Early prediction of putamen imaging features in HIV-associated neurocognitive impairment syndrome

Abstract

Background: To explore the correlation between the volume of putamen and brain cognitive impairment in patients with HIV and to predict the feasibility of early-stage HIV brain cognitive impairment through radiomics.

Method: Retrospective selection of 90 patients with HIV infection, including 36 asymptomatic neurocognitive impairment (ANI) patients and 54 pre-clinical ANI patients in Beijing YouAn Hospital. All patients received comprehensive neuropsychological assessment and MRI scanning. 3D Slicer software was used to acquire volume of interest (VOI) and radiomics features. Clinical variables and volume of putamen were compared between patients with ANI and pre-clinical ANI. The Kruskal Wallis test was used to analysis multiple comparisons between groups. The relationship between cognitive scores and VOI was compared using linear regression. For radiomics, principal component analysis (PCA) was used to reduce model overfitting and calculations and then a support vector machine (SVM) was used to build a binary classification model. For model performance evaluation, we used an accuracy, sensitivity, specificity and receiver operating characteristic curve (ROC).

Result: There were no significant differences in clinical variables between ANI group and pre-clinical-ANI group (P>0.05). The volume of bilateral putamen was significantly different between AHI group and pre-clinical group (P<0.05), but there was only a trend in the left putamen between ANI-treatment group and pre-clinical treatment group(P = 0.063). Reduced cognitive scores in Verbal Fluency, Attention/Working Memory, Executive Functioning, memory and Speed of Information Processing were negatively correlated with the increased VOI (P<0.05), but the correlation was relatively low. In diagnosing the ANI from pre-clinical ANI, the mean area under the ROC curves (AUC) were 0.85 ± 0.22, the mean sensitivity and specificity were 63.12 ± 5.51 and 94.25% ± 3.08%.

Conclusion: The volumes of putamen in patients with ANI may be larger than patients with pre-clinical ANI, the change of the volume of the putamen may have a certain process; there is a relationship between putamen and cognitive impairment, but the exact mechanism is unclear. Radiomics may be a useful tool for predicting early stage HAND in patients with HIV.

Keywords: Asymptomatic neurocognitive impairment; HIV; Putamen; Radiomics.

Fig. 1

The process of segmentation. a VOIs were drawn in all axial MRI images. b VOI masks extracted from MRI scans

Fig. 2

The relationship of the putamen between groups. *Fig. 2:1 = ANI-treatment group, 2 = ANI-treatment naive group,3 = non-ANI treatment group,4 = non-ANI treatment naive group. a:multiple comparisons between groups in the right putamen; b:multiple comparisons between groups in the left putamen; c:multiple comparisons between groups in the bilateral putamen

Fig. 3

ROC curves analysis of radiomics signatures for differentiating ANI