Clinical effects of physiologic lesion testing in influencing treatment strategy for multi-vessel coronary artery disease.

Background: The application of fractional flow reserve (FFR) and instantaneous wave-free ratio (iFR) in multivessel coronary artery disease (CAD) patients has not been definitively explored. We herein assessed how treatment strategies were decided based on FFR/iFR values in vessels selected clinically. Specifically, we sought to determine whether treatment selection was based on whether the vessel tested was the clinical target stenosis. Methods: 270 consecutive patients with angiographically determined multivessel disease who underwent FFR/iFR testing were included. Patients were classified initially based on their angiographic findings, then re-evaluated from FFR/iFR results (normal or abnormal). Tested lesions were classified into target or non-target lesions based on clinical and non-invasive evaluations. Results: Abnormal FFR/iFR values were demonstrated in 51.9 % of patients, in whom 51.4 % received coronary stenting (PCI) and 44.3 % had bypass surgery (CABG). With two-vessel CAD patients, medical therapy was preferred when the target lesion was normal (72.6 %), while PCI was preferred when it was abnormal (78.4 %). In non-target lesions, PCI was preferred regardless of FFR/iFR results (78.0 %). With three-vessel CAD patients, CABG was preferred when the target lesion was abnormal (68.5 %), and there was no difference in the selected modality when it was normal. Furthermore, the incidence of tested lesions was higher in the left anterior descending (LAD) compared to other coronary arteries, and two-vessel CAD patients with LAD stenoses were more frequently treated by PCI. Conclusion: The use of invasive physiologic testing in multivessel CAD patients may alter the preferred treatment strategy, leading to an overall increase in PCI selection.

Deep learning generates synthetic cancer histology for explainability and education.

Artificial intelligence methods including deep neural networks (DNN) can provide rapid molecular classification of tumors from routine histology with accuracy that matches or exceeds human pathologists. Discerning how neural networks make their predictions remains a significant challenge, but explainability tools help provide insights into what models have learned when corresponding histologic features are poorly defined. Here, we present a method for improving explainability of DNN models using synthetic histology generated by a conditional generative adversarial network (cGAN). We show that cGANs generate high-quality synthetic histology images that can be leveraged for explaining DNN models trained to classify molecularly-subtyped tumors, exposing histologic features associated with molecular state. Fine-tuning synthetic histology through class and layer blending illustrates nuanced morphologic differences between tumor subtypes. Finally, we demonstrate the use of synthetic histology for augmenting pathologist-in-training education, showing that these intuitive visualizations can reinforce and improve understanding of histologic manifestations of tumor biology.