Artificial Intelligence–Enhanced Perfusion Scoring Improves the Diagnostic Accuracy of Myocardial Perfusion Imaging
Bibliographic record
Abstract
We previously demonstrated that a deep learning (DL) model of myocardial perfusion SPECT imaging improved accuracy for detection of obstructive coronary artery disease (CAD). We aimed to improve the clinical translatability of this artificial intelligence (AI) approach using the results to derive enhanced total perfusion deficit (TPD) and 17-segment summed scores. Methods: We used a cohort of patients undergoing myocardial perfusion imaging within 180 d of invasive coronary angiography. Obstructive CAD was defined as any stenosis of at least 70% or at least 50% in the left main coronary artery. We used per-vessel DL predictions to modulate polar map pixel scores. These transformed polar maps were then used to derive TPD-DL and summed stress score–DL. We compared diagnostic performance using area under the receiver operating characteristic curve (AUC). Results: In the 555 patients held out for testing, the median age was 65 y (interquartile range, 57–73 y), and 381 (69%) were male. Obstructive CAD was present in 329 (59%) patients. The prediction performance for obstructive CAD of stress TPD-DL (AUC, 0.837; 95% CI, 0.804–0.870) was higher than AI prediction alone (AUC, 0.795; 95% CI, 0.758–0.831; P = 0.005) and traditional stress TPD (AUC, 0.737; 95% CI, 0.696–0.778; P < 0.001). Summed stress score–DL had the second highest prediction performance (AUC, 0.822; 95% CI, 0.788–0.857) and higher AUC than traditional quantitative summed stress score (AUC, 0.728; 95% CI, 0.686–0.769; P < 0.001). At a threshold of 5%, the sensitivity and specificity of TPD rose from 72% to 79% and from 62% to 70%, respectively. Conclusion: Integrating AI predictions with traditional quantitative approaches leads to a simplified AI approach, presenting clinicians with familiar measures but operating with higher accuracy than traditional quantitative scoring. This approach may facilitate integration of new AI methods into clinical practice.
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 | 0.002 | 0.007 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| 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".