Automated ICD-9 Medical Code Assignment from Given Free Text Using Deep Learning Approach

Abstract

AbstractInternational Classification of Diseases (ICD) is an authenticated medical management categorization structure of various diseases and health situations for medical and governance causes. Medical coding is a process which allocates a standard ICD codes to a patient visit, his health condition and diagnoses given by the doctor. It is a essential method which is important for sick person’s health concern and invoicing purpose. Hand-operated coding is costly and prone to error. Also, it takes too much time. Assigning correct codes to each patient admission is complicated and dedicated process. Combined approach using natural language processing and deep learning techniques is one of the best ways to solve the problem. So, based on overall complexity and diagnosis, we propose a deep learning solution of bidirectional long short-term memory (bi-LSTM) with attention layer methodology which can automatically assign International Classification of Diseases diagnostic codes to the diagnosis given by the doctor. In the proposed approach, word embedding is used to create hidden representation of diagnosis descriptions given and International Classification of Diseases (ICD) codes and propose an attention layer technique to find out the conflict between the numbers of diagnoses and respective codes. In bi-LSTM model, signal propagates backward as well as forward in time. The proposed approach will assign most probable ICD codes to the diagnoses using bi-LSTM.KeywordsBidirectional long short term memory (Bi-LSTM)International classification of diseases (ICD)Deep learning (DL)Natural language processing (NLP)diagnoses