Dry eye disease is a common and potentially debilitating medical condition. Meibum secreted from the meibomian glands is the largest contributor to the outermost, protective lipid layer of the tear film. Dysfunction of the meibomian glands is the most common cause of dry eye disease. As meibomian gland dysfunction progresses, gradual atrophy of the glands is observed. The meibomian glands are commonly visualized through meibography, a technique requiring specialist equipment and knowledge that might not be available to the physician. In the present project we use machine learning on clinical tabular data to predict the degree of meibomian gland dropout. Moreover, we employ explainable artificial intelligence on the best performing algorithms for feature importance evaluation. The best performing algorithms were AdaBoost, multilayer perceptron and LightGBM which outperformed the majority vote baseline classifier in every included evaluation metric for both multioutput and binary classification. Through explainable artificial intelligence known associations are validated and novel connections identified and discussed.

Diabetic retinopathy is a common complication of diabetes, and monitoring the progression of retinal abnormalities using fundus imaging is crucial. Because the images must be interpreted by a medical expert, it is infeasible to screen all individuals with diabetes for diabetic retinopathy. Deep learning has shown impressive results for automatic analysis and grading of fundus images. One drawback is, however, the lack of interpretability, which hampers the implementation of such systems in the clinic. Explainable artificial intelligence methods can be applied to explain the deep neural networks. Explanations based on concepts have shown to be intuitive for humans to understand, but have not yet been explored in detail for diabetic retinopathy grading. This work investigates and compares two concept-based explanation techniques for explaining deep neural networks developed for automatic diagnosis of diabetic retinopathy: Quantitative Testing with Concept Activation Vectors and Concept Bottleneck Models. We found that both methods have strengths and weaknesses, and choice of method should take the available data and the end user’s preferences into account.