Machine Learning Algorithm Accurately Diagnoses Parkinsonian Syndromes, Study Shows

In a recent study presented at the 2023 International Congress of Parkinson’s Disease and Movement Disorders, held August 27-31, in Copenhagen, Denmark, findings showed that a trained machine learning algorithm for PET scans achieved comparably high specificities and sensitivities in the diagnosis of Parkinsonian syndromes in both patient populations from the United States and Slovenia.¹ Since the classifiers performed similarly, the results suggest that metabolic patterns recognized at 1 institution can be used successfully at different sites for Parkinsonian syndromes.²

In the analysis, both set patterns of the model used for the US (86%) and Slovenia (85%) populations had successful reached a high diagnostic accuracy overall. Notably, the 2 models also had specificity and sensitivity of 82–83% and 94%, respectively, for Parkinson disease (PD), 96% and 70–74% for progressive supranuclear palsy (PSP), and 94–95% and 72% in multiple system atrophy (MSA).

READ MORE: IncobotulinumtoxinA Shows Long-Term Treatment Effect on Parkinson Disease Sialorrhea

“It’s become quite clear that there are reproducible differences in overall patterns of FDG-PET between parkinsonian syndromes, and that these changes can be seen very early in the disease process,” Ronald Postuma, MD, MSc, professor of neurology at the Montreal Neurological Institute, said in a statement.¹ In this study, researchers aimed to test an automated differential diagnostic algorithm for parkinsonian syndromes based on machine learning and FDG-PET scans using subject scores for sets of validated disease patterns from 2 independent sites.

Investigators assessed 265 FDG PET scans from the 2 sites in Slovenia and the United States among patients with parkinsonism syndromes (PD, n = 161; MSA, n = 57; PSP, n = 47) where the final clinical diagnosis was unknown at the time of the imaging assessment. Of note, the machine learning model was used according to the support vector machine and included 2 sets of features. These features comprised of expression values for 3 previously validated disease patterns recognized in Slovenian cohorts,^3,4 and analogous expression values for the original US cohorts.^5,6

“One of the difficulties in the field is the number of different patterns that have been described, which are often partially overlapping. This overlap suggests that machine learning procedures may be very useful, as they do not necessarily need to prespecify a specific pattern. The diagnostic performance here, although not perfect, is nonetheless quite encouraging, and equals that of many other biomarker/ imaging based diagnostic tools,” Postuma said in a statement.¹

In clinical assessment, FDG PET has been utilized to recognize the different disease-specific metabolic patterns of PD especially for the major parkinsonian syndromes like MSA and PSP. Using corresponding disease patterns, early differential diagnosis can be enhanced,⁷ but the authors in this study proved that the diagnosis accuracy can be further improved with using a machine learning approach.

Click here for more coverage of MDS 2023.

REFERENCES
1. Machine learnings has potential to assist with diagnosis in Parkinsonian syndromes. News Release. International Parkinson and Movement Disorder Society. Published August 27, 2023. Accessed August 28, 2023. https://www.movementdisorders.org/MDS/News/Newsroom/News-Releases/2023-Congress/ML-parkinsonisms
2. M. Perovnik, T. Rus, A. Vo, N. Nguyen, P. Tomše, J. Jamšek, C. Tang, M. Trošt, D. Eidelberg. Machine learning diagnosis of parkinsonian syndromes: network approach with two different sites [abstract]. Mov Disord. 2023; 38 (suppl 1). https://www.mdsabstracts.org/abstract/machine-learning-diagnosis-of-parkinsonian-syndromes-network-approach-with-two-different-sites/. Accessed August 29, 2023.
3. Tomše P, Jensterle L, Grmek M, Zaletel K, Pirtošek Z, Dhawan V, et al. Abnormal metabolic brain network associated with Parkinson’s disease: replication on a new European sample. Neuroradiology 2017;59:507–15. https://doi.org/10.1007/s00234-017-1821-3.
4. Tomše P, Rebec E, Studen A, Perovnik M, Rus T, Ležaić L, et al. Abnormal metabolic covariance patterns associated with multiple system atrophy and progressive supranuclear palsy. Phys Medica 2022;98:131–8. https://doi.org/10.1016/j.ejmp.2022.04.016.
5. Tang CC, Poston KL, Eckert T, Feigin A, Frucht S, Gudesblatt M, et al. Differential diagnosis of parkinsonism: a metabolic imaging study using pattern analysis. Lancet Neurol 2010;9:149–58. https://doi.org/10.1016/S1474-4422(10)70002-8.
6. Rus T, Tomše P, Jensterle L, Grmek M, Pirtošek Z, Eidelberg D, et al. Differential diagnosis of parkinsonian syndromes: a comparison of clinical and automated – metabolic brain patterns’ based approach. Eur J Nucl Med Mol Imaging 2020;47:2901–10. https://doi.org/10.1007/s00259-020-04785-z.
7. Perovnik M, Rus T, Schindlbeck KA, Eidelberg D. Functional brain networks in the evaluation of patients with neurodegenerative disorders. Nat Rev Neurol 2023;19:73–90. https://doi.org/10.1038/s41582-022-00753-3.