Automated computer-assisted categorization of radiology reports

OBJECTIVE: The objective of our study was to create and validate an automated computerized method for the categorization of narrative text radiograph reports. MATERIALS AND METHODS: Using commercially available software with embedded Boolean logic, we created a text search algorithm to categorize reports of radiography examinations into "fracture,"normal," and "neither normal nor fracture." The algorithm was refined and optimized through repeated testing on 512 consecutive ankle radiography reports from a single clinical imaging center. The final algorithm was applied on a different set of 750 consecutive radiography reports of the spine and extremities produced at three different clinical imaging sites and interpreted by 44 different radiologists. Expert reviewers assessed the accuracy of the final classification. The chi-square test or Fisher's exact test was performed to determine the reproducibility of results across different clinical imaging sites. RESULTS: The computerized classification was highly accurate for the classification of radiography reports into "normal" (specificity, 91.6%; sensitivity, 91.3%), "neither normal nor fracture"(sensitivity, 87.8%; specificity, 94.9%), and "fracture"(sensitivity, 94.1%; specificity, 98.1%) categories. This performance showed no significant difference across the three sites (p >0.05). CONCLUSION: Computerized categorization of narrative-text radiography reports is highly sensitive and specific and can be used to classify reports from different imaging sites generated by different radiologists. This method can be an extremely powerful tool in future cost-effectiveness studies, health care policy studies, operations assessments, and quality control.     

Automatic structuring of radiology free-text reports.

A natural language processor was developed that automatically structures the important medical information (eg, the existence, properties, location, and diagnostic interpretation of findings) contained in a radiology free-text document as a formal information model that can be interpreted by a computer program. The input to the system is a free-text report from a radiologic study. The system requires no reporting style changes on the part of the radiologist. Statistical and machine learning methods are used extensively throughout the system. A graphical user interface has been developed that allows the creation of hand-tagged training examples. Various aspects of the difficult problem of implementing an automated structured reporting system have been addressed, and the relevant technology is progressing well. Extensible Markup Language is emerging as the preferred syntactic standard for representing and distributing these structured reports within a clinical environment. Early successes hold out hope that similar statistically based models of language will allow deep understanding of textual reports. The success of these statistical methods will depend on the availability of large numbers of high-quality training examples for each radiologic subdomain. The acceptability of automated structured reporting systems will ultimately depend on the results of comprehensive evaluations.     

Critical issues in radiology requests and reports

Purpose

This study evaluated the appropriateness and accuracy of 500 radiology requests and their matched reports in order to identify recurring errors in both areas.

Materials and methods

A randomly chosen sample consisting of 167 computed tomography (CT), 166 ultrasonography (US) and 167 radiographic examinations were collected and analysed according to national referral guidelines and to the principles of justification and optimisation (Law no. 187/2000).

Results

We identified a high rate of inappropriate requests (27.6%) and requests lacking a clinical question (22%). There was good precision in the anamnestic data (80.6%) and in the formulation of the diagnostic question (76.8%). Almost all requests were handwritten, and 12.5% lacked the referring physician??s stamp and/or signature. No report mentioned the clinical information received or the equipment used. The use of contrast medium was always reported. Conclusions were reported in 9.8% of these reports. When further investigation would have been necessary, the radiologist omitted to report this in 60% of cases.

Conclusions

Some important weaknesses emerged, especially regarding requests for radiological examinations (22% lacked the clinical question, 27.6% were inappropriate), potentially limiting the effectiveness of the diagnostic process and leading to negative effects on the correct risk management process. There emerges a need for better collaboration between clinicians and radiologists.     

Intelligent image retrieval based on radiology reports

European Radiology - To create an advanced image retrieval and data-mining system based on in-house radiology reports. Radiology reports are semantically analysed using natural language processing...     

基于影像报告的智能图像检索

目的创建一个基于内部影像报告的高级图像检索和数据挖掘系统。方法首先,采用自然语言处理(NLP)技术对影像报告进行语义分析,并将其储存于当前最先进的搜索引擎中。然后,在图像存储与传输系统(PACS)中利用报告中的序列和影像号检索图像,并存储以供今后查看。采用基于网络前端作为查询图像、显示检索结果和报告文本的界面。以一本全面的放射学词典作为基本术语来源,搜索算法可同时搜索同义词、缩写及相关主题。结果采用不同系统设置对108份人工书写的报告进行分析。其中,全句法和语义分析效果最佳,其精确度和查全率分别为0.929和0.952。此系统自2010年10月成功运行以来,已有258824份报告被索引,405146幅预览图像存储于数据库中。结论采用数据挖掘和NLP技术能够快速访问大量的图像和影像报告资源库,具有很高的精确度和查全率。因此,该系统已成为日常临床工作、教学和研究的重要工具。     

Multicenter survey clarifying phrases in emergency radiology reports

Emergency Radiology - Interactions between radiologists and emergency physicians are often diminished as imaging volume increases and more radiologists read off site. We explore how several...     

Readability of radiology reports: implications for patient-centered care

IntroductionRadiology reports, although written primarily for healthcare providers, are read increasingly by patients and their family. This study sought to assess the readability of radiology reports.Materials and methodsFrom 108,228 consecutive radiology reports from a large US health system, we excluded duplicate reports, reports of research exams, and reports with missing data. For each report, we measured the numbers of words and sentences, and computed a “reading grade level” (RGL) as the mean of three readability indices: Flesch-Kincaid Grade Level, Gunning Fog index, and Simple Measure of Gobbledygook. Analysis of variance (ANOVA) evaluated the effects of modality, patient setting, examination urgency, and combinations thereof on RGL.ResultsThe 97,052 reports in the study cohort had a mean (±standard deviation) of 17.6 ± 12.8 sentences and 203 ± 161 words. Patient setting, modality, and examination urgency all had significant independent effects on RGL (all with p < 0.001). There were 4094 reports (4.2%) at a reading grade level of 8 or lower.ConclusionRadiology reports often contain complex concepts and polysyllabic terms unfamiliar to lay readers. Only 4% of all radiology reports in our sample were readable at the 8th grade level, which is the reading level of the average US adult. Although radiology reports are written for physicians and other healthcare providers, radiologists might explore using simpler, more structured language to address the goals of patient-centered care.     

A Bayesian approach for the categorization of radiology reports

RATIONALE AND OBJECTIVE: We sought to develop a Bayesian-filter that could distinguish positive radiology computed tomography (CT) reports of appendicitis from negative reports with no appendicitis. MATERIALS AND METHODS: Standard unstructured electronic text radiology reports containing the key word appendicitis were obtained using a Java-based text search engine from a hospital General Electric PACS system. A total of 500 randomly selected reports from multiple radiologists were then manually categorized and merged into two separate text files: 250 positive reports and 250 negative findings of appendicitis. The two text files were then processed by the freely available UNIX-based software dbacl 1.9, a digramic Bayesian classifier for text recognition, on a Linux based Pentium 4 system. The software was then trained on the two separate merged text files categories of positive and negative appendicitis. The ability of the Bayesian filter to discriminate between reports of negative and positive appendicitis images was then tested on 100 randomly selected reports of appendicitis: 50 positive cases and 50 negative cases. RESULTS: The training time for the Bayesian filter was approximately 2 seconds. The Bayesian filter subsequently was able to categorize 50 of 50 positive reports of appendicitis and 50 of 50 reports of negative appendicitis, in less than 10 seconds. CONCLUSION: A Bayesian-filter system can be used to quickly categorize radiology report findings and automatically determine after training, with a high degree of accuracy, whether the reports have text findings of a specific diagnosis. The Bayesian filter can potentially be applied to any type of radiologic report finding and any relevant category.