AI and Parkinson's Disease

Blog Author: Archita Khaire

Parkinson's disease (PD) is a neurodegenerative disorder caused by damage to dopamine-producing neurons in the mid-brain. Dopamine neurons originate in the midbrain and are found in either the substantia nigra or the ventral tegmental area, which is located adjacent to the substantia nigra.
When the neurons in the substantia nigra are damaged in large numbers, the loss of dopamine prevents normal function in basal ganglia and causes the motor symptoms of Parkinson's Disease tremor, rigidity, impaired balance, and loss of spontaneous movement.

Image Credit: https://parkinsonsdisease.net/

Currently there is no cure for Parkinson's disease. It doesn’t always affect how long you live but has a severely negative impact on the quality of life of patients and their caregivers. Early detection of PD can improve the symptoms dramatically. 

 Image Credit: https://www.pcori.org/research-results/pcori-stories/improving-life-women-parkinsons-disease

Early Detection of PD using AI
There are multiple early signs that can indicate onset of Parkinson's disease. Health data collected around these signs is used to build machine learning algorithms that can predict the PD. 
1. Speech impairments like dysphonia (defective use of the voice), hypophonia (reduced volume), monotone (reduced pitch range), and dysarthria (difficulty with articulation of sounds or syllables) are used to detect PD.  UCI Machine Learning Repository: Parkinsons Telemonitoring Data Set was created by Athanasios Tsanas and Max Little of the University of Oxford, in collaboration with 10 medical centers in the US and Intel Corporation who developed the tele-monitoring device to record the speech signals.

2. Parkinson's tremors usually start on one side of the body, commonly in the hands, and progress to the other side.​ Scientists have developed a Convolutional Neural Network (CNN), which aims at learning features from a signal extracted during the individual's exam by means of a smart pen composed of a series of sensors that can extract information from handwritten dynamics.

Image Credit: DOI:10.1109/SIBGRAPI.2016.054
​
3. Difficulty in walking is another sign of growing Parkinson's disease. Deep learning algorithms are used in Gait analysis to classify the movement properties into two classes PD and non-PD using features such as imbalance, frequency of falls etc.

4. Researchers are attempting to detect PD using human olfactory data. In patients with OD, loss of smell starts many years early before onset of motor skill symptoms. There is potential to diagnose PD much early with this method.

Leveraging AI to treat PD
Parkinson’s is caused by the death of dopamine-producing nerve cells. Scientists at Stanford school of medicine identified Miro1, mitochondrial protein which resists removal of damaged cells. They analyzed millions of drugs using AI to identify a compound that could bind with Miro and enhance the cell mechanism to remove damaged nerve cells.
​
Researchers at John Hopkins have identified the glucagon-like peptide receptor, or GLP1 receptor as potential receptor for slowing down PD. They are leveraging to AI to check if diabetes drug could be used to prevent neurological cell deaths.
​​

References:

• https://www.parkinson.org/
• https://parkinsonsdisease.net/
• Sajal, M.S.R., Ehsan, M.T., Vaidyanathan, R. et al. Telemonitoring Parkinson’s disease using machine learning by combining tremor and voice analysis. Brain Inf. 7, 12 (2020). https://doi.org/10.1186/s40708-020-00113-1
• https://github.com/pqrst/ParkinsonsDiseaseDataAnalysis
• R., Pereira & Weber, Silke & Hook, Christian & de Rosa, Gustavo & Papa, João. (2016). Deep Learning-Aided Parkinson's Disease Diagnosis from Handwritten Dynamics. 10.1109/SIBGRAPI.2016.054. 
• Juutinen, M., Wang, C., Zhu, J., Haladjian, J., Ruokolainen, J., Puustinen, J., & Vehkaoja, A. (2020). Parkinson's disease detection from 20-step walking tests using inertial sensors of a smartphone: Machine learning approach based on an observational case-control study. PloS one, 15(7), e0236258. https://doi.org/10.1371/journal.pone.0236258.
• Lo C, Arora S, Ben-Shlomo Y, Barber TR, Lawton M, Klein JC, Kanavou S, Janzen A, Sittig E, Oertel WH, Grosset DG, Hu MT. Olfactory Testing in Parkinson Disease and REM Behavior Disorder: A Machine Learning Approach. Neurology. 2021 Apr 13;96(15):e2016-e2027. doi: 10.1212/WNL.0000000000011743. Epub 2021 Feb 24. PMID: 33627500; PMCID: PMC8166425.
• Prashanth R, Roy SD, Mandal PK, Ghosh S. 2014. Parkinson’s disease detection using olfactory loss and REM sleep disorder features. Conf Proc IEEE Eng Med Biol Soc. 2014:5764–5767.
• Kaur, R., Chen, Z., Motl, R., Hernandez, M. E., & Sowers, R. (2021). Predicting Multiple Sclerosis From Gait Dynamics Using an Instrumented Treadmill: A Machine Learning Approach. IEEE Transactions on Biomedical Engineering, 68(9), 2666-2677. doi:10.1109/tbme.2020.3048142
• ​Wattendorf, E., Welge-Lussen, A., Fiedler, K., Bilecen, D., Wolfensberger, M., Fuhr, P., . . . Westermann, B. (2009). Olfactory Impairment Predicts Brain Atrophy in Parkinsons Disease. Journal of Neuroscience, 29(49), 15410-15413. doi:10.1523/jneurosci.1909-09.2009
• Lötsch, J., Kringel, D., & Hummel, T. (2018, October 27). Machine Learning in Human Olfactory Research. Retrieved from https://academic.oup.com/chemse/article/44/1/11/5145702
• ​Hsieh CH;Li L;Vanhauwaert R;Nguyen KT;Davis MD;Bu G;Wszolek ZK;Wang X;. (n.d.). Miro1 Marks Parkinson's Disease Subset and Miro1 Reducer Rescues Neuron Loss in Parkinson's Models. Retrieved from https://pubmed.ncbi.nlm.nih.gov/31564441/​