical Sciences, New Delhi, India

Abstract: An early investigation of neuropathy in diabetic patients is a crucial factor to monitor the progress of risk status (nerve damage, sensory loss, pain, numbness, irritation and other symptoms) in the foot. Nerve conduction test and 10 gm Semmes Weinstein monofilament test are the gold standards for the detection but are challenging for the patients. Hence, the detection techniques available depend on experimental results and are highly subjective. This paper presents the supervised technique, Linear Discriminant Analysis (LDA), to identify neuropathy and non-neuropathy at early stages using three parameters (hardness, Two-Point Discrimination (TPD) and Power Ratio (PR)) of different areas under the foot sole. Hardness, PR and TPD in the areas of both left and right feet of 13 diabetic patients (5 males and 8 females) are assessed. The results show that foot areas 3, and 4, and areas 5, 6 and 7 are the more sensitive (risk) zone for detection, with the classification accuracy of 92.3% and 84.6%. Therefore, threshold technique proves to be quite accurate for the identification of neuropathy and non-neuropathy in diabetic patients. However, classification accuracy can be further increased by training with a large number of data.

Keywords: diabetes mellitus; neuropathy; Semmes Weinstein monofilament; power ratio; hardness; two point discrimination; supervised techniques; LDA; linear discriminant analysis; diabetic patients; risk status; feet; nerve conduction; foot sole.

DOI: 10.1504/IJBET.2011.044411

International Journal of Biomedical Engineering and Technology, 2011 Vol.7 No.4, pp.315 - 326

Published online: 21 Jan 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article