Quantitative Evaluation of Large Language Models to Streamline Radiology Report Impressions: A Multimodal Retrospective Analysis.

Background The complex medical terminology of radiology reports may cause confusion or anxiety for patients, especially given increased access to electronic health records. Large language models (LLMs) can potentially simplify radiology report readability. Purpose To compare the performance of four publicly available LLMs (ChatGPT-3.5 and ChatGPT-4, Bard [now known as Gemini], and Bing) in producing simplified radiology report impressions. Materials and Methods In this retrospective comparative analysis of the four LLMs (accessed July 23 to July 26, 2023), the Medical Information Mart for Intensive Care (MIMIC)-IV database was used to gather 750 anonymized radiology report impressions covering a range of imaging modalities (MRI, CT, US, radiography, mammography) and anatomic regions. Three distinct prompts were employed to assess the LLMs' ability to simplify report impressions. The first prompt (prompt 1) was "Simplify this radiology report." The second prompt (prompt 2) was "I am a patient. Simplify this radiology report." The last prompt (prompt 3) was "Simplify this radiology report at the 7th grade level." Each prompt was followed by the radiology report impression and was queried once. The primary outcome was simplification as assessed by readability score. Readability was assessed using the average of four established readability indexes. The nonparametric Wilcoxon signed-rank test was applied to compare reading grade levels across LLM output. Results All four LLMs simplified radiology report impressions across all prompts tested (P < .001). Within prompts, differences were found between LLMs. Providing the context of being a patient or requesting simplification at the seventh-grade level reduced the reading grade level of output for all models and prompts (except prompt 1 to prompt 2 for ChatGPT-4) (P < .001). Conclusion Although the success of each LLM varied depending on the specific prompt wording, all four models simplified radiology report impressions across all modalities and prompts tested. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Rahsepar in this issue.

Artificial Intelligence to Improve Patient Understanding of Radiology Reports.

Diagnostic imaging reports are generally written with a target audience of other providers. As a result, the reports are written with medical jargon and technical detail to ensure accurate communication. With implementation of the 21st Century Cures Act, patients have greater and quicker access to their imaging reports, but these reports are still written above the comprehension level of the average patient. Consequently, many patients have requested reports to be conveyed in language accessible to them. Numerous studies have shown that improving patient understanding of their condition results in better outcomes, so driving comprehension of imaging reports is essential. Summary statements, second reports, and the inclusion of the radiologist's phone number have been proposed, but these solutions have implications for radiologist workflow. Artificial intelligence (AI) has the potential to simplify imaging reports without significant disruptions. Many AI technologies have been applied to radiology reports in the past for various clinical and research purposes, but patient focused solutions have largely been ignored. New natural language processing technologies and large language models (LLMs) have the potential to improve patient understanding of their imaging reports. However, LLMs are a nascent technology and significant research is required before LLM-driven report simplification is used in patient care.

Practical Guide to Natural Language Processing for Radiology.

Natural language processing (NLP) is the subset of artificial intelligence focused on the computer interpretation of human language. It is an invaluable tool in the analysis, aggregation, and simplification of free text. It has already demonstrated significant potential in the analysis of radiology reports. There are abundant open-source libraries and tools available that facilitate its application to the benefit of radiology. Radiologists who understand its limitations and potential will be better positioned to evaluate NLP models, understand how they can improve clinical workflow, and facilitate research endeavors involving large amounts of human language. The advent of increasingly affordable and powerful computer processing, the large quantities of medical and radiologic data, and advances in machine learning algorithms have contributed to the large potential of NLP. In turn, radiology has significant potential to benefit from the ability of NLP to convert relatively standardized radiology reports to machine-readable data. NLP benefits from standardized reporting, but because of its ability to interpret free text by using context clues, NLP does not necessarily depend on it. An overview and practical approach to NLP is featured, with specific emphasis on its applications to radiology. A brief history of NLP, the strengths and challenges inherent to its use, and freely available resources and tools are covered to guide further exploration and study within the field. Particular attention is devoted to the recent development of the Word2Vec and BERT (Bidirectional Encoder Representations from Transformers) language models, which have exponentially increased the power and utility of NLP for a variety of applications. Online supplemental material is available for this article. ©RSNA, 2021.

Natural language processing in radiology: Clinical applications and future directions.

Natural language processing (NLP) is a wide range of techniques that allows computers to interact with human text. Applications of NLP in everyday life include language translation aids, chat bots, and text prediction. It has been increasingly utilized in the medical field with increased reliance on electronic health records. As findings in radiology are primarily communicated via text, the field is particularly suited to benefit from NLP based applications. Furthermore, rapidly increasing imaging volume will continue to increase burden on clinicians, emphasizing the need for improvements in workflow. In this article, we highlight the numerous non-clinical, provider focused, and patient focused applications of NLP in radiology. We also comment on challenges associated with development and incorporation of NLP based applications in radiology as well as potential future directions.

Gliomas: predicting time to progression or survival with cerebral blood volume measurements at dynamic susceptibility-weighted contrast-enhanced perfusion MR imaging.

PURPOSE: To retrospectively determine whether relative cerebral blood volume (CBV) measurements can be used to predict clinical outcome in patients with high-grade gliomas (HGGs) and low-grade gliomas (LGGs) and specifically whether patients who have gliomas with a high initial relative CBV have more rapid progression than those who have gliomas with a low relative CBV. MATERIALS AND METHODS: Approval for this retrospective HIPAA-compliant study was obtained from the Institutional Board of Research Associates, with waiver of informed consent. One hundred eighty-nine patients (122 male and 67 female patients; median age, 43 years; range, 4-80 years) were examined with dynamic susceptibility-weighted contrast material-enhanced perfusion magnetic resonance (MR) imaging and were followed up clinically with MR imaging (median follow-up, 334 days). Log-rank tests were used to evaluate the association between relative CBV and time to progression by using Kaplan-Meier curves. Binary logistic regression was used to determine whether age, sex, and relative CBV were associated with an adverse event (progressive disease or death). RESULTS: Values for the mean relative CBV for patients according to each clinical response were as follows: 1.41 +/- 0.13 (standard deviation) for complete response (n = 4), 2.36 +/- 1.78 for stable disease (n = 41), 4.84 +/- 3.32 for progressive disease (n = 130), and 3.82 +/- 1.93 for death (n = 14). Kaplan-Meier estimates of median time to progression in days indicated that patients with a relative CBV of less than 1.75 had a median time to progression of 3585 days, whereas patients with a relative CBV of more than 1.75 had a time to progression of 265 days. Age and relative CBV were also independent predictors for clinical outcome. CONCLUSION: Dynamic susceptibility-weighted contrast-enhanced perfusion MR imaging can be used to predict median time to progression in patients with gliomas, independent of pathologic findings. Patients who have HGGs and LGGs with a high relative CBV (>1.75) have a significantly more rapid time to progression than do patients who have gliomas with a low relative CBV.

Lung cancer staging: clinical and radiologic perspectives.

Published in 2009, the 7th edition of the American Joint Committee on Cancer TNM staging system is the culmination of an extensive worldwide initiative to standardize and validate lung cancer staging. Unlike prior editions, the new staging system is now inclusive of small cell carcinoma and carcinoid tumors. In addition, significant changes were made to the T and M descriptors, resulting in improved prognostic stratification of disease. This review article highlights these changes, the rationale for their inclusion in the new staging manual, and the role of the radiologist in determining stage.

Imaging Features following Thermal Ablation of Lung Malignancies.

Percutaneous image-guided thermal ablation is gaining attraction as an effective alternative to surgical resection for patients with primary and secondary malignancies of the lung. Currently, no standard follow-up imaging protocol has been established or uniformly accepted. The early identification of residual or recurrent tumor would in theory enable the practitioner to offer expeditious retreatment or alternative treatment. This review elaborates on the imaging findings following thermal ablation, both heat- and cold-based, of nonresectable pulmonary malignancies.

Relative cerebral blood volume measurements of low-grade gliomas predict patient outcome in a multi-institution setting.

BACKGROUND/PURPOSE: The prognostic value of defining subcategories of gliomas is still controversial. This study aims to determine the utility of relative cerebral blood volume (rCBV) in predicting clinical response in patients with low-grade glioma at multiple institutions. MATERIALS AND METHODS: Sixty-nine patients were studied with dynamic susceptibility contrast-enhanced perfusion MRI at two institutions. The pathologic diagnoses of the low-grade gliomas were 34 astrocytomas, 20 oligodendroglioma, 9 oligoastrocytomas, 1 ganglioglioma and 5 with indeterminate histology. Wilcoxon tests were used to compare patients in different response categories with respect to baseline rCBV. Kaplan-Meier curve and log-rank tests were used to predict the association of rCBV with time to progression. RESULTS: At both institutions, patients with an adverse event (progressive disease or death) had a significantly higher baseline rCBV than those without (complete response or stable disease) (p value=0.0138). The odds ratio for detecting an adverse event when using rCBV was 1.87 (95% confidence interval: 1.14-3.08). rCBV was significantly negatively associated with time to progression (p=0.005). The median time to progression among subjects with rCBV>1.75 was 365 days, while there was 95% confidence that the median time to progression was at least 889 days among subjects with rCBV<1.75. CONCLUSION: Our study suggests not only that rCBV measurements correlate well with time to progression or death, but also that the findings can be replicated across institutions, which supports the application of rCBV as an adjunct to pathology in predicting glioma biology.

Fibronectin protects prostate cancer cells from tumor necrosis factor-alpha-induced apoptosis via the AKT/survivin pathway.

Integrins are cell surface heterodimeric transmembrane receptors that, in addition to mediating cell adhesion to extracellular matrix proteins modulate cell survival. This mechanism may be exploited in cancer where evasion from apoptosis invariably contributes to cellular transformation. The molecular mechanisms responsible for matrix-induced survival signals begin to be elucidated. Here we report that the inhibitor of apoptosis survivin is expressed in vitro in human prostate cell lines with the highest levels present in aggressive prostate cancer cells such as PC3 and LNCaP-LN3 as well as in vivo in prostatic adenocarcinoma. We also show that interference with survivin in PC3 prostate cancer cells using a Cys84--> Ala dominant negative mutant or survivin antisense cDNA causes nuclear fragmentation, hypodiploidy, cleavage of a 32-kDa proform caspase-3 to active caspase-3, and proteolysis of the caspase substrate poly(ADP-ribose) polymerase. We demonstrate that in the aggressive PC3 cell line, adhesion to fibronectin via beta1 integrins results in up-regulation of survivin and protection from apoptosis induced by tumor necrosis factor-alpha (TNF-alpha). In contrast, survivin is not up-regulated by cell adhesion in the non-tumorigenic LNCaP cell line. Dominant negative survivin counteracts the ability of fibronectin to protect cells from undergoing apoptosis, whereas wild-type survivin protects non-adherent cells from TNF-alpha-induced apoptosis. Evidence is provided that expression of beta1A integrin is necessary to protect non-adherent cells transduced with survivin from TNF-alpha-induced apoptosis. In contrast, the beta1C integrin, which contains a variant cytoplasmic domain, is not able to prevent apoptosis induced by TNF-alpha in non-adherent cells transduced with survivin. Finally, we show that regulation of survivin levels by integrins are mediated by protein kinase B/AKT. These findings indicate that survivin is required to maintain a critical anti-apoptotic threshold in prostate cancer cells and identify integrin signaling as a crucial survival pathway against death receptor-mediated apoptosis.