Feasibility of Determining Low Skeletal Muscle Mass Using Digital Images of Lower Extremities in Elderly Women

A Preliminary Study Using Convolutional Neural Network Classification and Edge Detection

Abstract

Objective: To preliminarily test whether digital images of the lower leg and edge detection can be used to determine elderly women with low Skeletal Muscle Index (SMI).

Methods: An elderly woman who was hospitalized was the subject of this study. Digital images of the subject's lower leg taken with a digital camera were used for edge detection using the Canny method. The criteria for low SMI were grouped using the criterion of 5.7 kg/m² proposed by the Asian Working Group, and analysis by convolutional neural network was performed on digital and edge-detected images of the lower leg, respectively.

Results: Thirty-two subjects were included in the study. The C statistic for determining low SMI in digital and edge-detected images of the lower leg was 0.83 (95% CI: 0.83–1.00) and 0.92 (95% CI: 0.92–1.00), respectivel.

Conclusion: Digital images of the lower leg may be used to classify those with low SMI.