Joint feature-sample selection and robust diagnosis of Parkinson's disease from MRI data

Abstract

Parkinson's disease (PD) is an overwhelming neurodegenerative disorder caused by deterioration of a neurotransmitter, known as dopamine. Lack of this chemical messenger impairs several brain regions and yields various motor and non-motor symptoms. Incidence of PD is predicted to double in the next two decades, which urges more research to focus on its early diagnosis and treatment. In this paper, we propose an approach to diagnose PD using magnetic resonance imaging (MRI) data. Specifically, we first introduce a joint feature-sample selection (JFSS) method for selecting an optimal subset of samples and features, to learn a reliable diagnosis model. The proposed JFSS model effectively discards poor samples and irrelevant features. As a result, the selected features play an important role in PD characterization, which will help identify the most relevant and critical imaging biomarkers for PD. Then, a robust classification framework is proposed to simultaneously de-noise the selected subset of features and samples, and learn a classification model. Our model can also de-noise testing samples based on the cleaned training data. Unlike many previous works that perform de-noising in an unsupervised manner, we perform supervised de-noising for both training and testing data, thus boosting the diagnostic accuracy. Experimental results on both synthetic and publicly available PD datasets show promising results. To evaluate the proposed method, we use the popular Parkinson's progression markers initiative (PPMI) database. Our results indicate that the proposed method can differentiate between PD and normal control (NC), and outperforms the competing methods by a relatively large margin. It is noteworthy to mention that our proposed framework can also be used for diagnosis of other brain disorders. To show this, we have also conducted experiments on the widely-used ADNI database. The obtained results indicate that our proposed method can identify the imaging biomarkers and diagnose the disease with favorable accuracies compared to the baseline methods.

Keywords: Diagnosis; Joint feature-sample selection; Matrix completion; Parkinson's disease; Robust linear discriminant analysis; Sparse regression.

Fig. 1

An illustration of the brain regions affected by PD in different stages of the disease. Darker blue denotes the earlier and more severely affected regions.

Fig. 2

Overview of our proposed method. First, the MR images are processed and tissue-segmented. Then, the anatomical automatic labeling (AAL) atlas is non-linearly registered to each subject’s original MR image, and then the WM, GM and CSF volumes of each ROI are calculated as features. These features form X and the corresponding labels form Y. Through our proposed joint feature-sample selection (JFSS), we discard some uninformative features and samples, leading to $\widehat{\mathbf{X}}$ and $\widehat{\mathbf{Y}}$. Then, we train a robust classifier (i.e.,, Robust LDA), in which we jointly decompose $\widehat{\mathbf{X}}$ into cleaned data $\widehat{\mathbf{D}}$ and its noise component E, and classify the cleaned data.

Fig. 3

All 98 ROIs used in this study: 90 ROIs from the AAL atlas (Tzourio-Mazoyer et al., 2002), 4 ROIs defined in brainstem, 2 ROIs in substantial nigra (L/R), and 2 ROIs in red nucleus (L/R).

Fig. 4

Comparisons of results on synthetic data, for three different runs. The diagram shows the mean accuracy for different methods as a function of ρ.

Fig. 5

Accuracy as a function of the JFSS hyperparameter (λ₁), experimented on synthetic data, with ρ=100.

Fig. 6

Comparisons of results by the proposed (JFSS + RLDA) and the baseline methods.

Fig. 7

Accuracy as a function of the JFSS hyperparameter (λ₁), for the Parkinson’s disease diagnosis experiment.

Fig. 8

Accuracy of the competing methods, as a function of their hyperparameter.

Fig. 9

Top and most frequent selected ROIs by our method.