Automated body composition analysis of clinically acquired computed tomography scans using neural networks

BACKGROUND & AIMS

The quantity and quality of skeletal muscle and adipose tissue is an important prognostic factor for clinical outcomes across several illnesses. Clinically acquired computed tomography (CT) scans are commonly used for quantification of body composition, but manual analysis is laborious and costly. The primary aim of this study was to develop an automated body composition analysis framework using CT scans.

METHODS

CT scans of the 3rd lumbar vertebrae from critically ill, liver cirrhosis, pancreatic cancer, and clear cell renal cell carcinoma patients, as well as renal and liver donors, were manually analyzed for body composition. Ninety percent of scans were used for developing and validating a neural network for the automated segmentation of skeletal muscle and adipose tissues. Network accuracy was evaluated with the remaining 10 percent of scans using the Dice similarity coefficient (DSC), which quantifies the overlap (0 = no overlap, 1 = perfect overlap) between human and automated segmentations.

RESULTS

Of the 893 patients, 44% were female, with a mean (±SD) age and body mass index of 52.7 (±15.8) years old and 28.0 (±6.1) kg/m², respectively. In the testing cohort (n = 89), DSC scores indicated excellent agreement between human and network-predicted segmentations for skeletal muscle (0.983 ± 0.013), and intermuscular (0.900 ± 0.034), visceral (0.979 ± 0.019), and subcutaneous (0.986 ± 0.016) adipose tissue. Network segmentation took ~350 milliseconds/scan using modern computing hardware.

CONCLUSIONS

Our network displayed excellent ability to analyze diverse body composition phenotypes and clinical cohorts, which will create feasible opportunities to advance our capacity to predict health outcomes in clinical populations.

Clinically Practical Approach for Screening of Low Muscularity Using Electronic Linear Measures on Computed Tomography Images in Critically Ill Patients

BACKGROUND

Computed tomography (CT) scans performed during routine hospital care offer the opportunity to quantify skeletal muscle and predict mortality and morbidity in intensive care unit (ICU) patients. Existing methods of muscle cross-sectional area (CSA) quantification require specialized software, training, and time commitment that may not be feasible in a clinical setting. In this article, we explore a new screening method to identify patients with low muscle mass.

METHODS

We analyzed 145 scans of elderly ICU patients (≥65 years old) using a combination of measures obtained with a digital ruler, commonly found on hospital radiological software. The psoas and paraspinal muscle groups at the level of the third lumbar vertebra (L3) were evaluated by using 2 linear measures each and compared with an established method of CT image analysis of total muscle CSA in the L3 region.

RESULTS

There was a strong association between linear measures of psoas and paraspinal muscle groups and total L3 muscle CSA (R² = 0.745, P < 0.001). Linear measures, age, and sex were included as covariates in a multiple logistic regression to predict those with low muscle mass; receiver operating characteristic (ROC) area under the curve (AUC) of the combined psoas and paraspinal linear index model was 0.920. Intraclass correlation coefficients (ICCs) were used to evaluate intrarater and interrater reliability, resulting in scores of 0.979 (95% CI: 0.940-0.992) and 0.937 (95% CI: 0.828-0.978), respectively.

CONCLUSIONS

A digital ruler can reliably predict L3 muscle CSA, and these linear measures may be used to identify critically ill patients with low muscularity who are at risk for worse clinical outcomes.

Skeletal muscle predicts ventilator-free days, ICU-free days, and mortality in elderly ICU patients

INTRODUCTION

As the population ages, the number of injured elderly is increasing. We sought to determine if low skeletal muscle mass adversely affected outcome in elderly patients following trauma.

METHODS

Patients ≥ 65 years of age with an admission abdominal computed tomography scan and requiring intensive care unit (ICU) stay at a Level I trauma center in 2009-2010 were reviewed. Muscle cross-sectional area at the 3rd lumbar vertebra was quantified and muscle index, a normalized measure of muscle mass, was calculated and related to clinical parameters including ventilator-free days, ICU-free days, and mortality. Using previously established sex-specific, muscle index cut-points, patients were then categorized as sarcopenic or non-sarcopenic and differences in clinical outcomes between these two groups were also compared. We also examined muscle index as a continuous variable relative to the same clinical outcomes.

RESULTS

There were 149 severely injured elderly patients (median age 79 years) enrolled in this study of which 71% were sarcopenic. Of the patients who were sarcopenic, 9% were underweight, 44% normal weight, and 47% overweight/obese as per body mass index (BMI) classifications. The overall mortality rate was 27% and univariate analysis demonstrated higher mortality among those who were sarcopenic (32% vs. 14%, P = 0.018). After controlling for age, sex, and injury severity, multiple logistic regression demonstrated that increased muscle index was significantly associated with decreased mortality (OR per unit muscle index = 0.93, 95% CI: 0.875-0.997, P = 0.025). In addition, multivariate linear regression showed that sarcopenia, but not muscle index, was associated with decreased ventilator-free (P = 0.004) and ICU-free days (P = 0.002). Neither BMI, serum albumin nor total adipose tissue on admission were indicative of survival, ventilator-free or ICU-free days.

CONCLUSIONS

Sarcopenia is highly prevalent in the elderly population with traumatic injuries. Traditional measures of nutritional assessment, such as BMI and serum albumin, do not accurately predict outcome in the injured elderly. Sarcopenia, however, represents a potential new predictor for mortality, discharge disposition, and ICU utilization. Measurement of muscularity allows for the early identification of at-risk patients who may benefit from aggressive and multidisciplinary nutritional and rehabilitative strategies.