Pinpointing neurologic dysfunction in fundus photographs using deep learning analysis
Deep learning analysis of fundus photographs has the potential to assess neurologic dysfunction in individuals with Parkinson’s disease.
(Image Credit: AdobeStock/ipuwadol)
Sangil Ahn, BS, and colleagues reported that deep learning analysis of fundus photographs has the potential to assess neurologic dysfunction in individuals with Parkinson’s disease (PD).1 He is from the Department of Electrical and Computer Engineering, Sungkyunkwan University, Suwon, Republic of Korea.
The ability to assess neurologic disabilities in this manner would eliminate the need for complex neurologic testing by using a noninvasive widely available technology to assess patients, the investigators explained.
In their study, the goal was to develop an algorithm that can predict Hoehn and Yahr (H-Y) scale and Unified Parkinson’s Disease Rating Scale part III (UPDRS-III) scores using fundus photography in patients with PD.
Ahn and colleagues analyzed the fundus photographs of study patients with Parkinson’s disease and participants with atypical motor abnormalities that was not PD who visited the neurology department of Kangbuk Samsung Hospital from October 7, 2020, to April 30, 2021. They developed a convolutional neural network to predict both the H-Y scale and UPDRS-III scores based on fundus photography findings and the participants’ demographic characteristics, they explained.
The study population was comprised of 266 patients with PD (134 men; mean age, 70.8 years), and 349 individuals with non-PD atypical motor abnormalities (236 women; mean age, 70.7 years). The investigators analyzed 539 fundus images from the PD patients.
For the internal validation data set, the sensitivity values were 83.23% (95% confidence interval [CI], 82.07%-84.38%) and 82.61% (95% CI, 81.38%-83.83%) for the H-Y scale and UPDRS-III scores, respectively. The respective specificity values were 66.81% (95% CI, 64.97%-68.65%) and 65.75% (95% CI, 62.56%-68.94%).
For the external validation data set, the sensitivity and specificity were 70.73% (95% CI, 66.30%-75.16%) and 66.66% (95% CI, 50.76%-82.25%), respectively.
The calculated area under the ROC curve and accuracy were 0.67 (95% CI, 0.55-0.79) and 70.45% (95% CI, 66.85%-74.04%), respectively.
The investigators commented, “This decision analytical model reveals amalgamative insights into the neurologic dysfunction among PD patients by providing information on how to apply a deep learning method to evaluate the association between the retina and brain. Study data may help clarify recent research findings regarding dopamine pathologic cascades between the retina and brain among patients with PD; however, further research is needed to expand the clinical implication of this algorithm.”
