Is Dimensionality Reduction Indispensable for Radiomics Analysis? Construction a New Radiomics Analysis Scheme Independent of Dimension Reduction

Abstract

Background: Dimension reduction has always been considered indispensable to analysis. However, its non-substitutability may still cause major challenges. In this study, we tried to propose a new radiomics analysis scheme independent of dimension reduction, and evaluated its clinical application value by comparing it with the general analysis scheme dependent on dimension reduction. Methods: The non-contrast computed tomography (NCCT) images of 274 acute brainstem infarction (ABI) were analyzed using radiomic features algorithm in Matlab R2013a, and two types of radiomic indicators were established (with global feature management and dimension reduction). Multivariate logistic regression was used to evaluate the abilities of ABI detection. Multiple linear regression was used to investigate the correlations amongst clinical characteristics. And, the abilities to predict the prognosis of ABI were evaluated by the generalized linear model (GLM) and multivariate logistic regression. Findings: Both types of schemes could effectively detect ABI, and showed significant correlations amongst clinical characteristics, as well as predicted the prognosis of ABI. And, although the ABI detecting ability of the new sheme was slightly weaker than that of the general analysis scheme (AUC: 0.893 vs. 0.835), it presented a stronger performance in predicting prognosis (AUC: 0.864, OR = 0.917, p = 0.021 vs. AUC:0.806, OR = 0.972, p = 0.007). Interpretation: This study proved that the scheme of global feature management could be a powerful supplement to the radiomics methodology. Also, this study provided an unprecedented suggestion to effectively improve decision-support in ABI using radiomics. Funding Statement: This work was supported by National Natural Science Foundation of China (81871343, 81525014), Jiangsu Provincial Key Research and Development Plan (BE2018693), and Natural Science Foundation of Jiangsu Province (BK20181226). Declaration of Interests: The authors declare that they have no competing interests. Ethics Approval Statement: This study was approved by the institutional review board of Jiangsu University (JSDXPROSPERO-20180898), and written informed consent was obtained from all patients, or their legal guardians.