Thomas J. Schultz

Informatics in Radiology: Render: An Online Searchable Radiology Study Repository

Radiology departments are a rich source of information in the form of digital radiology reports and images obtained in patients with a wide spectrum of clinical conditions. A free text radiology report and image search application known as Render was created to allow users to find pertinent cases for a variety of purposes. Render is a radiology report and image repository that pools researchable information derived from multiple systems in near real time with use of (a) Health Level 7 links for radiology information system data, (b) periodic file transfers from the picture archiving and communication system, and (c) the results of natural language processing (NLP) analysis. Users can perform more structured and detailed searches with this application by combining different imaging and patient characteristics such as examination number; patient age, gender, and medical record number; and imaging modality. Use of NLP analysis allows a more effective search for reports with positive findings, resulting in the retrieval of more cases and terms having greater relevance. From the retrieved results, users can save images, bookmark examinations, and navigate to an external search engine such as Google. Render has applications in the fields of radiology education, research, and clinical decision support.

Extraction of Recommendation Features in Radiology with Natural Language Processing: Exploratory Study

OBJECTIVE The purposes of this study were to validate a natural language processing program for extraction of recommendation features, such as recommended time frames and imaging technique, from electronic radiology reports and to assess patterns of recommendation features in a large database of radiology reports. MATERIALS AND METHODS This study was performed on a radiology reports database covering the years 1995-2004. From this database, 120 reports with and without recommendations were selected and randomized. Two radiologists independently classified these reports according to presence of recommendations, time frame, and imaging technique suggested for follow-up or repeated examinations. The natural language processing program then was used to classify the reports according to the same criteria used by the radiologists. The accuracy of classification of recommendation features was determined. The program then was used to determine the patterns of recommendation features for different patients and imaging features in the entire database of 4,211,503 reports. RESULTS The natural language processing program had an accuracy of 93.2% (82/88) for identifying the imaging technique recommended by the radiologists for further evaluation. Categorization of recommended time frames in the reports with the 88 recommendations obtained with the program resulted in 83 (94.3%) accurate classifications and five (5.7%) inaccurate classifications. Recommendations of CT were most common (27.9%, 105,076 of 376,918 reports) followed by those for MRI (17.8%). In most (85.4%, 322,074/376,918) of the reports with imaging recommendations, however, radiologists did not specify the time frame. CONCLUSION Accurate determination of recommended imaging techniques and time frames in a large database of radiology reports is possible with a natural language processing program. Most imaging recommendations are for high-cost but more accurate radiologic studies.

Natural Language Processing Using Online Analytic Processing for Assessing Recommendations in Radiology Reports

PURPOSE The study purpose was to describe the use of natural language processing (NLP) and online analytic processing (OLAP) for assessing patterns in recommendations in unstructured radiology reports on the basis of patient and imaging characteristics, such as age, gender, referring physicians, radiology subspecialty, modality, indications, diseases, and patient status (inpatient vs outpatient). MATERIALS AND METHODS A database of 4,279,179 radiology reports from a single tertiary health care center during a 10-year period (1995-2004) was created. The database includes reports of computed tomography, magnetic resonance imaging, fluoroscopy, nuclear medicine, ultrasound, radiography, mammography, angiography, special procedures, and unclassified imaging tests with patient demographics. A clinical data mining and analysis NLP program (Leximer, Nuance Inc, Burlington, Massachusetts) in conjunction with OLAP was used for classifying reports into those with recommendations (I(REC)) and without recommendations (N(REC)) for imaging and determining I(REC) rates for different patient age groups, gender, imaging modalities, indications, diseases, subspecialties, and referring physicians. In addition, temporal trends for I(REC) were also determined. RESULTS There was a significant difference in the I(REC) rates in different age groups, varying between 4.8% (10-19 years) and 9.5% (>70 years) (P <.0001). Significant variations in I(REC) rates were observed for different imaging modalities, with the highest rates for computed tomography (17.3%, 100,493/581,032). The I(REC) rates varied significantly for different subspecialties and among radiologists within a subspecialty (P < .0001). For most modalities, outpatients had a higher rate of recommendations when compared with inpatients. CONCLUSION The radiology reports database analyzed with NLP in conjunction with OLAP revealed considerable differences between recommendation trends for different imaging modalities and other patient and imaging characteristics.

Use of Radcube for Extraction of Finding Trends in a Large Radiology Practice

The purpose of our study was to demonstrate the use of Natural Language Processing (Leximer), along with Online Analytic Processing, (NLP-OLAP), for extraction of finding trends in a large radiology practice. Prior studies have validated the Natural Language Processing (NLP) program, Leximer for classifying unstructured radiology reports based on the presence of positive radiology findings (FPOS) and negative radiology findings (FNEG). The FPOS included new relevant radiology findings and any change in status from prior imaging. Electronic radiology reports from 1995–2002 and data from analysis of these reports with NLP-Leximer were saved in a data warehouse and exported to a multidimensional structure called the Radcube. Various relational queries on the data in the Radcube were performed using OLAP technique. Thus, NLP-OLAP was applied to determine trends of FPOS in different radiology exams for different patient and examination attributes. Pivot tables were exported from NLP-OLAP interface to Microsoft Excel for statistical analysis. Radcube allowed rapid and comprehensive analysis of FPOS and FNEG trends in a large radiology report database. Trends of FPOS were extracted for different patient attributes such as age groups, gender, clinical indications, diseases with ICD codes, patient types (inpatient, ambulatory), imaging characteristics such as imaging modalities, referring physicians, radiology subspecialties, and body regions. Data analysis showed substantial differences between FPOS rates for different imaging modalities ranging from 23.1% (mammography, 49,163/212,906) to 85.8% (nuclear medicine, 93,852/109,374; p < 0.0001). In conclusion, NLP-OLAP can help in analysis of yield of different radiology exams from a large radiology report database.

Integrating digital teaching-file systems with off-the-shelf presentation software to facilitate speaker-led conferences

Objective: Develop methods for automated transfer of images and associated text from a teaching-file repository into presentation material for speaker-led conferences.erials/Methods:Our institution uses a Microsoft Windows (Microsoft Corp, Redmond, WA) software application to maintain a digital teaching-file database that can store and retrieve content in a case-centric fashion. Virtually any number of images can be stored with any given case. Cases and their associated images can be retrieved via a module that supports searches by American College of Radiology (ACR) code and by free-text Boolean queries on the history, findings, diagnosis, and discussion components of a case. In addition to the software system serving directly as an interactive teaching tool, the digital teaching file itself serves as an image repository and resource for attending radiologists who create their own presentations and lectures. To better support this use, software modules were developed for interprocess communication and automated creation of Powerpoint slides. These modules are fully integrated with the teaching-file software application. A single image or a set of selected images can be automatically made into individual slides with two mouse clicks. Images are automatically centered and optimally sized. A slide title is automatically rendered from the user’s preference of the case history or diagnosis (stored with the case), or via the entry of freeform text. We describe the programming techniques that are used, as well as how several features of the operating system and Powerpoint itself can be integrated with a customized software application to facilitate this objective.Results: The creation of presentation-ready Powerpoint slides is fully automated from within our teaching-file application, and the time required to create a presentation compared to the conventional method of manually seeking and inserting files from within Powerpoint itself, on a per-slide basis, is drastically reduced. The benefits are magnified by having all imagery stored within an organized and searchable database system so that desired images can be easily located.Conclusion: A digital teaching-file system can serve as a useful image repository for purposes ancillary to direct computerized instruction. Software that supports these uses, such as the automated creation of presentation material for speaker-led conferences, facilitates the radiologist’s role as an educator.

The primary interpretation workstation: Information beyond image data

With the advent of picture archival and communication systems (PACS), the importance of design surrounding primary review workstations has become apparent. To help acceptance of filmless medical imaging, workstations must be developed that serve the needs of both radiologists and referring clinicians. This report will discuss integral requirements of workstation design, including information creation, medical management, medical knowledge, and enabling technologies.

Radiation shielding calculation for digital breast tomosynthesis rooms with an updated workload survey

PURPOSE To present shielding calculations for clinical digital breast tomosynthesis (DBT) rooms with updated workload data from a comprehensive survey and to provide reference shielding data for DBT rooms. METHODS The workload survey was performed from eight clinical DBT (Hologic Selenia Dimensions) rooms at Massachusetts General Hospital (MGH) for the time period between 10/1/2014 and 10/1/2015. Radiation output related information tags from the DICOM header, including mAs, kVp, beam filter material and gantry angle, were extracted from a total of 310 421 clinical DBT acquisitions from the PACS database. DBT workload distributions were determined from the survey data. In combination with previously measured scatter fraction data, unshielded scatter air kerma for each room was calculated. Experiment measurements with a linear-array detector were also performed on representative locations for verification. Necessary shielding material and thickness were determined for all barriers. For the general purpose of DBT room shielding, a set of workload-distribution-specific transmission data and unshielded scatter air kerma values were calculated using the updated workload distribution. RESULTS The workload distribution for Hologic DBT systems could be simplified by five different kVp/filter combinations for shielding purpose. The survey data showed the predominance of 45° gantry location for medial-lateral-oblique views at MGH. When taking into consideration the non-isotropic scatter fraction distribution together with the gantry angle distribution, accurate and conservative estimate of the unshielded scatter air kerma levels were determined for all eight DBT rooms. Additional shielding was shown to be necessary for two 4.5 cm wood doors. CONCLUSIONS This study provided a detailed workload survey and updated transmission data and unshielded scatter air kerma values for Hologic DBT rooms. Example shielding calculations were presented for clinical DBT rooms.

Integrating digital educational content created and stored within disparate software environments: An extensible markup language (XML) solution in real-world use

Objective: To provide a standardized and scaleable mechanism for exchanging digital radiologic educational content between software systems that use disparate authoring, storage, and presentation technologies.Materials/Methods: Our institution uses two distinct software systems for creating educational content for radiology. Each system is used to create in-house educational content as well as commercial educational products. One system is an authoring and viewing application that facilitates the input and storage of hierarchical knowledge and associated imagery, and is capable of supporting a variety of entity relationships. This system is primarily used for the production and subsequent viewing of educational CD-ROMS. Another software system is primarily used for radiologic education on the world wide web. This system facilitates input and storage of interactive knowledge and associated imagery, delivering this content over the internet in a Socratic manner simulating in-person interaction with an expert. A subset of knowledge entities common to both systems was derived. An additional subset of knowledge entities that could be bidirectionally mapped via algorithmic transforms was also derived. An extensible markup language (XML) object model and associated lexicon were then created to represent these knowledge entities and their interactive behaviors. Forward-looking attention was exercised in the creation of the object model in order to facilitate straightforward future integration of other sources of educational content. XML generators and interpreters were written for both systems.Results: Deriving the XML object model and lexicon was the most critical and time-consuming aspect of the project. The coding of the XML generators and interpreters required only a few hours for each environment. Subsequently, the transfer of hundreds of educational cases and thematic presentations between the systems can now be accomplished in a matter of minutes. The use of of context as well as content, thus providing “presentationready” outcomes.Conclusion: The automation of knowledge exchange between dissimilar digital teaching environments magnifies the efforts of educators and enriches the learning experience for participants. XML is a powerful and useful mechanism for transfering educational content, as well as the context and interactive behaviors of such content, between disparate systems.