In this paper, we present a novel method to detect mania in bipolar individuals that utilizes pictures of the individual’s eyes. Both mania and depression can be detrimental to the individual and in some cases, can be life threatening. It is often difficult for an individual to identify that they are in a state of hypomania or mania. To understand if automated methods for detecting mania are possible, we created a study website and curated a dataset of eye images from individuals with bipolar disorder. These images were labeled as either manic, depressed or stable, where manic images were further broken down into euphoric or dysphoric. Using convolutional neural networks, random forest, and logistic regression, we were able to produce models detecting euphoric mania with accuracy rates averaging between 23% and 73%. Although our results may have promising accuracies, further investigation of the precision and recall values indicated overfitting in the models
Wheeler, Jessica; Jecha, Jean; Kottegoda, Manjula; Teo, Sharon; and Larson, Eric C.
"Bipolar Mania Eye Image Classification,"
SMU Data Science Review: Vol. 1
, Article 1.
Available at: https://scholar.smu.edu/datasciencereview/vol1/iss1/1
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License