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

Purpose

Thigh muscle group segmentation is important for assessing muscle anatomy, metabolic disease, and aging. Many efforts have been put into quantifying muscle tissues with magnetic resonance (MR) imaging, including manual annotation of individual muscles. However, leveraging publicly available annotations in MR images to achieve muscle group segmentation on single-slice computed tomography (CT) thigh images is challenging.

Approach

We propose an unsupervised domain adaptation pipeline with self-training to transfer labels from three-dimensional MR to single CT slices. First, we transform the image appearance from MR to CT with CycleGAN and feed the synthesized CT images to a segmenter simultaneously. Single CT slices are divided into hard and easy cohorts based on the entropy of pseudo-labels predicted by the segmenter. After refining easy cohort pseudo-labels based on anatomical assumption, self-training with easy and hard splits is applied to fine-tune the segmenter.

Results

On 152 withheld single CT thigh images, the proposed pipeline achieved a mean Dice of 0.888 (0.041) across all muscle groups, including gracilis, hamstrings, quadriceps femoris, and sartorius muscle.

Conclusions

To our best knowledge, this is the first pipeline to achieve domain adaptation from MR to CT for thigh images. The proposed pipeline effectively and robustly extracts muscle groups on two-dimensional single-slice CT thigh images. The container is available for public use in GitHub repository available at: https://github.com/MASILab/DA_CT_muscle_seg.

Skeletal muscle analyses: agreement between non-contrast and contrast CT scan measurements of skeletal muscle area and mean muscle attenuation

Low skeletal muscle area (SMA) and muscle radiation attenuation (MRA) have been associated with poor prognosis in various patient populations. Both non-contrast and contrast CT scans are used to determine SMA and MRA. The effect of the use of a contrast agent on SMA and MRA is unknown. Therefore, we investigated agreement between these two scan options. SMA and MRA of 41 healthy individuals were analysed on a paired non-contrast and contrast single CT scan, and agreement between paired scan results was assessed with use of Bland-Altman plots, intraclass correlation coefficients (ICCs), standard error of measurements (SEM) and smallest detectable differences at a 95% confidence level (SDD₉₅ ). Analyses were stratified by tube voltage. Difference in SMA between non-contrast and contrast scans made with a different tube voltage was 7·0 ± 7·5 cm² ; for scans made with the same tube voltage this was 2·3 ± 1·7 cm² . Agreement was excellent for both methods: ICC: 0·952, SEM: 7·2 cm² , SDD₉₅ : 19·9 cm² and ICC: 0·997, SEM: 2·0 cm² , SDD₉₅ : 5·6 cm² , respectively. MRA of scans made with a different tube voltage differed 1·3 ± 11·3 HU, and agreement was poor (ICC: 0·207, SEM: 7·9 HU, SDD₉₅ : 21·8 HU). For scans made with the same tube voltage the difference was 6·7 ± 3·2 HU, and agreement was good (ICC: 0·682, SEM: 5·3 HU, SDD₉₅ : 14·6 HU). In conclusion, SMA and MRA can be slightly influenced by the use of contrast agent. To minimise measurement error, image acquisition parameters of the scans should be similar.