Home Medizin Deep Learning wirft ein neues Licht auf die Parkinson-Erkennung durch das Auge

Deep Learning wirft ein neues Licht auf die Parkinson-Erkennung durch das Auge

von NFI Redaktion

A recent scientific study discusses the potential of retinal fundus imaging as a diagnostic screening method for Parkinson’s disease (PD). The study, entitled „Deep Learning predicts the incidence and prevalence of Parkinson’s disease using fundus imaging from the UK Biobank,“ was published in Scientific Reports. The image source is credited to recep art / Shutterstock.com.

Background:
Parkinson’s disease is associated with a gradual decline in motor control and several non-motor symptoms, attributed to the progressive loss of dopaminergic neurons in the substantia nigra of the brain. The number of PD-related deaths has more than doubled since 2000, largely due to a lack of high-quality interventions in the elderly. Further research is needed to better understand the pathology of Parkinson’s disease and develop early diagnostic systems.

Retinal imaging, often referred to as a „window to the brain,“ offers a feasible way to assess neuropathological processes associated with many neurodegenerative diseases. Despite recent advances, clinical outcomes related to retinal degeneration are not always conclusive, requiring further research to improve diagnostic accuracy. For this purpose, artificial intelligence (AI) algorithms, including deep learning models and conventional machine learning algorithms, have proven to be efficient diagnostic tools.

Study Details:
The study aimed to systematically profile the classification performance in different stages of Parkinson’s disease progression, including incident and prevalent Parkinson’s disease, using AI algorithms without methods for feature selection or external quantitative measurements. The robustness was achieved through deep learning and conventional machine learning methods.

Results:
Deep neural networks outperformed conventional machine learning models and demonstrated remarkable performance in predicting PD incidence before official diagnosis, with a sensitivity of 80% within 5.07 years. The accuracy achieved in distinguishing patients with prevalent and incident Parkinson’s disease from corresponding healthy controls was 68%.

Conclusions:
The study concluded that deep learning models surpassed conventional machine learning models in accurately predicting Parkinson’s disease based on retinal fundus images. It is expected that this work will serve as a reference for future research and algorithm selection in clinical settings. However, the study’s limitations include the size of the dataset and the population’s specific demographics, restricting the generalizability of the results.

Journal Reference:
Tran, C., Shen, K., Liu, K., et al. (2024). Deep Learning predicts the incidence and prevalence of Parkinson’s disease using fundus imaging from the UK Biobank. Scientific Reports, 14(1), 1-12. doi:10.1038/s41598-024-54251-1.

Related Posts

Adblock Detected

Please support us by disabling your AdBlocker extension from your browsers for our website.