Large Language Models in Cancer Imaging: Applications and Future Perspectives DOI Open Access
Mickaël Tordjman,

Ian Bolger,

Murat Yüce

et al.

Journal of Clinical Medicine, Journal Year: 2025, Volume and Issue: 14(10), P. 3285 - 3285

Published: May 8, 2025

Recently, there has been tremendous interest on the use of large language models (LLMs) in radiology. LLMs have employed for various applications cancer imaging, including improving reporting speed and accuracy via generation standardized reports, automating classification staging abnormal findings incorporating appropriate guidelines, calculating individualized risk scores. Another is their ability to improve patient comprehension imaging reports with simplification medical terms possible translations multiple languages. Additional future include multidisciplinary tumor board standardizations, aiding management, preventing predicting adverse events (contrast allergies, MRI contraindications) research. However, limitations such as hallucinations variable performances could present obstacles widespread clinical implementation. Herein, we a review current well pitfalls limitations.

Language: Английский

Multimodal Large Language Model With Knowledge Retrieval Using Flowchart Embedding for Forming Follow-Up Recommendations for Pancreatic Cystic Lesions DOI
Zheren Zhu, Jin Liu, Cheng William Hong

et al.

American Journal of Roentgenology, Journal Year: 2025, Volume and Issue: unknown

Published: April 16, 2025

Background: The American College of Radiology (ACR) Incidental Findings Committee (IFC) algorithm provides guidance for pancreatic cystic lesions (PCL) management. Its implementation using plain-text large language model (LLM) solutions is challenging given that key components include multimodal data (e.g., figures and tables). Objective: To evaluate a LLM approach incorporating knowledge retrieval flowchart embedding forming follow-up recommendations PCL Methods: This retrospective study included patients who underwent abdominal CT or MRI from September 1, 2023 to 2024 which the report mentioned PCL. Reports' findings sections were inputted (GPT-4o). For task 1 [198 (mean age, 69.0±13.0 years; 110 women, 88 men)], assessed features (presence, size location, main duct communication, worrisome high-risk stigmata) formed recommendation three methods [default knowledge; retrieval-augmented generation (RAG) ACR IFC PDF document; LLM's image-to-text conversion in-context integration document's flowcharts tables]. 2 [85 initial 69.2±10.8 48 37 men], an additional relevant prior was inputted; interval change provided adjusted schedule accounting imaging embedding. Three radiologists accuracy in consensus independently; one radiologist 2. Results: with had 98.0-99.0%. Accuracy default knowledge, RAG, 42.4%, 23.7%, 89.9% (p<.001); 39.9%, 24.2%, 91.9% 3 40.9%, 25.3%, (p<.001). 2, demonstrated 96.5% schedules 81.2%. Conclusion: Multimodal aided automated provision adherent clinical document. Clinical Impact: framework could be extended other incidental through use documents as input.

Language: Английский

Citations

2
The Role of AI in Lymphoma: An Update DOI Creative Commons

James Cairns,

Russell Frood, Chirag Patel

et al.

Seminars in Nuclear Medicine, Journal Year: 2025, Volume and Issue: unknown

Published: March 1, 2025

Malignant lymphomas encompass a range of malignancies with incidence rising globally, particularly age. In younger populations, Hodgkin and Burkitt predominate, while older populations more commonly experience subtypes such as diffuse large B-cell, follicular, marginal zone, mantle cell lymphomas. Positron emission tomography/computed tomography (PET/CT) using [18F] fluorodeoxyglucose (FDG) is the gold standard for staging, treatment response assessment, prognostication in lymphoma. However, interpretation PET/CT complex, time-consuming, reliant on expert imaging specialists, exacerbating challenges associated workforce shortages worldwide. Artificial intelligence (AI) offers transformative potential across multiple aspects this setting. AI applications appointment planning have demonstrated utility reducing nonattendance rates improving departmental efficiency. Advanced reconstruction techniques leveraging convolutional neural networks (CNNs) enable reduced injected activities radiopharmaceutical patient dose whilst maintaining diagnostic accuracy, benefiting patients requiring scans. Automated segmentation tools, predominantly 3D U-Net architectures, improved quantification metrics total metabolic tumour volume (TMTV) lesion glycolysis (TLG), facilitating stratification. Despite these advancements, remain, including variability performance, impact Deauville Score interpretation, standardization TMTV/TLG measurements. Emerging language models (LLMs) also show promise enhancing reporting, converting free-text reports into structured formats, communication. Further research required to address limitations AI-induced errors, physiological uptake differentiation, integration clinical workflows. With robust validation harmonization, could significantly enhance lymphoma care, precision, workflow efficiency, outcomes.

Language: Английский

Citations

1
Multi-modal large language models in radiology: principles, applications, and potential DOI
Yiqiu Shen,

Yanqi Xu,

Jiajian Ma

et al.

Abdominal Radiology, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 2, 2024

Language: Английский

Citations

4
Künstliche Intelligenz in der Radiologie DOI
Moritz C. Halfmann, Peter Mildenberger, Tobias Jorg

et al.

Deleted Journal, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 4, 2025

Aufgrund des andauernden, rapiden Fortschritts künstlicher Intelligenz (KI) inklusive Large Language Models (LLMs) werden Radiolog*innen in absehbarer Zeit vor die Herausforderung der verantwortungsvollen klinischen Integration dieser Modelle gestellt. Ziel Arbeit ist es, einen Überblick über aktuelle Entwicklungen zum Thema LLMs, mögliche Einsatzgebiete Radiologie sowie ihre (zukünftige) Relevanz und Limitationen zu liefern. In Übersichtsarbeit wurden Publikationen LLMs für spezifische Anwendungen Medizin analysiert. Zusätzlich wurde Literatur den Herausforderungen im Zusammenhang mit einer LLM-Nutzung gesichtet zusammengefasst. Neben einem generellen radiologischen Anwendungsbeispielen von verschiedene besonders spannende Arbeiten empfohlen. Um anstehende klinische ermöglichen, müssen sich Thematik auseinandersetzen, Anwendungsgebiete möglicher kennen, um Hinblick auf Patientensicherheit, Ethik Datenschutz bewältigen können.

Citations

0
Assessing large language models for Lugano classification of malignant lymphoma in Japanese FDG-PET reports DOI Creative Commons
Rintaro Ito, Keita Kato,

Kosuke Nanataki

et al.

Deleted Journal, Journal Year: 2025, Volume and Issue: 9(1)

Published: March 9, 2025

This study evaluates the performance of four large language models (LLMs) in classifying malignant lymphoma stages using Lugano classification from free-text FDG-PET reports Japanese Specifically, we assess GPT-4o, Claude 3.5 Sonnet, Llama 3 70B, and Gemma 2 27B their ability interpret unstructured radiology texts. In a retrospective single-center study, 80 patients who underwent staging FDG-PET/CT for were included. The "Findings" sections analyzed without pre-processing. Each LLM assigned based on these reports. Performance was compared to reference standard determined by expert radiologists. Statistical analyses involved overall accuracy, weighted kappa agreement. GPT-4o achieved highest accuracy at 75% (60/80 cases) with substantial agreement (weighted κ = 0.801). Sonnet had 61.3% (49/80, 0.763). 70B showed accuracies 58.8% 57.5%, respectively, all indicating outperformed other LLMs assigning demonstrated potential advanced clinical While immediate utility automatically predicting stage an existing report may be limited, results highlight value understanding standardizing data.

Language: Английский

Citations

0
Harnessing the Power of Generative AI to Enhance Radiologist Efficiency and Accuracy DOI
Paul Babyn, Scott Adams

Radiology, Journal Year: 2025, Volume and Issue: 314(3)

Published: March 1, 2025

HomeRadiologyVol. 314, No. 3 PreviousNext Reviews and CommentaryEditorialHarnessing the Power of Generative AI to Enhance Radiologist Efficiency AccuracyPaul S. Babyn1 , Scott J. Adams1Paul Adams1Author Affiliations1Department Medical Imaging, Royal University Hospital, Saskatchewan, 103 Hospital Dr, Saskatoon, SK, Canada S7N 0W8Address correspondence to: P.S.B. (email: [email protected])Paul Adams1Published Online:Mar 11 2025https://doi.org/10.1148/radiol.250339See also article by Hong et al in this issue.MoreSectionsFull textPDF ToolsAdd favoritesCiteTrack CitationsPermissionsReprints ShareShare onFacebookXLinked In References1. Shen Y, Xu Ma J, al. Multi-modal large language models radiology: principles, applications, potential. Abdom Radiol (NY) 2024. 10.1007/s00261-024-04708-8. Published online December 2, Google Scholar2. Kim K, Cho Jang R, Updated primer on generative artificial intelligence medical imaging for professionals. Korean J 2024;25(3):224–242. Medline Scholar3. EK, Roh B, Park Value using a model chest radiography reporting: reader study. Radiology 2025;314(3):e241646. Scholar4. Sacoransky E, Kwan BYM, Soboleski D. ChatGPT assistive structured radiology systematic review. Curr Probl Diagn 2024;53(6):728–737. Scholar5. Zhang L, Liu M, Wang Constructing generate impressions from findings reports. 2024;312(3):e240885. Scholar6. Amin KS, Davis MA, Doshi Haims AH, Khosla P, Forman HP. Accuracy ChatGPT, Bard, Microsoft Bing simplifying 2023;309(2):e232561. Scholar7. Lee S, Youn H, Yoon SH. CXR-LLaVA: multimodal interpreting x-ray images. Eur 2025. 10.1007/s00330-024-11339-6. January 15, Scholar8. Shin HJ, Han Ryu EK. The impact reading times radiologists radiographs. NPJ Digit Med 2023;6(1):82. Scholar9. Yu F, Endo Krishnan Evaluating progress automatic report generation. Patterns 2023;4(9):100802. Scholar10. Gefter WB, Prokop Seo JB, Raoof Langlotz CP, Hatabu H. Human-AI symbiosis: path forward improve role patient care. 2024;310(1):e232778. ScholarArticle HistoryReceived: Jan 29 2025Revision requested: Feb received: 12 2025Accepted: 13 2025Published online: Mar 2025 FiguresReferencesRelatedDetailsAccompanying This ArticleValue Using Model Chest Radiography Reporting: A Reader StudyMar 2025Radiology Vol. Metrics Altmetric Score PDF download

Language: Английский

Citations

0
GPT-4o in radiology: In-context learning based automatic generation of radiology impressions DOI Creative Commons
Mohammed A. Mahyoub, Yong Wang, Mohammad T. Khasawneh

et al.

Natural Language Processing Journal, Journal Year: 2025, Volume and Issue: unknown, P. 100145 - 100145

Published: April 1, 2025

Language: Английский

Citations

0
Roles and Potential of Large Language Models in Healthcare: A Comprehensive Review DOI Creative Commons
Chi‐Hung Lin, Chang‐Fu Kuo

Biomedical Journal, Journal Year: 2025, Volume and Issue: unknown, P. 100868 - 100868

Published: April 1, 2025

Large Language Models (LLMs) are capable of transforming healthcare by demonstrating remarkable capabilities in language understanding and generation. They have matched or surpassed human performance standardized medical examinations assisted diagnostics across specialties like dermatology, radiology, ophthalmology. LLMs can enhance patient education providing accurate, readable, empathetic responses, they streamline clinical workflows through efficient information extraction from unstructured data such as notes. Integrating LLM into practice involves user interface design, clinician training, effective collaboration between Artificial Intelligence (AI) systems professionals. Users must possess a solid generative AI domain knowledge to assess the generated content critically. Ethical considerations ensure privacy, security, mitigating biases, maintaining transparency critical for responsible deployment. Future directions include interdisciplinary collaboration, developing new benchmarks that incorporate safety ethical measures, advancing multimodal integrate text imaging data, creating LLM-based agents complex decision-making, addressing underrepresented rare diseases, integrating with robotic precision procedures. Emphasizing safety, integrity, human-centered implementation is essential maximizing benefits LLMs, while potential risks, thereby helping these tools rather than replace expertise compassion healthcare.

Language: Английский

Citations

0
Evaluation of large language models in generating pulmonary nodule follow-up recommendations DOI
Jianbo Wen,

Wanyue Huang,

Huzheng Yan

et al.

European Journal of Radiology Open, Journal Year: 2025, Volume and Issue: 14, P. 100655 - 100655

Published: April 30, 2025

Language: Английский

Citations

0
The Role of Large Language Models (LLMs) in Breast Imaging Today and in the Near Future DOI
Simone Schiaffino, Tianyu Zhang, Ritse M. Mann

et al.

Journal of Magnetic Resonance Imaging, Journal Year: 2025, Volume and Issue: unknown

Published: May 4, 2025

ABSTRACT This narrative review focuses on the integration of large language models (LLMs), such as GPT‐4 and Gemini, into breast imaging. LLMs excel in understanding, processing, generating human‐like text, with potential applications ranging widely from decision‐making to radiology reporting support. show promise addressing current critical challenges, including rising demands for imaging services concurrent an increasing shortage radiologist workforce. Their ability integrate clinical guidelines generate standardized, evidence‐based reports has improve diagnostic consistency reduce inter‐reader variability. Emerging multimodal capabilities further extend their utility, enabling textual visual data tasks tumor classification decision‐making. Despite these advancements, significant challenges remain. often suffer limitations hallucinations, biases training datasets, domain‐specific knowledge gaps. These issues can affect reliability, particularly nuanced like Breast Imaging Reporting Data System categorization image assessment. Moreover, ethical concerns about privacy, biased outputs, regulatory compliance must be addressed before effective deployment setting. Current studies suggest that while complement human expertise, performance still lags behind radiologists key areas, requiring complex medical reasoning or direct analysis. Looking ahead, are poised play a crucial role by optimizing workflows, supporting multidisciplinary meetings, improving patient education. However, successful will depend proper context training, robust validation, oversight, supervision safeguard. Evidence Level 5. Technical Efficacy Stage 2.

Language: Английский

Citations

0