Category: Bone Densitometry or Body Composition

35 - 3D Optical Models for Predicting the Risk of Osteopnia, Sarcopenia, and Obesity in Women

Background: Sarcopenic obesity (OB-S) is a category of obesity in the elderly defined as excess adiposity with reduced muscle mass. OB-S increases the risk of fractures and falls in older individuals. However, three DXA scans are needed to determine muscle and bone status (whole body, hip, and spine). Furthermore, many obese patients are presumed to have high bone density due to the influence of high weight and may not be identified for any DXA scan. Whole body, three-dimensional optical (3DO) scanning is an automated, highly accessible technology shown to accurately predict fat and lean mass status. In this study, we investigate the ability of 3DO to predict the status of Osteopenia (OP), Obesity (OB), and Sarcopenia (S) in women.
Purpose. To create models for predicting OB, S, and OP status, individually and in combination, in women from 3DO scans.

The Risk Factors for Breast Cancer Lymphedema Study is a recruitment of 815 women that have had breast cancer and are at risk of developing lymphedema. Each woman received spine, hip, and whole body dual-energy X-ray absorptiometry (DXA) scans using a Hologic Horizon scanner (Hologic, Inc., Bedford, MA) and one 3DO scan using a Fit3D Proscanner (Fit3D, Redwood City, CA). Subjects were classified as sarcopenic if their appendicular skeletal mass index was lower than 5.45 kg/m2, obese if their body percent fat was greater than 40% body fat (BF), and osteopenic if their hip or spine T-score was ˂ -1.0. We noted the combinational categories of each condition for a total of 8 categories: normal metabolic status, OP, OB, S, OP-OB, OP-S, OB-S, and OP-OB-S. 3DO scans were registered to a standard CAESAR template following the methods of Allen et al. (1). Shape variation of 3DO scans was quantified using principal component analysis. Probabilities of OP, OB, and S were derived using logistic regression using the shape PCA descriptors. Visual PCA shape models were created for each category.

At this time, we have evaluated 85 women with mixed menopausal status, ethnicity, age = 60.6 ± 8.8 years, and BMI = 25.7 ± 4.9 kg/m2. The first 9 PCA components described 95% of the total shape variance. Logistic regression models combining these PCA modes for predicting OP, OB, and S status had AUROC values 0.78, 0.90, and 0.89 respectively. The model was able to distinguish sarcopenic obesity vs. non-sarcopenic obesity with an AUROC of 0.93. Visual PCA shape models are shown in Figure 1 with the number of women in each category. At 80% sensitivity, specificities of classifying OP, OB, S, and OB-S were 60%, 86%, 81%, and 88% respectively.

3DO whole body imaging shows promise as a fast, inexpensive and automated screening technology for identifying women with OB, S, and OB-S. However, classification of OP status was weak. Women with osteopenia do not look markedly different than women with normal bone status.

*Liu and Sommer had equal contribution in this abstract.

Bo Fan

San Francisco, California

Christine Miaskowski

University of California, San Francisco
San Francisco, California

Bennett K. Ng

PhD Candidate
UCSF Department of Radiology and Biomedical Imaging
San Francisco, California

John Shepherd

Adjunct Professor
San Francisco, California

Professor John Shepherd is a Professor in the Department of Radiology and the Director of the Body Composition, Exercise Physiology, and Energy Metabolism Lab at the University of California, San Francisco. He is a Fulbright fellow to the Karolinska Institute in Stockholm, and the current President of the International Society for Clinical Densitometry. He is an expert in quantitative breast imaging as well as musculoskeletal imaging using X-ray absorptiometry techniques. Dr. Shepherd received his BS in Engineering Physics from Texas Tech University and his PhD in Engineering Physics from the University of Virginia followed by a postdoctoral fellowship in Biophysics at Princeton University. He is also a Certified Clinical Densitometrist.

Dr. Shepherd’s research interests involve quantitative imaging methods for tissue composition using X-rays. He is the PI for the Shape Up! Study to examine 3D optical whole body scans in 1500 indivuduals from 5 to 85 years, and the PI of the 3CB study to extend mammography to measure the composition of invasive lesions. He has been the DXA CORE director for NHANES study since 1999. His current research interests include shape and appearance modeling and deep learning methods to big imaging datasets including DXA bone denisty scans. He has published over 150 peer reviewed articles in these fields.

Markus Sommer

Assistant CRC
San Francisco, California

En Yong Liu

Exercise Physiologist
University of California San Francisco
San Francisco, California