Brain and Behavior (Nov 2023)

Textural features of the frontal white matter could be used to discriminate amnestic mild cognitive impairment patients from the normal population

  • Wei Zheng,
  • Ronghua Mu,
  • Fuzhen Liu,
  • Xiaoyan Qin,
  • Xin Li,
  • Peng Yang,
  • Xin Li,
  • Yahui Liang,
  • Xiqi Zhu

DOI
https://doi.org/10.1002/brb3.3222
Journal volume & issue
Vol. 13, no. 11
pp. n/a – n/a

Abstract

Read online

Abstract Objective We aim to develop a radiomics model based on 3‐dimensional (3D)‐T1WI images to discriminate amnestic mild cognitive impairment (aMCI) patients from the normal population by measuring changes in frontal white matter. Methods In this study, 126 patients with aMCI and 174 normal controls (NC) were recruited from the local community. All subjects underwent routine magnetic resonance imaging examination (including 3D‐T1WI ). Participants were randomly divided into a training set (n = 242, aMCI:102, NC:140) and a testing set (n = 58, aMCI:24, NC:34). Texture features of the frontal lobe were extracted from 3D‐T1WI images. The least absolute shrinkage and selection operator (LASSO) was used to reduce feature dimensions and develop a radiomics signature model. Diagnostic performance was assessed in the training and testing sets using the receiver operating characteristic (ROC) curve analysis. The area under the ROC curve (AUC), sensitivity, and specificity were also calculated. The efficacy of the radiomics model in discriminating aMCI patients from the normal population was assessed by decision curve analysis (DCA). Results A total of 108 frontal lobe texture features were extracted from 3D‐T1WI images. LASSO selected 58 radiomic features for the final model, including log‐sigma (n = 18), original (n = 8), and wavelet (n = 32) features. The performance of radiomic features extracted from 3D T1 imaging for distinguishing aMCI patients from controls was: in the training set, AUC was 1.00, and the accuracy, sensitivity, and specificity were 100%, 98%, and 100%, respectively. In the testing set, AUC was 0.82 (95% CI:0.69–0.95), and the accuracy, sensitivity, and specificity were 69%, 92%, and 55%, respectively. The DCA demonstrated that the model had favorable clinical predictive value. Conclusions Textural features of white matter in the frontal lobe showed potential for distinguishing aMCI from the normal population, which could be a surrogate protocol to aid aMCI screening in clinical setting.

Keywords