Automated Assessment of Sarcopenia and Body Composition from Cardiac Magnetic Resonance Scans
Bibliographic record
Abstract
Background: Sarcopenia, defined as age-related loss of muscle mass, is a risk factor for mortality and morbidity in patients with cardiovascular disease (CVD).Methods to quantify muscle mass are not always available in clinical workflows, however, cardiac magnetic resonance (CMR) provides high-resolution axial imaging of the thorax, which can be examined to assess body composition.This study aims to utilize deep learning segmentation on CMR scans to develop an automated technique for assessing body composition and sarcopenia.Methods: We assembled a retrospective cohort of adult patients that underwent clinically indicated CMR exams at multiple centers.We extracted the axial black-blood HASTE images as acquired on 1.5 T and 3 T Siemens scanners, with approximately 20 slices spanning the entire thoracic cavity.We manually segmented the skeletal muscle and subcutaneous fat tissues in a representative subset of 50 CMR scans (990 axial slices).We then used these segmentations to retrain a deep learning (DL) model that had previously undergone validation on over 1,000 patient CT scans (known as 'transfer learning').We applied additional active learning training techniques, which yielded a final training set consisting of 100 CMR exams (2024 axial slices).The final DL model, DeepSarcMR, utilized a U-Net structure with 4 skip connections and 17.2M parameters, and we trained it using the Adam optimizer along with a cross-entropy and Dice loss function.Clinical statistical analysis was performed with the predicted thoracic skeletal muscle and subcutaneous fat volumes.CVD diagnoses were obtained through ICD-10 codes.Results: Ten-fold cross validation yielded 0.961 average accuracy on the validation set.We used the best fold with 0.964 accuracy to perform inference on a total of 2506 CMR scans.Within the included cohort (N=2027), the mean skeletal muscle volume was 3104 ± 510 cm 3 and 2053 ±
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".