Objective: Alzheimer's disease, an irreversible neurological condition, demands timely diagnosis for effective clinical intervention. This study employs radiomics analysis to assess image features in default mode network cerebral perfusion imaging among individuals with cognitive impairment.
Methods: A radiomics analysis of cerebral perfusion imaging was conducted on 117 patients with cognitive impairment. They were divided into training and validation sets in a 7:3 ratio. Least Absolute Shrinkage and Selection Operator (LASSO) and Random Forest were employed to select and model image features, followed by logistic regression analysis of LASSO and Random Forest results. Diagnostic performance was assessed by calculating the area under the curve (AUC).
Results: In the training set, LASSO achieved AUC of 0.978, Random Forest had an AUC of 0.933. In the validation set, LASSO had AUC of 0.859, Random Forest had AUC of 0.986. By conducting Logistic Regression analysis in combination with LASSO and Random Forest, we identified a total of five radiomics features, with four related to morphology and one to textural features, originating from the medial prefrontal cortex and middle temporal gyrus. In the training set, Logistic Regression achieved AUC of 0.911, while in the validation set, it attained AUC of 0.925.
Conclusion: The medial prefrontal cortex and middle temporal gyrus are the two brain regions within the default mode network that hold the highest significance for Alzheimer's disease diagnosis. Radiomics analysis contributes to the clinical assessment of Alzheimer's disease by delving into image data to extract deeper layers of information.
Keywords: Alzheimer's disease; Cerebral blood flow perfusion imaging; Default mode network (DMN); Radiomics.
© 2024 The Authors. Journal of Applied Clinical Medical Physics published by Wiley Periodicals, LLC on behalf of The American Association of Physicists in Medicine.