The Latest Articles on Machine Learning and Parkinson's Disease
Our staff editors continue to share exciting, interesting, and thought-provoking reading material in the recommended articles series.
This week, we would like to share several latest articles on Machine Learning and Parkinson's Disease.
Title: Machine Learning Based Analysis of the Upper Limb Freezing During Handwriting in Parkinson's Disease Patients
Authors: Vassili Gorbatsov, Elli Valla, Sven Nõmm, Kadri Medijainen, Pille Taba, Aaro Toomela
Type: Research Article
Abstract:
Freezing of the upper limb in Parkinson's disease patients occurring during writing tests constitutes the research subject of the present paper. Digitisation of the writing and drawing tests coupled with artificial intelligence techniques have demonstrated accurate results in supporting the diagnostics of Parkinson's disease. In the digital setting, the analysis of freezing episodes did not get much attention. The main goal of the present paper is to determine if the neighbourhood of the point where freezing occurred possesses sufficient discriminating power to distinguish between the Parkinson's disease patients and healthy control individuals. For each freezing episode, time intervals of one second before and after are considered. These intervals are described by the hand movement's kinematic and pressure parameters. These parameters are used as features for the standard machine learning workflow that applies a nested cross-validation loop. The paper's main findings have demonstrated that analysis of the freezing neighbourhoods allows distinguishing Parkinson's disease patients from age matched healthy controls. The best results were achieved based on the movements occurring one second after the freezing episode. Kinematic and pressure-based features describing these movements have allowed training classifiers whose accuracy, precision, and recall have reached the values of 0.86, 0.86 and 0.93, respectively. Furthermore, the achieved results are comparable to those available in the literature.
Access this article: https://doi.org/10.1016/j.ifacol.2022.10.237
Title: Towards a telehealth infrastructure supported by machine learning on edge/fog for Parkinson's movement screening
Authors: Shehjar Sadhu, Dhaval Solanki, Nicholas Constant, Vignesh Ravichandran, Gozde Cay, Manob Jyoti Saikia, Umer Akbar, Kunal Mankodiya
Type: Research Article
Abstract:
Approximately 10 million people worldwide live with Parkinson's disease (PD), a progressive and incurable neurological movement disorder. Symptomatic treatment is available for PD but requires patients to make periodic clinic visits (2–3 times per year) for symptom assessment. Advanced telehealth technologies can enhance clinical care for PD but warrant an Internet-of-Things-based (IoT) infrastructure that can enable symptom monitoring in out-of-clinic settings, such as homes. In this paper, we present an edge and Fog device-based IoT framework and a machine learning-based telehealth infrastructure that can detect and classify hand movement tasks based on a clinical test (UPDRS) for remote symptom assessment. We used a pair of smart gloves integrated with finger flex sensors, an inertial measurement unit (IMU), and a wireless embedded system (i.e., an edge device) to record the hand movements. The edge device (ESP32) detected the activity on the edge node and transmitted the data to the Fog node for classification. The Fog node (Raspberry pi) hosted the Machine Learning (ML) based activity classification models to classify UPDRS-based hand movement tasks. In this paper, we present the development of edge-fog-supported ML infrastructure. To develop ML models, we utilized the data of 9 participants (five healthy and four people with PD) who performed hand movements tasks while wearing the smart gloves. We developed and tested different classification models such as K-nearest neighbors (KNN), Support Vector Machine (SVM), and Decision Tree (DT) to identify which one fits best for such an edge-fog infrastructure for remote symptom assessment. Results showed that the SVM model outperformed others by giving 94% training, 94% testing, and 93% validation accuracy with a mean inference time of 560 μs on the Fog node. Our preliminary results are promising to further our research in deploying the edge-fog-driven ML-based telehealth infrastructure in real-world settings.
Access this article: https://doi.org/10.1016/j.smhl.2022.100351
Title: Generative Adversarial Networks as a Data Augmentation Tool for CNN-Based Parkinson's Disease Diagnostics
Authors: Erik Dzotsenidze, Elli Valla, Sven Nõmm, Kadri Medijainen, Pille Taba, Aaro Toomela
Type: Research Article
Abstract:
Growing research interest has arisen towards automated neurodegenerative disease diagnostics based on the information extracted from the digital drawing tests. Since the performance of modern modelling techniques (machine learning, deep learning) relies heavily on the size of training data available, data scarcity is one of the most significant problems in computer-aided diagnostics. This paper proposes using Generative Adversarial Networks to synthesise digital drawing tests acquired from Parkinson's patients and healthy controls. Four different architectures (StyleGAN2-ADA, StyleGAN2-ADA + LeCam, StyleGAN3 and ProjectedGAN) are evaluated and compared with the traditional data augmentation methods. Convolutional neural networks are utilised for Parkinson's disease diagnostics. Our results indicate that GAN-generated images’ addition outperforms the standard augmentation methods in classifying Parkinson's disease in some settings. Therefore, the proposed framework could serve as a potential decision support tool for clinicians in computer-aided fine-motor analysis for neurodegenerative disease diagnostics.
Access this article: https://doi.org/10.1016/j.ifacol.2022.10.240
Title: Parkinson's disease gene prioritising using an efficient and biologically appropriate network-based consensus strategy
Authors: Baby Kumari, Pankaj Singh Dholaniya
Type: Research Article
Abstract:
The complex multifactorial diseases such as Parkinson’s disease (PD) are characterised by two factors: genomics and the variability in environment. The goal of studying the control vs diseases cohorts is to understand that what biological pathways are perturbed in the manifestation of the disease. Although many genetic mutations are associated with the disease, they have varying rate of risk among different patients. This is due the different genomic landscape and environmental conditions that varies from individual to individual. This raises a question that how different set of mutations are linked to the manifestation of the same disease? This can be answered by identifying the consensus gene/biological pathways that are perturbed under disease condition even if the disease has manifested through different set of mutations. To understand this we propose a network-based strategy for gene prioritisation that combines the knowledge of disease associated mutations and the expression pattern of genes in different samples from patients and healthy-controls. The network comprises of three layers, first being the genes with PD associated mutations. The third or target layer comprises of the core pathways associated with PD pathology and the middle layer is composed interacting partners between these two layers. A scoring mechanism for edges was developed using gene expression profiles across different samples and the busiest routes/paths were discovered between the first and the third layer. The results highlight a subset of genes from the middle layer of the network and could classify between the control vs patients sample using machine learning algorithm and are significantly associated with PD pathways.
Access this article: https://doi.org/10.1016/j.jocs.2022.101879
Title: Non-REM sleep with hypertonia in Parkinsonian Spectrum Disorders: A pilot investigation
Authors: Daniel J. Levendowski, Daniel J. Levendowskia, Bradley F. Boeve, Debby Tsuang, Joanne M. Hamilton, David Salat, Chris Berka, Joyce K. Lee-Iannotti, David Shprecher, Philip R. Westbrook, Gandis Mazeika, Leslie Yack, Sarah Payne, Paul C. Timm, Thomas C. Neylan, Erik K.St. Louis
Type: Research Article
Abstract:
Introduction
From an ongoing multicenter effort toward differentiation of Parkinsonian spectrum disorders (PSD) from other types of neurodegenerative disorders, the sleep biomarker non-rapid-eye-movement sleep with hypertonia (NRH) emerged.
Methods
This study included in the PSD group patients with dementia with Lewy bodies/Parkinson disease dementia (DLB/PDD = 16), Parkinson disease (PD = 16), and progressive supranuclear palsy (PSP = 13). The non-PSD group included patients with Alzheimer disease dementia (AD = 24), mild cognitive impairment (MCI = 35), and a control group with normal cognition (CG = 61). In-home, multi-night Sleep Profiler studies were conducted in all participants. Automated algorithms detected NRH, characterized by elevated frontopolar electromyographic power. Between-group differences in NRH were evaluated using Logistic regression, Mann-Whitney U and Chi-squared tests.
Results
NRH was greater in the PSD group compared to non-PSD (13.9 ± 11.0% vs. 3.1 ± 4.7%, P < 0.0001). The threshold NRH ≥ 5% provided the optimal between-group differentiation (AUC = 0.78, P < 0.001). NRH was independently associated with the PSD group after controlling for age, sex, and SSRI/SNRI use (P < 0.0001). The frequencies of abnormal NRH by subgroup were PSP = 92%, DLB/PDD = 81%, PD = 56%, MCI = 26%, AD = 17%, and CG = 16%. The odds of abnormal NRH in each PSD subgroup ranged from 3.7 to 61.2 compared to each non-PSD subgroup. The night-to-night and test-retest intraclass correlations were excellent (0.78 and 0.84, both P < 0.0001).
Conclusions
In this pilot study, NRH appeared to be a novel candidate sleep biomarker for PSD-related neurodegeneration. Future studies in larger cohorts are needed to confirm these findings, understand the etiology of NRH magnitude/duration, and determine whether it is an independent prodromal marker for specific neurodegenerative pathologies.
Access this article: https://doi.org/10.1016/j.sleep.2022.09.025
