Predicting High Imaging Utilization Based on Initial Radiology Reports: A Feasibility Study of Machine Learning

Construction of English and American Literature Corpus Based on Machine Learning Algorithm

In China, the application of corpus in language teaching, especially in English and American literature teaching, is still in the preliminary research stage, and there are various shortcomings, which have not been paid due attention by front-line educators. Constructing English and American literature corpus according to certain principles can effectively promote English and American literature teaching. The research of this paper is devoted to how to automatically build a corpus of English and American literature. In the process of keyword extraction, key phrases and keywords are effectively combined. The similarity between atomic events is calculated by the TextRank algorithm, and then the first N sentences with high similarity are selected and sorted. Based on ML (machine learning) text classification method, a combined classifier based on SVM (support vector machine) and NB (Naive Bayes) is proposed. The experimental results show that, from the point of view of accuracy and recall, the classification effect of the combined algorithm proposed in this paper is the best among the three methods. The best classification results of accuracy, recall, and F value are 0.87, 0.9, and 0.89, respectively. Experimental results show that this method can quickly, accurately, and persistently obtain high-quality bilingual mixed web pages.

Optimization of Radiology Workflow with Artificial Intelligence

The potential of artificial intelligence (AI) in radiology goes far beyond image analysis. AI can be used to optimize all steps of the radiology workflow by supporting a variety of nondiagnostic tasks, including order entry support, patient scheduling, resource allocation, and improving the radiologist's workflow. This article discusses several principal directions of using AI algorithms to improve radiological operations and workflow management, with the intention of providing a broader understanding of the value of applying AI in the radiology department.

An Audio Personal Health Library of Clinic Visit Recordings for Patients and Their Caregivers (HealthPAL): User-Centered Design Approach

Background

Providing digital recordings of clinic visits to patients has emerged as a strategy to promote patient and family engagement in care. With advances in natural language processing, an opportunity exists to maximize the value of visit recordings for patients by automatically tagging key visit information (eg, medications, tests, and imaging) and linkages to trustworthy web-based resources curated in an audio-based personal health library.

Objective

This study aims to report on the user-centered development of HealthPAL, an audio personal health library.

Methods

Our user-centered design and usability evaluation approach incorporated iterative rounds of video-recorded sessions from 2016 to 2019. We recruited participants from a range of community settings to represent older patient and caregiver perspectives. In the first round, we used paper prototypes and focused on feature envisionment. We moved to low-fidelity and high-fidelity versions of the HealthPAL in later rounds, which focused on functionality and use; all sessions included a debriefing interview. Participants listened to a deidentified, standardized primary care visit recording before completing a series of tasks (eg, finding where a medication was discussed in the recording). In the final round, we recorded the patients’ primary care clinic visits for use in the session. Findings from each round informed the agile software development process. Task completion and critical incidents were recorded in each round, and the System Usability Scale was completed by participants using the digital prototype in later rounds.

Results

We completed 5 rounds of usability sessions with 40 participants, of whom 25 (63%) were women with a median age of 68 years (range 23-89). Feedback from sessions resulted in color-coding and highlighting of information tags, a more prominent play button, clearer structure to move between one’s own recordings and others’ recordings, the ability to filter recording content by the topic discussed and descriptions, 10-second forward and rewind controls, and a help link and search bar. Perceived usability increased over the rounds, with a median System Usability Scale of 78.2 (range 20-100) in the final round. Participants were overwhelmingly positive about the concept of accessing a curated audio recording of a clinic visit. Some participants reported concerns about privacy and the computer-based skills necessary to access recordings.

Conclusions

To our knowledge, HealthPAL is the first patient-centered app designed to allow patients and their caregivers to access easy-to-navigate recordings of clinic visits, with key concepts tagged and hyperlinks to further information provided. The HealthPAL user interface has been rigorously co-designed with older adult patients and their caregivers and is now ready for further field testing. The successful development and use of HealthPAL may help improve the ability of patients to manage their own care, especially older adult patients who have to navigate complex treatment plans.

Automatic recommendation of medical departments to outpatients based on text analyses and medical knowledge graph

In many countries, outpatients generally visit a major hospital without a referral from health professionals due to the shortage of family physicians. Not knowing at which medical specialty department to register, outpatients have to wait in long queues to consult receptionists. We propose to alleviate this situation via a computer system offering an automatic recommendation of departments (ARD) to outpatients, which identifies the appropriate medical department for outpatients according to their chief complaints. Besides, ARD systems can boost the emerging services of online hospital registration and online medical diagnosis, which require that the outpatients know the correct department first. ARD is a typical problem of text classification. Nevertheless, off-the-shelf tools for text processing may not suit ARD, because the chief complaints of outpatients are generally brief and contain much noisy information. To solve this problem, we propose ARD-K, a deep learning framework incorporating external medical knowledge sources. We also propose a dual-attention mechanism to mitigate the interference of noisy words and knowledge entities. The performance of ARD-K is compared with some off-the-shelf techniques on a real-world dataset. The results demonstrate the effectiveness of ARD-K for the automatic recommendation of departments to outpatients.

Natural language processing for automated annotation of medication mentions in primary care visit conversations

OBJECTIVES

The objective of this study is to build and evaluate a natural language processing approach to identify medication mentions in primary care visit conversations between patients and physicians.

MATERIALS AND METHODS

Eight clinicians contributed to a data set of 85 clinic visit transcripts, and 10 transcripts were randomly selected from this data set as a development set. Our approach utilizes Apache cTAKES and Unified Medical Language System controlled vocabulary to generate a list of medication candidates in the transcribed text and then performs multiple customized filters to exclude common false positives from this list while including some additional common mentions of the supplements and immunizations.

RESULTS

Sixty-five transcripts with 1121 medication mentions were randomly selected as an evaluation set. Our proposed method achieved an F-score of 85.0% for identifying the medication mentions in the test set, significantly outperforming existing medication information extraction systems for medical records with F-scores ranging from 42.9% to 68.9% on the same test set.

DISCUSSION

Our medication information extraction approach for primary care visit conversations showed promising results, extracting about 27% more medication mentions from our evaluation set while eliminating many false positives in comparison to existing baseline systems. We made our approach publicly available on the web as an open-source software.

CONCLUSION

Integration of our annotation system with clinical recording applications has the potential to improve patients' understanding and recall of key information from their clinic visits, and, in turn, to positively impact health outcomes.

Comparative effectiveness of convolutional neural network (CNN) and recurrent neural network (RNN) architectures for radiology text report classification

This paper explores cutting-edge deep learning methods for information extraction from medical imaging free text reports at a multi-institutional scale and compares them to the state-of-the-art domain-specific rule-based system - PEFinder and traditional machine learning methods - SVM and Adaboost. We proposed two distinct deep learning models - (i) CNN Word - Glove, and (ii) Domain phrase attention-based hierarchical recurrent neural network (DPA-HNN), for synthesizing information on pulmonary emboli (PE) from over 7370 clinical thoracic computed tomography (CT) free-text radiology reports collected from four major healthcare centers. Our proposed DPA-HNN model encodes domain-dependent phrases into an attention mechanism and represents a radiology report through a hierarchical RNN structure composed of word-level, sentence-level and document-level representations. Experimental results suggest that the performance of the deep learning models that are trained on a single institutional dataset, are better than rule-based PEFinder on our multi-institutional test sets. The best F1 score for the presence of PE in an adult patient population was 0.99 (DPA-HNN) and for a pediatrics population was 0.99 (HNN) which shows that the deep learning models being trained on adult data, demonstrated generalizability to pediatrics population with comparable accuracy. Our work suggests feasibility of broader usage of neural network models in automated classification of multi-institutional imaging text reports for a variety of applications including evaluation of imaging utilization, imaging yield, clinical decision support tools, and as part of automated classification of large corpus for medical imaging deep learning work.

Making Sense of Big Textual Data for Health Care: Findings from the Section on Clinical Natural Language Processing

Objectives: To summarize recent research and present a selection of the best papers published in 2016 in the field of clinical Natural Language Processing (NLP). Method: A survey of the literature was performed by the two section editors of the IMIA Yearbook NLP section. Bibliographic databases were searched for papers with a focus on NLP efforts applied to clinical texts or aimed at a clinical outcome. Papers were automatically ranked and then manually reviewed based on titles and abstracts. A shortlist of candidate best papers was first selected by the section editors before being peer-reviewed by independent external reviewers. Results: The five clinical NLP best papers provide a contribution that ranges from emerging original foundational methods to transitioning solid established research results to a practical clinical setting. They offer a framework for abbreviation disambiguation and coreference resolution, a classification method to identify clinically useful sentences, an analysis of counseling conversations to improve support to patients with mental disorder and grounding of gradable adjectives. Conclusions: Clinical NLP continued to thrive in 2016, with an increasing number of contributions towards applications compared to fundamental methods. Fundamental work addresses increasingly complex problems such as lexical semantics, coreference resolution, and discourse analysis. Research results translate into freely available tools, mainly for English.

Combining Contrast Mining with Logistic Regression To Predict Healthcare Utilization in a Managed Care Population

BACKGROUND

Because 5% of patients incur 50% of healthcare expenses, population health managers need to be able to focus preventive and longitudinal care on those patients who are at highest risk of increased utilization. Predictive analytics can be used to identify these patients and to better manage their care. Data mining permits the development of models that surpass the size restrictions of traditional statistical methods and take advantage of the rich data available in the electronic health record (EHR), without limiting predictions to specific chronic conditions.

OBJECTIVE

The objective was to demonstrate the usefulness of unrestricted EHR data for predictive analytics in managed healthcare.

METHODS

In a population of 9,568 Medicare and Medicaid beneficiaries, patients in the highest 5% of charges were compared to equal numbers of patients with the lowest charges. Contrast mining was used to discover the combinations of clinical attributes frequently associated with high utilization and infrequently associated with low utilization. The attributes found in these combinations were then tested by multiple logistic regression, and the discrimination of the model was evaluated by the c-statistic.

RESULTS

Of 19,014 potential EHR patient attributes, 67 were found in combinations frequently associated with high utilization, but not with low utilization (support>20%). Eleven of these attributes were significantly associated with high utilization (p<0.05). A prediction model composed of these eleven attributes had a discrimination of 84%.

CONCLUSIONS

EHR mining reduced an unusably high number of patient attributes to a manageable set of potential healthcare utilization predictors, without conjecturing on which attributes would be useful. Treating these results as hypotheses to be tested by conventional methods yielded a highly accurate predictive model. This novel, two-step methodology can assist population health managers to focus preventive and longitudinal care on those patients who are at highest risk for increased utilization.