Convolutional Neural Network Based Clustering And Manifold Learning Method For Diabetic Plantar Pressure Imaging Dataset

Keywords

Clustering; Convolutional neural network; Diabetic mellitus; Parameterized manifold learning; Plantar pressure imaging

Abstract

Foot plantar pressure characteristics can be used to investigate and characterize diabetic patients. The current work proposed an effective method for analyzing plantar pressure images in order to obtain the key areas of foot plantar pressure characteristics. A collected data of plantar pressure of diabetic patients is involved to evaluate the proposed method based on image analysis. Initially, the plantar pressure imaging dataset was preprocessed by using watershed transformation to determine the region of interest (ROI) as well as to decrease the computation complexity. Afterward, the convolutional neural network (CNN) based K-mean clustering and parameterized manifold learning using an improved isometric mapping algorithm (ISOMAP) were applied to attain segments of the imaging dataset. The proposed method was discussed and was compared on ten areas of plantar including toes, mid-foots and heels. For the clustering result, the experiments established superior performance with root mean square error (RMSE) of 70%, average accuracy of 80% and 80% time consuming. Furthermore, the proposed manifold learning method achieved an average accuracy of 87.2%, which was superior to other seven algorithms including multi-dimensional scaling (MDS), principal components analysis (PCA), locally linear embedding (LLE), Hessian LLE, Laplacian eigenmap method (LE), diffusion map, and local tangent space alignment (LTSA). The proposed approach established potential application on shoe-last customization of diabetic foot.

Publication Date

6-1-2017

Publication Title

Journal of Medical Imaging and Health Informatics

Volume

7

Issue

3

Number of Pages

639-651

Document Type

Article

Personal Identifier

scopus

DOI Link

https://doi.org/10.1166/jmihi.2017.2082

Socpus ID

85020031003 (Scopus)

Source API URL

https://api.elsevier.com/content/abstract/scopus_id/85020031003

This document is currently not available here.

Share

COinS