Characterization of an AI tool for identification of high-respiratory motion patients in cardiac SPECT
Bibliographic record
Abstract
Myocardial perfusion imaging (MPI) with SPECT is a common tool in the diagnosis of coronary artery disease. Image quality is degraded in the presence of respiratory motion (RM). Through a data-driven gating, retrospective motion compensation (MC) can be performed, but these methods are computationally intensive and time-consuming, making implementation in a clinical workflow impractical. In this work, an AI tool for rapid identification of patients that would benefit from MC is presented. This tool was trained on 2763 SPECT scans acquired from 1025 patients. Scans were retrospectively binned into respiratory gates and respiratory motion (RM) alignment vectors were found based on a minimized RMS difference between reconstructed gates. Using the AI model, extent of RM was estimated. An additional 773 MPI scans from 264 patients were reconstructed with and without MC. AI estimated motion extent was tested as a predictor of a change in perfusion score after MC. Using a motion cutoff of 10 mm, the tool showed an 80% true positive rate (TPR) for identifying stress scans that would show a significant change in perfusion score. Use of this tool takes less than 10 seconds. The tool was tested against sex- and state-based effects across rest, stress and prone SPECT MPI, and data were down-sampled to investigate the impact of noise on AI predictions. AI-estimated motion extent was shown to be best correlated to RMS values for stress scans (r2of 0.794 for stress compared to 0.683 and 0.629 for rest and prone). Predictions based on scans of female patients were also shown to have a weaker correlation with RMS values than male patients (0.662 averaged across states versus 0.696). The tool’s TPR rapidly decreased as noise increased. Overall, the tool was shown to reduce MC workload by up to 54% relative to correcting RM in all scans.
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 | 0.003 | 0.013 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 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 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
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".