Author: Lindsay Hammons, Lisa Baumann Kreuziger, Haidy G. Nasief, Matthew Scheidt, Farrell Sean, Antonio Sosa Lozano 👨🔬
Affiliation: Division of Hematology and Oncology, University of Washington, Vascular and Interventional Radiology, Medical college of wisconsin, Department of Radiation Oncology, Medical College of Wisconsin 🌍
Purpose: Venous thromboembolism, which includes pulmonary embolism (PE), is the third leading cause of acute cardiovascular syndrome behind myocardial infarction and stroke. Current research categorizes acute PEs into low-risk, intermediate-risk (defined as symptomatic PE without cardiogenic shock), and high-risk (defined as symptomatic with cardiogenic shock) events. There is no consensus, however, among the many major clinical guidelines regarding the most effective intervention(s) for intermediate-risk PE, as that category is clinically too broad. Classically, Radiomics have been employed to develop a fingerprint associated with tissue microenvironments using advanced medical imaging. The purpose of this study is to develop a radiomic fingerprint of acute PE clot types and burden, which could be used to further risk-stratify intermediate-risk PEs.
Methods: In this study, 44 individual PE clots were identified from 18 patients and segmented in MIM Software™ by an experienced hematology-oncology fellow and cross-validated by a senior radiologist. The clots were divided into groups according to their origin (saddle(SD), main(MN), segmental, or sub_segmental(SSEG)). Radiomic features were extracted from each clot type. ANOVA and regression analysis were used to identify significant features to build classification model. Cross- validated Bi-Layered Neural networks were developed to predict different clot types. The performance of each model was judged by the CV-AUC under the ROC curve.
Results: The results showed that a Bi-Layered Neural network with three fully connected layers incorporating volume, max and Kurtosis can differentiate SD from SSEG with a CV-AUC of 0.98 and MN from SD with a CV-AUC of 0.96 and MN from SSEG with CV-AUC of 0.81.
Conclusion: Developing a Radiomic fingerprint to predict PE clot type is feasible. Large verification studies are needed to develop this into an invaluable tool to predict PE clot burden associated with right ventricular cardiac function to aid in determining type of intervention to for intermediate-risk clots.