Digitization of medicine: how radiology can take advantage of the digital revolution

Evaluation of large language models performance against humans for summarizing MRI knee radiology reports: A feasibility study

OBJECTIVES

This study addresses the critical need for accurate summarization in radiology by comparing various Large Language Model (LLM)-based approaches for automatic summary generation. With the increasing volume of patient information, accurately and concisely conveying radiological findings becomes crucial for effective clinical decision-making. Minor inaccuracies in summaries can lead to significant consequences, highlighting the need for reliable automated summarization tools.

METHODS

We employed two language models - Text-to-Text Transfer Transformer (T5) and Bidirectional and Auto-Regressive Transformers (BART) - in both fine-tuned and zero-shot learning scenarios and compared them with a Recurrent Neural Network (RNN). Additionally, we conducted a comparative analysis of 100 MRI report summaries, using expert human judgment and criteria such as coherence, relevance, fluency, and consistency, to evaluate the models against the original radiologist summaries. To facilitate this, we compiled a dataset of 15,508 retrospective knee Magnetic Resonance Imaging (MRI) reports from our Radiology Information System (RIS), focusing on the findings section to predict the radiologist's summary.

RESULTS

The fine-tuned models outperform the neural network and show superior performance in the zero-shot variant. Specifically, the T5 model achieved a Rouge-L score of 0.638. Based on the radiologist readers' study, the summaries produced by this model were found to be very similar to those produced by a radiologist, with about 70% similarity in fluency and consistency between the T5-generated summaries and the original ones.

CONCLUSIONS

Technological advances, especially in NLP and LLM, hold great promise for improving and streamlining the summarization of radiological findings, thus providing valuable assistance to radiologists in their work.

Deep-learning based classification of a tumor marker for prognosis on Hodgkin's disease

PURPOSE

Hodgkin's disease is a common malignant disorder in adolescent patients. Although most patients are cured, approximately 10%-15% of patients experience a relapse or have resistant disease. Furthermore, there are no definitive molecular predictors for early identification of patients at high risk of treatment failure to first line therapy. The aim of this study was to evaluate the deep learning-based classifier model of medical image classification to predict clinical outcome that may help in appropriate therapeutic decisions.

METHODS

Eighty-three FFPE biopsy specimens from patients with Hodgkin's disease were stratified according to the patient's qPET scores, stained with picrosirius red dye and digitalized by whole slide image scanning. The resulting whole slide images were cut into tiles and annotated by two classes based on the collagen fibers' degree of coloring with picrosirius red. The neural network (YOLOv4) was then trained with the annotated data. Training was performed with 30 cases. Prognostic power of the weakly stained picrosirius red fibers was evaluated with 53 cases. The same neural network was trained with MMP9 stained tissue slides from the same cases and the quantification results were compared with the variant from the picrosirius red cases.

RESULTS

There was a weak monotonically increasing relationship by parametric ANOVA between the qPET groups and the percentages of weakly stained fibers (p = .0185). The qPET-positive cases showed an average of 18% of weakly stained fibers, and the qPET-negative cases 10%-14%. Detection performance showed an AUC of 0.79.

CONCLUSIONS

Picrosirius red shows distinct associations as a prognostic metric candidate of disease progression in Hodgkin's disease cases using whole slide images but not sufficiently as a prognostic device.

Comprehensive strategies of machine-learning-based quantitative structure-activity relationship models

Early quantitative structure-activity relationship (QSAR) technologies have unsatisfactory versatility and accuracy in fields such as drug discovery because they are based on traditional machine learning and interpretive expert features. The development of Big Data and deep learning technologies significantly improve the processing of unstructured data and unleash the great potential of QSAR. Here we discuss the integration of wet experiments (which provide experimental data and reliable verification), molecular dynamics simulation (which provides mechanistic interpretation at the atomic/molecular levels), and machine learning (including deep learning) techniques to improve QSAR models. We first review the history of traditional QSAR and point out its problems. We then propose a better QSAR model characterized by a new iterative framework to integrate machine learning with disparate data input. Finally, we discuss the application of QSAR and machine learning to many practical research fields, including drug development and clinical trials.

Patient Access to an Online Portal for Outpatient Radiological Images and Reports: Two Years' Experience

To assess the incidence of outpatient examinations delivered through a web portal in the Latium Region in 2 years and compare socio-demographic characteristics of these users compared to the total of examinations performed. All radiological exams (including MRI, X-ray and CT) performed from March 2017 to February 2019 were retrospectively analysed. For each exam, anonymized data of users who attended the exam were extracted and their characteristics were compared according to digital access to the reports. Overall, 9068 exams were performed in 6720 patients (55.8% males, median age 58 years, interquartile range (IQR) 46-70) of which 90.2% residents in Rome province, mainly attending a single radiological examination (77.3%). Among all exams, 446 (4.9%) were accessed, of which 190 (4.4%) in the first and 5.4% in the second year (p < 0.041). MRI was the type of exams mostly accessed (175, 7.0%). Being resident in the provinces of the Latium Region other than Rome was associated with a higher access rate (OR = 1.84, p = 0.001). Considering the overall costs sustained to implement a web portal which allows users a personal access to their own reports, if all users would have accessed/downloaded their exams, an overall users' and hospital savings up to €255,808.28 could have been determined. The use of a web portal could represent a consistent economical advantage for the user, the hospital and the environment. Even if increasing over time, the use of web portal is still limited and strategies to increase the use of such systems should be implemented.

Challenges of Implementing Picture Archiving and Communication System in Multiple Hospitals: Perspectives of Involved Staff and Users

Today, despite the advantages of the PACS system, its implementation in some healthcare organizations faces many challenges. One of the important factors in the successful implementation of a PACS system is identifying and prioritizing the challenges from the perspectives of involved staff and user of this system. Therefore, the aim of this study was to determine and compare the challenges of implementing PACS from perspectives these users in educational hospitals. This study was conducted on all IT and medical equipment staff, and radiology residents (n = 140) in Kerman University of Medical Sciences (KUMS) and Shiraz University of Medical Sciences (SUMS) in 2016. The data were collected through two researcher-made questionnaires. Their validity was approved by radiologists, IT staff, and medical informatics specialists and their reliability through calculation of Cronbach's Alpha (0.969 and 0.795). We used Multivariate Analysis of Variance (MANOVA) to compare the scores given by three groups of participants in the challenges and Univariate Analysis of Variance (ANOVA) to compare the scores in two universities. The participants believed that technical challenges were more important than other challenges (x̄=3.74, SD = 0.7). IT experts (x̄=3.87, SD = 1) and radiology residents (x̄=3.95, SD = 0.9) gave the higher scores to the "shortage of high quality monitors" factor and medical equipment experts (x̄=4.26, SD = 0.87) to the "low speed of communication networks" factor among all technical challenges. The mean scores given to technical (x̄=76.1, SD = 13.5) and managerial (x̄=16, SD = 5.9) challenges in SUMS were more than the scores of the same challenges in KUMS (x̄=69.9, SD = 15.7) and (x̄=11.9, SD = 6.4) (p < 0.05). The technical challenges are the most common challenges to PACS implementation, and different universities experience different levels of technical challenges. Eliminating implementation challenges can reduce the risk of failure in the utilization process. Based on the results of this study, providing necessary infrastructures such as appropriate monitors and upgraded IT equipment can prevent many of the PACS implementation challenges.

The need for standardization of nuclear cardiology reporting and data system (NCAD-RADS): Learning from coronary artery disease (CAD), breast imaging (BI), liver imaging (LI), and prostate imaging (PI) RADS

Newer structured reporting manners, the reporting and data system (RADS), have made vast steps in improving standardized and structured reporting, allowing better communication between radiologists and referring providers. This has been implemented in several fields: breast (BI-RADS), lung (Lung-RADS), liver (LI-RADS), thyroid (TI-RADS), prostate (PI-RADS), and in cardiovascular radiology (CAD-RADS). The field of nuclear cardiology began its efforts of standardization years ago; however, a widespread standardized reporting structure has not yet been adopted. Such an approach in nuclear cardiology, the nuclear cardiology reporting and data system (NCAD-RADS), will assist radiologists and treating clinicians in conveying and understanding reports and determining the appropriate next steps in management. By linking explicit findings to defined recommendations, patients will receive more consistent and appropriate care.

Conversion of Radiology Reporting Templates to the MRRT Standard

In 2013, the Integrating the Healthcare Enterprise (IHE) Radiology workgroup developed the Management of Radiology Report Templates (MRRT) profile, which defines both the format of radiology reporting templates using an extension of Hypertext Markup Language version 5 (HTML5), and the transportation mechanism to query, retrieve, and store these templates. Of 200 English-language report templates published by the Radiological Society of North America (RSNA), initially encoded as text and in an XML schema language, 168 have been converted successfully into MRRT using a combination of automated processes and manual editing; conversion of the remaining 32 templates is in progress. The automated conversion process applied Extensible Stylesheet Language Transformation (XSLT) scripts, an XML parsing engine, and a Java servlet. The templates were validated for proper HTML5 and MRRT syntax using web-based services. The MRRT templates allow radiologists to share best-practice templates across organizations and have been uploaded to the template library to supersede the prior XML-format templates. By using MRRT transactions and MRRT-format templates, radiologists will be able to directly import and apply templates from the RSNA Report Template Library in their own MRRT-compatible vendor systems. The availability of MRRT-format reporting templates will stimulate adoption of the MRRT standard and is expected to advance the sharing and use of templates to improve the quality of radiology reports.

Development of an IHE MRRT-compliant open-source web-based reporting platform

OBJECTIVES

To develop a platform that uses structured reporting templates according to the IHE Management of Radiology Report Templates (MRRT) profile, and to implement this platform into clinical routine.

METHODS

The reporting platform uses standard web technologies (HTML / JavaScript and PHP / MySQL) only. Several freely available external libraries were used to simplify the programming. The platform runs on a standard web server, connects with the radiology information system (RIS) and PACS, and is easily accessible via a standard web browser.

RESULTS

A prototype platform that allows structured reporting to be easily incorporated into the clinical routine was developed and successfully tested. To date, 797 reports were generated using IHE MRRT-compliant templates (many of them downloaded from the RSNA's radreport.org website). Reports are stored in a MySQL database and are easily accessible for further analyses.

CONCLUSION

Development of an IHE MRRT-compliant platform for structured reporting is feasible using only standard web technologies. All source code will be made available upon request under a free license, and the participation of other institutions in further development is welcome.

KEY POINTS

• A platform for structured reporting using IHE MRRT-compliant templates is presented. • Incorporating structured reporting into clinical routine is feasible. • Full source code will be provided upon request under a free license.

Building for tomorrow today: opportunities and directions in radiology resident research

RATIONALE AND OBJECTIVES

With rapid scientific and technological advancements in radiological research, there is renewed emphasis on promoting early research training to develop researchers who are capable of tackling the hypothesis-driven research that is typically funded in contemporary academic research enterprises. This review article aims to introduce radiology residents to the abundant radiology research opportunities available to them and to encourage early research engagement among trainees.

MATERIALS AND METHODS

To encourage early resident participation in radiology research, we review the various research opportunities available to trainees spanning basic, clinical, and translational science opportunities to ongoing research in information technology, informatics, and quality improvement research.

CONCLUSIONS

There is an incredible breadth and depth of ongoing research at academic radiology departments across the country, and the material presented herein aspires to highlight both subject matter and opportunities available to radiology residents eager to engage in radiologic research. The opportunities for interested radiology residents are as numerous as they are broad, spanning the basic sciences to clinical research to informatics, with abundant opportunities to shape our future practice of radiology.

Towards social radiology as an information infrastructure: reconciling the local with the global

The current widespread use of medical images and imaging procedures in clinical practice and patient diagnosis has brought about an increase in the demand for sharing medical imaging studies among health professionals in an easy and effective manner. This article reveals the existence of a polarization between the local and global demands for radiology practice. While there are no major barriers for sharing such studies, when access is made from a (local) picture archive and communication system (PACS) within the domain of a healthcare organization, there are a number of impediments for sharing studies among health professionals on a global scale. Social radiology as an information infrastructure involves the notion of a shared infrastructure as a public good, affording a social space where people, organizations and technical components may spontaneously form associations in order to share clinical information linked to patient care and radiology practice. This article shows however, that such polarization establishes a tension between local and global demands, which hinders the emergence of social radiology as an information infrastructure. Based on an analysis of the social space for radiology practice, the present article has observed that this tension persists due to the inertia of a locally installed base in radiology departments, for which common teleradiology models are not truly capable of reorganizing as a global social space for radiology practice. Reconciling the local with the global signifies integrating PACS and teleradiology into an evolving, secure, heterogeneous, shared, open information infrastructure where the conceptual boundaries between (local) PACS and (global) teleradiology are transparent, signaling the emergence of social radiology as an information infrastructure.