Label efficient segmentation of single slice thigh CT with two-stage pseudo labels

Abstract

. Purpose: Muscle, bone, and fat segmentation from thigh images is essential for quantifying body composition. Voxelwise image segmentation enables quantification of tissue properties including area, intensity, and texture. Deep learning approaches have had substantial success in medical image segmentation, but they typically require a significant amount of data. Due to the high cost of manual annotation, training deep learning models with limited human label data is desirable, but it is a challenging problem. Approach: Inspired by transfer learning, we proposed a two-stage deep learning pipeline to address the thigh and lower leg segmentation issue. We studied three datasets, 3022 thigh slices and 8939 lower leg slices from the BLSA dataset and 121 thigh slices from the GESTALT study. First, we generated pseudo labels for thigh based on approximate handcrafted approaches using CT intensity and anatomical morphology. Then, those pseudo labels were fed into deep neural networks to train models from scratch. Finally, the first stage model was loaded as the initialization and fine-tuned with a more limited set of expert human labels of the thigh. Results: We evaluated the performance of this framework on 73 thigh CT images and obtained an average Dice similarity coefficient (DSC) of 0.927 across muscle, internal bone, cortical bone, subcutaneous fat, and intermuscular fat. To test the generalizability of the proposed framework, we applied the model on lower leg images and obtained an average DSC of 0.823. Conclusions: Approximated handcrafted pseudo labels can build a good initialization for deep neural networks, which can help to reduce the need for, and make full use of, human expert labeled data.

Semiautomated segmentation of lower extremity MRI reveals distinctive subcutaneous adipose tissue in lipedema: a pilot study

“…Fully-automated deep learning and atlas-based segmentation techniques may be impractical for applications in lipedema as a result of the uncommon diagnosis of lipedema that limits sample size, and the high anatomical variability in lower extremity soft tissue and vascular characteristics of this disease. It is possible that our classical approach could provide initial segmentations for a deep learning approach, similar to work by Yang et al 24 . for single slice thigh computed tomography images.…”

Single slice thigh CT muscle group segmentation with domain adaptation and self-training

“…We split one single CT slice into left and right thigh images with size 256×256 pixels by following the pipeline in Ref. 27. During preprocessing steps, 11 images were discarded due to low quality or abnormal anatomic appearance.…”

“…To reduce incorrect prediction induced by noise and appearance shift in synthetic images, we leverage muscle and bone masks from Ref. 27 to mask out erroneous predictions for the easy split as shown in Fig. 2(b).…”

Longitudinal variability analysis on low-dose abdominal CT with deep learning-based segmentation

“…1 Body composition measurements help monitor the changes related to growth and disease progressions. 2,3 Computed tomography is widely used to assess body composition. 4,5 Single abdominal slice is commonly selected to measure body distribution instead of CT volume to reduce radiation absorbed by human body.…”