A deep learning model based on the attention mechanism for automatic segmentation of abdominal muscle and fat for body composition assessment

Abstract

Background: Quantitative muscle and fat data obtained through body composition analysis are expected to be a new stable biomarker for the early and accurate prediction of treatment-related toxicity, treatment response, and prognosis in patients with lung cancer. The use of these biomarkers can enable the adjustment of individualized treatment regimens in a timely manner, which is critical to further improving patient prognosis and quality of life. We aimed to develop a deep learning model based on attention for fully automated segmentation of the abdomen from computed tomography (CT) to quantify body composition.Methods: A fully automatic segmentation deep learning model was designed based on the attention mechanism and using U-Net as the framework. Subcutaneous fat, skeletal muscle, and visceral fat were manually segmented by two experts to serve as ground truth labels. The performance of the model was evaluated using Dice similarity coefficients (DSCs) and Hausdorff distance at 95th percentile (HD95). Results:The mean DSC for subcutaneous fat and skeletal muscle were high for both the enhanced CT test set (0.93±0.06 and 0.96±0.02, respectively) and the plain CT test set (0.90±0.09 and 0.95±0.01, respectively).Nevertheless, the model did not perform well in the segmentation performance of visceral fat, especially for the enhanced CT test set. The mean DSC for the enhanced CT test set was 0.87±0.11, while the mean DSC for the plain CT test set was 0.92±0.03. We discuss the reasons for this result.Conclusions: This work demonstrates a method for the automatic outlining of subcutaneous fat, skeletal muscle, and visceral fat areas at L3.

Automatic deep learning method for third lumbar selection and body composition evaluation on CT scans of cancer patients

“…Analyzing body composition is a challenging task due to varied methods used (39). When comparing similar populations who have undergone automated body composition analysis at the L3 level, the average DICE scores for SAT, VAT, and SM are all >0.90 (26,30,32,33,(40)(41)(42)(43)(44)(45), which is very similar to our study. Some studies have attempted to identify lipid infiltration within muscle as an indicator of muscle quality, but the difficulty is obtaining accurate measurements and whose clinical relevance remains uncertain (27,46).…”

“…One popular deep learning model for medical image segmentation is U-net, which is designed for semantic segmentation tasks and has been successfully applied to body composition analysis (26,27,32,36). The U-net architecture includes a contracting path for feature extraction and a symmetric expanding path for precise localization, which allows for accurate segmentation of complex anatomic structures.…”

“…However, manual measurement of body composition through CT scans is time-consuming and requires expertise, which limits its practical use in daily clinical practice. Artificial intelligence using deep learning provides an opportunity to automate muscle mass assessment with high precision (26)(27)(28)(29)(30)(31), but validated tools for research and integration into the electronic medical record are scarce. Accordingly, studies proposing algorithms for detecting body composition from L3 have been conducted on homogeneous populations with few challenging cases, such as patients undergoing cementoplasty or those with intra-corporeal devices, which are common in the cancer population, especially among metastatic patients.…”

Simultaneous Learning of Erector Spinae Muscles for Automatic Segmentation of Site-Specific Skeletal Muscles in Body CT Images

“…This index is used because the skeletal muscle mass in the L3 cross-section correlates with the total body skeletal muscle mass [8]. Previously, the semi-automatic segmentation of skeletal muscles in L3 cross-sections has been performed using software-based thresholding [9], and automatic segmentation has been performed using deep learning [10,11,12]. In the automatic segmentation of skeletal muscles using deep learning, multiple skeletal muscles within the L3 cross-section are segmented as a single region [10,11].…”

“…Previously, the semi-automatic segmentation of skeletal muscles in L3 cross-sections has been performed using software-based thresholding [9], and automatic segmentation has been performed using deep learning [10,11,12]. In the automatic segmentation of skeletal muscles using deep learning, multiple skeletal muscles within the L3 cross-section are segmented as a single region [10,11]. In other words, the erector spinae, quadratus lumborum, psoas major, external oblique abdominis, internal oblique abdominis, transverse abdominis, and rectus abdominis were segmented into the same skeletal muscle region without any distinction.…”