Related Experiment Video
Updated: May 6, 2026
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
Hippocampal functional imaging-derived radiomics features for diagnosing cognitively impaired patients with
Wei Zeng1,2, Xiao Liang1,2, Jiali Guo3
1Jiangxi Provincial Key Laboratory for Precision Pathology and Intelligent Diagnosis, Department of Radiology, the First Affiliated Hospital, Jiangxi Medical College, Nanchang University, 17 Yongwaizheng Street, Nanchang, 330006, People's Republic of China.
Purpose:
The aim of this retrospective study was to investigate whether radiomics features derived from hippocampal functional imaging can effectively differentiate cognitively impaired patients from cognitively preserved patients with Parkinson's disease (PD).
Methods:
The study included a total of 89 clinically definite PD patients, comprising 55 who werecognitively impaired and 34 who were cognitively preserved. All participants underwent functional magnetic resonance imaging and clinical assessments. Preprocessed functional data were utilized to derive the amplitude of the low-frequency fluctuations (ALFF), regional homogeneity (ReHo), voxel-mirrored homotopic connectivity (VMHC), and degree centrality (DC). A standardized set of radiomics features was subsequently extracted from the bilateral hippocampi, resulting in a total of 819 features. Following feature selection, the radiomics score (rad-score) and logistic regression (LR) models were trained. Additionally, the Shapley additive explanations (SHAP) algorithm was employed to elucidate and interpret the predictions made by the LR models. Finally, the relationships between the radiomics features derived from hippocampal functional imaging and the scores of the clinical measures were explored to assess their clinical significance.
Results:
The rad-score and LR algorithm models constructed using a combination of wavelet features extracted from ReHo and VMHC data exhibited superior classification efficiency. These models demonstrated exceptional accuracy, sensitivity, and specificity in distinguishing cognitively impaired PD patients (CI-PD) from cognitively preserved PD (CP-PD) patients, with values of 0.889, 0.900, and 0.882, respectively. Furthermore, SHAP values indicated that wavelet features derived from ReHo and VMHC were critical for classifying CI-PD patients. Importantly, our findings revealed significant associations between radiomics wavelet features and scores on the Hamilton Anxiety Scale, Non-Motor Symptom Scale, and Montreal Cognitive Assessment in CI-PD patients (P < 0.05, with Bonferroni correction).
Conclusions:
Our novel rad-score model and LR model, which utilize radiomics features derived from hippocampal functional imaging, have demonstrated their value in diagnosing CI-PDpatients. These models can enhance the accuracy and efficiency of functional MRI diagnosis, suggesting potential clinical applications.
Clinical Trial Number:
Not applicable.
