Pathological Myopia detection from selective fundus image features

We explore feature selection methodology for automatic Pathological Myopia detection via learning from an optimal set of features. An mRMR optimized classifier is trained using the candidate feature set to find the optimized classifier. We tested the proposed methodology on eye records of approximately 800 subjects collected from a population study. The experimental results demonstrate that the new classifier is much efficient by using less than 25% of the initial candidate feature set. The ranked optimal feature set also enables the clinicians to better access the diagnostic process of the algorithm. The work is a further step towards the advancement of the automatic pathological myopia diagnosis.