Development of a fully automatic deep learning system for L3 selection and body composition assessment on computed tomography

Abstract

As sarcopenia research has been gaining emphasis, the need for quantification of abdominal muscle on computed tomography (CT) is increasing. Thus, a fully automated system to select L3 slice and segment muscle in an end-to-end manner is demanded. We aimed to develop a deep learning model (DLM) to select the L3 slice with consideration of anatomic variations and to segment cross-sectional areas (CSAs) of abdominal muscle and fat. Our DLM, named L3SEG-net, was composed of a YOLOv3-based algorithm for selecting the L3 slice and a fully convolutional network (FCN)-based algorithm for segmentation. The YOLOv3-based algorithm was developed via supervised learning using a training dataset (n = 922), and the FCN-based algorithm was transferred from prior work. Our L3SEG-net was validated with internal (n = 496) and external validation (n = 586) datasets. Ground truth L3 level CT slice and anatomic variation were identified by a board-certified radiologist. L3 slice selection accuracy was evaluated by the distance difference between ground truths and DLM-derived results. Technical success for L3 slice selection was defined when the distance difference was < 10 mm. Overall segmentation accuracy was evaluated by CSA error and DSC value. The influence of anatomic variations on DLM performance was evaluated. In the internal and external validation datasets, the accuracy of automatic L3 slice selection was high, with mean distance differences of 3.7 ± 8.4 mm and 4.1 ± 8.3 mm, respectively, and with technical success rates of 93.1% and 92.3%, respectively. However, in the subgroup analysis of anatomic variations, the L3 slice selection accuracy decreased, with distance differences of 12.4 ± 15.4 mm and 12.1 ± 14.6 mm, respectively, and with technical success rates of 67.2% and 67.9%, respectively. The overall segmentation accuracy of abdominal muscle areas was excellent regardless of anatomic variation, with CSA errors of 1.38–3.10 cm2. A fully automatic system was developed for the selection of an exact axial CT slice at the L3 vertebral level and the segmentation of abdominal muscle areas.

Multi-source data approach for personalized outcome prediction in lung cancer screening: update from the NELSON trial

“…The new multi-source models for participant selection and lung nodule discrimination will be externally validated on fractures [25]. Body composition will be evaluated as areas of subcutaneous fat and muscle, that will be semi-automatically measured with in-house developed software [26].…”

“…For body composition, an in-house (UMCU) developed automated AI algorithm will be used. Although not (yet) CE-marked, we have performed these measurements successfully in non-contrast CT scans and now have measurements in more than 1000 subjects [26].…”

Prognostic Value of Sarcopenia and Myosteatosis in Patients with Resectable Pancreatic Ductal Adenocarcinoma

“…Muscle quantity and quality were measured in a single slice of an axial CT image of the portal venous phase following the automatic selection of the CT slice at the inferior endplate level of the L3 vertebra ( Fig. 1 ) [ 23 ]. All skeletal muscles (psoas, paraspinal, transversus abdominis, rectus abdominis, quadratus lumborum, internal oblique, and external oblique muscles) in the selected image were automatically segmented using a convolutional neural network (AID-U™, iAID Inc.) with a Dice similarity coefficient of 0.96–0.97 [ 24 ].…”

Lifestyle interventions after colorectal cancer surgery using a mobile digital device: A study protocol for a randomized controlled trial

“…Body composition on CT will be evaluated with artificial intelligence software (AID-U™, iAID Inc, Seoul, Korea), a fully automatic deep learning system for both third lumbar vertebra (L3) selection and body composition assessment. [ 21 ] Two experienced operators (YK and KWK) will check the quality of segmentation results in all L3 level segmentation images. Skeletal muscle area including all muscles on the selected axial images, that is, psoas, paraspinal, transversus abdominis, rectus abdominis, quadratus lumborum, and internal and external obliques, the visceral fat area, and the subcutaneous fat area will be measured.…”