Machine Learning for Detecting Parkinson’s Disease by Resting-State Functional Magnetic Resonance Imaging: A Multicenter Radiomics Analysis

Abstract

Parkinson’s disease (PD) is one of the most common progressive degenerative diseases, and its diagnosis is challenging on clinical grounds. Clinically, effective and quantifiable biomarkers to detect PD are urgently needed. In our study, we analyzed data from two centers, the primary set was used to train the model, and the independent external validation set was used to validate our model. We applied amplitude of low-frequency fluctuation (ALFF)-based radiomics method to extract radiomics features (including first- and high-order features). Subsequently, t-test and least absolute shrinkage and selection operator (LASSO) were harnessed for feature selection and data dimensionality reduction, and grid search method and nested 10-fold cross-validation were applied to determine the optimal hyper-parameter λ of LASSO and evaluate the performance of the model, in which a support vector machine was used to construct the classification model to classify patients with PD and healthy controls (HCs). We found that our model achieved good performance [accuracy = 81.45% and area under the curve (AUC) = 0.850] in the primary set and good generalization in the external validation set (accuracy = 67.44% and AUC = 0.667). Most of the discriminative features were high-order radiomics features, and the identified brain regions were mainly located in the sensorimotor network and lateral parietal cortex. Our study indicated that our proposed method can effectively classify patients with PD and HCs, ALFF-based radiomics features that might be potential biomarkers of PD, and provided further support for the pathological mechanism of PD, that is, PD may be related to abnormal brain activity in the sensorimotor network and lateral parietal cortex.

Using histogram analysis of the intrinsic brain activity mapping to identify essential tremor

“…On account of this advantage, a few studies have achieved good classi cation performance for identifying Parkinson's disease from HCs or Alzheimer's Disease from HCs [12,[28][29][30]. Meanwhile, due to the shines of predicting the individual subject and the multivariate nature of machine learning algorithms, radiomics analysis has been successfully used for neurological disease research and provided quantitative and objective support for clinical diagnosis and disease prognosis [29,31,32]. Combined clinical symptoms or brain gray matter volumes with machine learning algorithms, few studies achieved good classi cation performance in classifying ET from HCs [33][34][35][36].…”

Combined brain network topological metrics with machine learning algorithms to identify essential tremor

“…Meanwhile, other models also achieved mean AUCs over 0.8 with different advantages. For example, SVM incorporates several advantageous properties to reduce overfitting and deliver good generalization performance despite a small sample size ( Mo et al, 2019 ; Shi et al, 2022b ), RF has strong adaptability to the data on account of ensemble strategy and the prediction accuracy is relatively accurate ( Chong-Wen et al, 2022 ), and NB performs well on small-scale data and has stable classification efficiency. However, this is only a preliminary and small-scale study, and the generalizability of the model across different groups, cultures, and ethnicities needs to be further verified before our model can be applied to clinical work.…”

Neuroimaging biomarkers for detecting schizophrenia: A resting-state functional MRI-based radiomics analysis

“…We extracted 48 radiomics features for each VOI, including 15 intensity-based histogram and 33 texture features, which measure the intensity of gray levels and the spatial dissimilarity of the intensity levels for all voxels in VOI, respectively. Additional details are described elsewhere ( Aerts et al., 2014 ; Feng et al., 2018 ; Shi et al., 2022 ; Zhao et al., 2020 ) and in Supplementary Methods . Finally, a total of 23,616 features [(15 + 33) × 246 × 2 = 23616] were extracted from each participant.…”

“…Radiomics, an emerging method of medical image analysis, has rapidly developed to identify brain signatures for neuropsychological diseases aiding in diagnosis, prediction, and identification of physiological mechanisms ( Feng et al., 2018 ; Shi et al., 2021b , 2022 ; Sun et al., 2018 ; Zhao et al., 2020 ). In radiomics analysis, medical images can be transformed into high-throughput, quantitative, and mineable features, including some invisible to the human visual system, by a series of data characterization algorithms ( Gillies et al., 2016 ; Lambin et al., 2012 ).…”

“…Machine learning techniques are also used to perform data mining on extracted features. Currently, there is growing use of rs-fMRI-based radiomics analysis for exploring neuroimaging biomarkers of neuropsychiatric disorders, including AD ( Feng et al., 2018 ; Zhao et al., 2020 ), attention deficit hyperactivity disorder ( Sun et al., 2018 ), hippocampal sclerosis ( Mo et al., 2019 ), and PD ( Shi et al., 2021b , Shi et al., 2022 ). To our knowledge, no studies have assessed SZ diagnosis with DC- and VMHC-based radiomics.…”