Clinical Analysis from Pattern Disentanglement Insight

Anonymous

Abstract: Diagnosis of a clinical condition can help medical professionals save time in the decision-making and prevent overlooking risks. Several machine learning models have been developed to predict clinical conditions, however, many existing models may have ineffective interpretability which is often desirable. In this paper, we explore the problem of text interpretability using free-text medical notes recorded in electronic health records (EHR). We propose an algorithm combining text mining and pattern discovery solution to discover strong association patterns between patient discharge summaries and the code of international classification of diseases (ICD9 code). The proposed approach offers a straightforward interpretation of the underlying relation of patient characteristics in an unsupervised machine learning setting and also outperforms the baseline clustering algorithm and is comparable to baseline supervised methods

Paper Type: short

Research Area: Interpretability and Analysis of Models for NLP