AUTHOR=Su Hang , Han Zhengyuan , Fu Yujie , Zhao Dong , Yu Fanhua , Heidari Ali Asghar , Zhang Yu , Shou Yeqi , Wu Peiliang , Chen Huiling , Chen Yanfan 

TITLE=Detection of pulmonary embolism severity using clinical characteristics, hematological indices, and machine learning techniques

JOURNAL=Frontiers in Neuroinformatics

VOLUME=Volume 16 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/neuroinformatics/articles/10.3389/fninf.2022.1029690

DOI=10.3389/fninf.2022.1029690

ISSN=1662-5196

ABSTRACT=Pulmonary embolism (PE) is a potentially life-threatening cardiopulmonary illness. PE can range asymptomatic blood clots to large emboli that can occlude the pulmonary arteries causing sudden cardiovascular collapse and death. In context with the potential life-threat caused by PE, it is necessary to conduct risk stratification after the diagnosis of PE to adjust follow-up treatment. We captured clinical characteristics, blood routine data and arterial blood gas analysis data from all 139 patients. Combining these data, this paper proposes a PE risk stratified prediction framework based on machine learning technology. The prediction framework is proposed by combining the improved binary cuckoo algorithm (SBCS) and the kernel extreme learning machine (KELM). In this paper, we conduct benchmark function experiments to verify the overall performance of SBCS. Then, to verify the performance of SBCS on the feature selection problem, experiments based on seven public data sets are conducted in this paper. To further demonstrate the usefulness and applicability of the SBCS-KELM framework, this paper conducts aided diagnosis experiments on PE data collected from hospital. The results of the experiments reveal that the chosen indicators, such as syncope, systolic blood pressure (SBP), oxygen saturation (SaO2%), white blood cell (WBC), neutrophil percentage (NEUT%), and others, are critical for the feature selection method proposed in this study to determine the severity of PE. The classification results showed that the accuracy of the prediction model was 99.26% and its sensitivity was 98.57%. It is expected to become a new and accurate method to distinguish the severity of PE.