Identify Rare Disease Patients from Electronic Health Records through Machine Learning Approach

Abstract

Uncommon maladies are extremely hard to recognize among expansive number of other conceivable judgments. Better accessibility of patient information and change in machine learning calculations engage us to handle this issue computationally. In this paper, we target one such uncommon ailment - cardiovascular amyloidosis. We mean to computerize the way toward distinguishing potential cardiovascular amyloidosis patients with the assistance of machine learning calculations and furthermore learn most prescient elements. With the assistance of experienced cardiologists, we arranged a highest quality level with 73 positive (heart amyloidosis) and 197 negative cases. We accomplished high normal cross-approval F1 score of 0.98 utilizing a group machine learning classifier. A portion of the prescient factors were: Age and Diagnosis of heart failure, trunk torment, congestive heart disappointment, hypertension, demure open edge glaucoma, and shoulder joint pain. Additionally studies are expected to approve the exactness of the framework over a whole wellbeing framework and its generalizability for different maladies.