Attention Based CNN-LSTM Network for Pulmonary Embolism Prediction on Chest Computed Tomography Pulmonary Angiograms

With more than 60,000 deaths annually in the United States, Pulmonary Embolism (PE) is among most fatal cardiovascular diseases. It caused by an artery blockage lung; confirming its presence time-consuming and prone to over-diagnosis. The utilization of automated PE detection systems critical for diagnostic accuracy efficiency. In this study we propose a two-stage attention-based CNN-LSTM network predicting PE, associated type (chronic, acute) corresponding location (leftsided, rightsided or central) on computed tomography (CT) examinations. We trained our model largest available public Computed Tomography Angiogram dataset (RSNA-STR CT (RSPECT) Dataset, N = 7279 studies) tested it in-house curated 106 studies. Our framework mirrors radiologic process via multi-slice approach so that pathologic sequela true pulmonary emboli may be meticulously assessed, enabling physicians better appraise morbidity when present. proposed method outperformed baseline CNN classifier single-stage network, achieving AUC 0.95 test set detecting study.