A recent study out of UCLA seemed to add to the body of evidence supporting AI-assisted CCTA as an alternative to SPECT for diagnosing obstructive coronary artery disease, although holes in the study suggest that CCTA AI has more to prove
- SPECT continues to be a first-line test to diagnose coronary ischemia, often instead of or before invasive coronary angiography.
- However, SPECT’s cost, necessary radiotracers, lower image quality, and radiation exposure have raised doubts about its continued dominance.
To determine if AI-CCTA could hold its own, researchers analyzed CCTA images from 175 patients using Cleerly’s AI-QCT ISCHEMIA model and found it outperformed SPECT when compared to invasive angiography.
- AI-CCTA showed a sensitivity of 75% and specificity of 70% for predicting coronary ischemia, outperforming SPECT (70% & 53%).
- AI-enhanced CCTA also surpassed SPECT’s negative predictive value among female patients (91% vs. 68%).
- However, AI-CCTA’s AUCs were only slightly higher than SPECT (0.81 vs. 0.76).
One important caveat is how the researchers reached these results, as one of our readers pointed out that “coronary angiography was used as the gold standard for comparison” but only 16 patients in the study received ICA.
- As a result, comparative analysis would only be based on those 16 patients.
- Yet, the study reports sample sizes of 17 and 25 for sensitivity and specificity, and also varied on the NPV endpoint.
- This led to all of the proportions reported not matching the study’s final sensitivity and specificity results.
Even though the study misses the mark and AI is still somewhat unproven, AI-CCTA has been gaining momentum versus SPECT in recent years.
- CCTA’s biggest step forward was the 2021 AHA/ACC chest pain guidelines that established it as a front-line coronary artery disease test, potentially instead of SPECT.
- Then a JACC study from 2023 found that AI-ISCHEMIA was superior to SPECT and similar to PET and FFRCT in ischemia prediction (AUCs: 0.91 vs. 0.71).
- Another 2024 JACC study found the AI-ISCHEMIA + CCTA approach more accurately diagnoses coronary ischemia (AUCs: 0.80 vs. 0.72).
The Takeaway
AI assisted CCTA is still relatively new and in some ways unproven, but its growing usage, new reimbursements, cost-effectiveness, and greater accuracy are worth keeping in mind as the modality matures in comparison to more established scans like SPECT. However, it will take stronger results than this study in order to drive a more meaningful shift in exam volumes.
