Back

When Diagnosis Is Made By Playing Shuffled Cards: A Non-Invasive, Real-Time, Adaptive Method to Functionally Evaluate Coronary Artery Based on Coronary Angiography

Dai, Y.; Zhu, P.; Xue, B.; Ling, Y.; Shi, X.; Geng, L.; Xie, Y.; Zhang, Q.; Liu, J.

2022-03-02 cardiovascular medicine
10.1101/2022.03.01.22271738 medRxiv
Show abstract

AimsThe great value concealed in the sequence of coronary angiography frames is not discovered in the world. We discovered and demonstrated the "Sequence Value" concealed in the coronary angiography and proposed DZL (Disarranged Zone Learning) to realize the sequence value in functional evaluation of coronary artery disease. Furthermore, we automated the DZL using a deep learning model to release huge medical resources. Methods and ResultsWe gave a novel definition of TIMI flow grade using the term temporal and spatial coupling tightness (TSCT) of the antegrade contrast agent. We used the TSCT to model the myocardial ischemia in a functional perspective and used the PCI conduction after CAG as the outcome event of myocardial ischemia. We proposed a novel method (Disarranged Zone Learning) to measure TSCT and we designed an experiment to validate its effectiveness. We further automated the novel method using an unsupervised deep learning model. The prediction accuracy of the model was applied as a proxy of myocardial ischemia. We further proposed Difference DZL to quantify the functional capability of any specific vessel segment. DZL overall AUC reaches 0.92. DZL automation reveals an AUC of 0.84 (95%CI, 0.81-0.87). ConclusionWe unprecedentedly discovered the "Sequence Value" concealed in coronary angiography. We then proposed a novel method termed DZL to functionally evaluate the coronary artery in a non-invasive, real-time and adaptive manner. DisclosureThe Authors declare that there is no conflict of interest.

Matching journals

The top 9 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.