Unsupervised Topic Modeling in a Large Free Text Radiology Report Repository

Abstract

Radiology report narrative contains a large amount of information about the patient's health and the radiologist's interpretation of medical findings. Most of this critical information is entered in free text format, even when structured radiology report templates are used. The radiology report narrative varies in use of terminology and language among different radiologists and organizations. The free text format and the subtlety and variations of natural language hinder the extraction of reusable information from radiology reports for decision support, quality improvement, and biomedical research. Therefore, as the first step to organize and extract the information content in a large multi-institutional free text radiology report repository, we have designed and developed an unsupervised machine learning approach to capture the main concepts in a radiology report repository and partition the reports based on their main foci. In this approach, radiology reports are modeled in a vector space and compared to each other through a cosine similarity measure. This similarity is used to cluster radiology reports and identify the repository's underlying topics. We applied our approach on a repository of 1,899,482 radiology reports from three major healthcare organizations. Our method identified 19 major radiology report topics in the repository and clustered the reports accordingly to these topics. Our results are verified by a domain expert radiologist and successfully explain the repository's primary topics and extract the corresponding reports. The results of our system provide a target-based corpus and framework for information extraction and retrieval systems for radiology reports.

Health Communication Through News Media During the Early Stage of the COVID-19 Outbreak in China: A Digital Topic Modeling Approach

“…LDA is a generative probabilistic topic modeling method that is widely applied in text mining 9 , medicine 10,11 , and social network analysis 12 due to its excellent capability of converting visual words into images and visual word document 13-15 . It is a generative statistical model with a three-level hierarchical Bayesian model.…”

Integrating Natural Language Processing and Machine Learning Algorithms to Categorize Oncologic Response in Radiology Reports

“…However, as TF and TF-IDF strategies can sometimes create extraordinarily high dimensional datasets, feature hashing has been shown to be an effective strategy for dimensionality reduction [10]. Bag-ofwords models have been used to successfully process radiology reports [11,12].…”

Natural Language Processing for Automated Annotation of Medication Mentions in Primary Care Visit Conversations

“…Prior work for finding medication mentions has focused on written clinical reports. [47][48][49][50][51] Our error analysis suggests that baseline approaches, which rely on dictionaries, struggle with patient-clinician conversational text because of language like filler words (for example, "aha" and "hmm") matching with abbreviations for medications, and the fact that common conversational words are often used as medication names.…”

“…These methods have been used to predict hospital readmissions, [43] future radiology utilizations, [44] and characterization of significance, change, and urgency of clinical findings in medical records. [45][46][47][48][49] As such, we plan to develop a recording system for patients that applies NLP methods to unstructured clinic visit recordings. [50] In this paper, we describe an approach to extract mentions of medication names in transcripts of clinic visit audio recordings.…”