Page 36 - AIH-2-4

Artificial Intelligence in Health Early Parkinson’s detection through CNNs

Despite the strong performance of the proposed Author contributions
models, it is not without any limitations. First, the model
was trained and evaluated solely on the PPMI dataset, This is a single-authored article.
and its generalizability to external datasets remains to Ethics approval and consent to participate
be validated. Second, the slice selection and averaging
strategies used in this study are based on fixed indices that Not applicable.
are applicable in the PPMI data, which may not optimally Consent for publication
capture relevant features in all subjects in other datasets.
Third, while CNNs perform well, they are inherently Not applicable.
black-box models, making it difficult to interpret specific
feature-driven decisions. Finally, although promising Availability of data
results were achieved in this controlled research setup, Not applicable.
further validation is necessary before deployment in
clinical environments. Further disclosure
5. Conclusion This work was carried out independently by the author.
The author is currently employed at Siemens Healthineers,
Accurate and early detection of PD is a challenging clinical Bangalore, India, however, the views expressed and the
problem. The numerous common symptoms shared by work presented here are solely those of the author and do
this class of Parkinsonism disorders represent a source not reflect the views of the company.
of misdiagnosis. Accurate identification of degenerative
Parkinsonism from other non-degenerative ones is crucial References
for effective patient treatment and management. In this 1. Jankovic J. Parkinson’s disease: Clinical features and
work, machine learning models that could classify early diagnosis. J Neurol Neurosurg Psychiatry. 2008;79(4):368-376.
PD from healthy normal and also SWEDD from PD using doi: 10.1136/jnnp.2007.131045
SPECT images were developed. These machine learning
models exhibit very high performance, of which the CNN 2. Fahn S. Description of Parkinson’s disease as a clinical
model especially achieves the best performance. These syndrome. Ann N Y Acad Sci. 2003;991:1-14.
predictive models carry enormous potential to be used in a doi: 10.1111/j.1749-6632.2003.tb07458.x
clinical setting and can act as an aid the clinical diagnostic 3. Tysnes OB, Storstein A. Epidemiology of Parkinson’s disease.
process. J Neural Transm (Vienna). 2017;124(8):901-905.

Acknowledgments doi: 10.1007/s00702-017-1686-y
4. Marek K, Jennings D, Lasch S, et al. The Parkinson
None.
progression marker initiative (PPMI). Prog Neurobiol.
Funding 2011;95(4):629-635.
doi: 10.1016/j.pneurobio.2011.09.005
PPMI, a public-private partnership, is funded by the
Michael J. Fox Foundation for Parkinson’s Research 5. Groveman BR, Orrù CD, Hughson AG, et al. Rapid
and other funding partners include AbbVie, Allergan, and ultra-sensitive quantitation of disease-associated
Amathus Therapeutics, Avid Radiopharmaceuticals, α-synuclein seeds in brain and cerebrospinal fluid by αSyn
RT-QuIC. Acta Neuropathol Commun. 2018;6:7.
Biogen Idec, BioLegend, Bristol-Myers Squibb, Celgene,
Denali Therapeutics, GE Healthcare, Genentech, doi: 10.1186/s40478-018-0508-2
GlaxoSmithKline, Eli Lilly and Company, Lundbeck, 6. Booij J, Tissingh G, Boer GJ, et al. [123I]FP-CIT SPECT
Merck & Co., Meso Scale Discovery, Pfizer, Piramal, Prevail shows a pronounced decline of striatal dopamine transporter
Therapeutics, Hoffmann-La Roche, Sanofi Genzyme, labelling in early and advanced Parkinson’s disease. J Neurol
Servier, Takeda Pharmaceutical Company, Teva, Verily Neurosurg Psychiatry. 1997;62(2):133-140.
Life Sciences, Voyager Therapeutics, and UCB (Union doi: 10.1136/jnnp.62.2.133
ChimiqueBelge).
7. Booth TC, Nathan M, Waldman AD, Quigley AM,
Conflict of interest Schapira AH, Buscombe J. The role of functional dopamine-
transporter SPECT imaging in Parkinsonian syndromes,
The author declares no competing interests. part 1. AJNR Am J Neuroradiol. 2015;36(2):236-244.

Volume 2 Issue 4 (2025) 30 doi: 10.36922/AIH025040005