← Back to Explore

Applications

JSHS · 2022

Overview

Sarcopenia, a condition distinguished by a reduction in skeletal muscle mass and function, significantly impacts the treatment success and survival-rate of over 50 million people. To diagnose sarcopenia in patients with head and neck cancer (HNC), skeletal muscle index (SMI) is measured from computed tomography (CT) of the cervical spine. However, this requires time-consuming manual segmentation which is unrobust, unquantified, and prone to inter-operator variability. In this research, a deep learning-based approach is developed for automatic sarcopenia diagnosis at the third cervical vertebrae level. Contrast-enhanced CT scans from 394 population- representative HNC patients at a single institution were utilized. First, single-slice selection and ground-truth skeletal muscle and adipose tissue annotations were manually performed on each scan. Then, a two-dimensional U-Net deep learning architecture, SarcoSeg, was built to auto-segment muscle and adipose from scans. SarcoSeg uses Sørensen–Dice coefficient generalized loss to alleviate class imbalance. From segmented cross-sectional area (CSA), SarcoSeg calculates SMI to predict sarcopenia status. Statistical analysis shows that SarcoSeg is a promising method for predicting sarcopenia from CT. Linear regression comparing SarcoSeg’s predicted segmentations to radiologist ground-truth segmentations of skeletal muscle and adipose yielded R-squared values of 0.96 and 1.00 respectively. In predicting sarcopenia, SarcoSeg’s specificity and positive predictive value were measured to be 0.98 and 0.96 respectively. Cohen’s kappa between the SarcoSeg sarcopenia diagnosis and ground-truth was measured to be 0.80, suggesting excellent agreement. This study enables oncologists to make quantified sarcopenia-related decisions for HNC patients towards precision care and improving quality-of-life and survival-rate.

Competition history

  • JSHS 2022 Category not listed

Resources

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Source: Junior Science and Humanities Symposium

Save projects to your library

Sign in with Google to keep track of projects you find interesting, organized into folders. An account also raises your daily allowance for “Has this been done?”, and lets you create a key for the MCP server with a much higher limit than anonymous use. Browsing stays public.

Continue with Google