New Horizons: The Potential Role of OpenAI's ChatGPT in Clinical Radiology

Can large language models pass official high-grade exams of the European Society of Neuroradiology courses? A direct comparison between OpenAI chatGPT 3.5, OpenAI GPT4 and Google Bard

We compared different LLMs, notably chatGPT, GPT4, and Google Bard and we tested whether their performance differs in subspeciality domains, in executing examinations from four different courses of the European Society of Neuroradiology (ESNR) notably anatomy/embryology, neuro-oncology, head and neck and pediatrics. Written exams of ESNR were used as input data, related to anatomy/embryology (30 questions), neuro-oncology (50 questions), head and neck (50 questions), and pediatrics (50 questions). All exams together, and each exam separately were introduced to the three LLMs: chatGPT 3.5, GPT4, and Google Bard. Statistical analyses included a group-wise Friedman test followed by a pair-wise Wilcoxon test with multiple comparison corrections. Overall, there was a significant difference between the 3 LLMs (p < 0.0001), with GPT4 having the highest accuracy (70%), followed by chatGPT 3.5 (54%) and Google Bard (36%). The pair-wise comparison showed significant differences between chatGPT vs GPT 4 (p < 0.0001), chatGPT vs Bard (p < 0. 0023), and GPT4 vs Bard (p < 0.0001). Analyses per subspecialty showed the highest difference between the best LLM (GPT4, 70%) versus the worst LLM (Google Bard, 24%) in the head and neck exam, while the difference was least pronounced in neuro-oncology (GPT4, 62% vs Google Bard, 48%). We observed significant differences in the performance of the three different LLMs in the running of official exams organized by ESNR. Overall GPT 4 performed best, and Google Bard performed worst. This difference varied depending on subspeciality and was most pronounced in head and neck subspeciality.

ChatGPT in radiology: A systematic review of performance, pitfalls, and future perspectives

PURPOSE

The purpose of this study was to systematically review the reported performances of ChatGPT, identify potential limitations, and explore future directions for its integration, optimization, and ethical considerations in radiology applications.

MATERIALS AND METHODS

After a comprehensive review of PubMed, Web of Science, Embase, and Google Scholar databases, a cohort of published studies was identified up to January 1, 2024, utilizing ChatGPT for clinical radiology applications.

RESULTS

Out of 861 studies derived, 44 studies evaluated the performance of ChatGPT; among these, 37 (37/44; 84.1%) demonstrated high performance, and seven (7/44; 15.9%) indicated it had a lower performance in providing information on diagnosis and clinical decision support (6/44; 13.6%) and patient communication and educational content (1/44; 2.3%). Twenty-four (24/44; 54.5%) studies reported the proportion of ChatGPT's performance. Among these, 19 (19/24; 79.2%) studies recorded a median accuracy of 70.5%, and in five (5/24; 20.8%) studies, there was a median agreement of 83.6% between ChatGPT outcomes and reference standards [radiologists' decision or guidelines], generally confirming ChatGPT's high accuracy in these studies. Eleven studies compared two recent ChatGPT versions, and in ten (10/11; 90.9%), ChatGPTv4 outperformed v3.5, showing notable enhancements in addressing higher-order thinking questions, better comprehension of radiology terms, and improved accuracy in describing images. Risks and concerns about using ChatGPT included biased responses, limited originality, and the potential for inaccurate information leading to misinformation, hallucinations, improper citations and fake references, cybersecurity vulnerabilities, and patient privacy risks.

CONCLUSION

Although ChatGPT's effectiveness has been shown in 84.1% of radiology studies, there are still multiple pitfalls and limitations to address. It is too soon to confirm its complete proficiency and accuracy, and more extensive multicenter studies utilizing diverse datasets and pre-training techniques are required to verify ChatGPT's role in radiology.

Transcending Language Barriers: Can ChatGPT Be the Key to Enhancing Multilingual Accessibility in Health Care?

OBJECTIVE

To explore the capabilities of Chat Generative Pre-trained Transformer (ChatGPT) for the purpose of simplifying and translating radiology reports into Spanish, Hindi, and Russian languages, with comparisons to its performance in simplifying to the English language.

METHODS

Fifty deidentified abdomen-pelvis CT reports were fed to ChatGPT (4.0), instructing it to simplify and translate the report. The processed reports were rated on factual correctness (category 1), potential harmful errors (category 2), completeness (category 3), and explanation of medical terms (category 4). The translated versions were also rated on the quality of translation (category 5). The scores in each category were compared between the translated versions and each translated version was compared with the English version in the first four categories. The original reports and the simplified English reports were rated on the Flesch Reading Ease Score and the Flesch Kincaid Grade Level.

RESULTS

The Spanish translation outperformed the Hindi and Russian version significantly in categories 1 and 3 (P < .05). All translated versions performed significantly worse compared with the English version in category 4 (P < .001). Notably, the Hindi translated version performed significantly worse in all four categories (P < .05). The Russian translated version was also significantly worse in category 3 (P < .05). In the first three categories, the Spanish translation, and in the first two categories, the Russian translation demonstrated no statistically significant difference from the English version. No statistically significant difference was observed in the Flesch Reading Ease Score and Flesch Kincaid Grade Level of the simplified English reports. Typographical errors in the original reports negatively affected the translation.

CONCLUSION

ChatGPT demonstrates potential ability in translating reports and communicating pertinent clinical information with limited errors. More training and tailoring are required for languages that are not as commonly used in medical literature. Large language models can be used for translating and simplifying radiology reports, potentially improving access to health care and helping reduce health care costs.

Evaluation of ChatGPT-Generated Educational Patient Pamphlets for Common Interventional Radiology Procedures

RATIONALE AND OBJECTIVES

This study aimed to evaluate the accuracy and reliability of educational patient pamphlets created by ChatGPT, a large language model, for common interventional radiology (IR) procedures.

METHODS AND MATERIALS

Twenty frequently performed IR procedures were selected, and five users were tasked to independently request ChatGPT to generate educational patient pamphlets for each procedure using identical commands. Subsequently, two independent radiologists assessed the content, quality, and accuracy of the pamphlets. The review focused on identifying potential errors, inaccuracies, the consistency of pamphlets.

RESULTS

In a thorough analysis of the education pamphlets, we identified shortcomings in 30% (30/100) of pamphlets, with a total of 34 specific inaccuracies, including missing information about sedation for the procedure (10/34), inaccuracies related to specific procedural-related complications (8/34). A key-word co-occurrence network showed consistent themes within each group of pamphlets, while a line-by-line comparison at the level of users and across different procedures showed statistically significant inconsistencies (P < 0.001).

CONCLUSION

ChatGPT-generated education pamphlets demonstrated potential clinical relevance and fairly consistent terminology; however, the pamphlets were not entirely accurate and exhibited some shortcomings and inter-user structural variabilities. To ensure patient safety, future improvements and refinements in large language models are warranted, while maintaining human supervision and expert validation.

Application of ChatGPT for Orthopedic Surgeries and Patient Care

Artificial intelligence (AI) has rapidly transformed various aspects of life, and the launch of the chatbot "ChatGPT" by OpenAI in November 2022 has garnered significant attention and user appreciation. ChatGPT utilizes natural language processing based on a "generative pre-trained transfer" (GPT) model, specifically the transformer architecture, to generate human-like responses to a wide range of questions and topics. Equipped with approximately 57 billion words and 175 billion parameters from online data, ChatGPT has potential applications in medicine and orthopedics. One of its key strengths is its personalized, easy-to-understand, and adaptive response, which allows it to learn continuously through user interaction. This article discusses how AI, especially ChatGPT, presents numerous opportunities in orthopedics, ranging from preoperative planning and surgical techniques to patient education and medical support. Although ChatGPT's user-friendly responses and adaptive capabilities are laudable, its limitations, including biased responses and ethical concerns, necessitate its cautious and responsible use. Surgeons and healthcare providers should leverage the strengths of the ChatGPT while recognizing its current limitations and verifying critical information through independent research and expert opinions. As AI technology continues to evolve, ChatGPT may become a valuable tool in orthopedic education and patient care, leading to improved outcomes and efficiency in healthcare delivery. The integration of AI into orthopedics offers substantial benefits but requires careful consideration and continuous improvement.

Assessing the Responses of Large Language Models (ChatGPT-4, Gemini, and Microsoft Copilot) to Frequently Asked Questions in Breast Imaging: A Study on Readability and Accuracy

Background Large language models (LLMs), such as ChatGPT-4, Gemini, and Microsoft Copilot, have been instrumental in various domains, including healthcare, where they enhance health literacy and aid in patient decision-making. Given the complexities involved in breast imaging procedures, accurate and comprehensible information is vital for patient engagement and compliance. This study aims to evaluate the readability and accuracy of the information provided by three prominent LLMs, ChatGPT-4, Gemini, and Microsoft Copilot, in response to frequently asked questions in breast imaging, assessing their potential to improve patient understanding and facilitate healthcare communication. Methodology We collected the most common questions on breast imaging from clinical practice and posed them to LLMs. We then evaluated the responses in terms of readability and accuracy. Responses from LLMs were analyzed for readability using the Flesch Reading Ease and Flesch-Kincaid Grade Level tests and for accuracy through a radiologist-developed Likert-type scale. Results The study found significant variations among LLMs. Gemini and Microsoft Copilot scored higher on readability scales (p < 0.001), indicating their responses were easier to understand. In contrast, ChatGPT-4 demonstrated greater accuracy in its responses (p < 0.001). Conclusions While LLMs such as ChatGPT-4 show promise in providing accurate responses, readability issues may limit their utility in patient education. Conversely, Gemini and Microsoft Copilot, despite being less accurate, are more accessible to a broader patient audience. Ongoing adjustments and evaluations of these models are essential to ensure they meet the diverse needs of patients, emphasizing the need for continuous improvement and oversight in the deployment of artificial intelligence technologies in healthcare.

Assessing ChatGPT 4.0's test performance and clinical diagnostic accuracy on USMLE STEP 2 CK and clinical case reports

While there is data assessing the test performance of artificial intelligence (AI) chatbots, including the Generative Pre-trained Transformer 4.0 (GPT 4) chatbot (ChatGPT 4.0), there is scarce data on its diagnostic accuracy of clinical cases. We assessed the large language model (LLM), ChatGPT 4.0, on its ability to answer questions from the United States Medical Licensing Exam (USMLE) Step 2, as well as its ability to generate a differential diagnosis based on corresponding clinical vignettes from published case reports. A total of 109 Step 2 Clinical Knowledge (CK) practice questions were inputted into both ChatGPT 3.5 and ChatGPT 4.0, asking ChatGPT to pick the correct answer. Compared to its previous version, ChatGPT 3.5, we found improved accuracy of ChatGPT 4.0 when answering these questions, from 47.7 to 87.2% (p = 0.035) respectively. Utilizing the topics tested on Step 2 CK questions, we additionally found 63 corresponding published case report vignettes and asked ChatGPT 4.0 to come up with its top three differential diagnosis. ChatGPT 4.0 accurately created a shortlist of differential diagnoses in 74.6% of the 63 case reports (74.6%). We analyzed ChatGPT 4.0's confidence in its diagnosis by asking it to rank its top three differentials from most to least likely. Out of the 47 correct diagnoses, 33 were the first (70.2%) on the differential diagnosis list, 11 were second (23.4%), and three were third (6.4%). Our study shows the continued iterative improvement in ChatGPT's ability to answer standardized USMLE questions accurately and provides insights into ChatGPT's clinical diagnostic accuracy.

Evaluating capabilities of large language models: Performance of GPT-4 on surgical knowledge assessments

BACKGROUND

Artificial intelligence has the potential to dramatically alter health care by enhancing how we diagnose and treat disease. One promising artificial intelligence model is ChatGPT, a general-purpose large language model trained by OpenAI. ChatGPT has shown human-level performance on several professional and academic benchmarks. We sought to evaluate its performance on surgical knowledge questions and assess the stability of this performance on repeat queries.

METHODS

We evaluated the performance of ChatGPT-4 on questions from the Surgical Council on Resident Education question bank and a second commonly used surgical knowledge assessment, referred to as Data-B. Questions were entered in 2 formats: open-ended and multiple-choice. ChatGPT outputs were assessed for accuracy and insights by surgeon evaluators. We categorized reasons for model errors and the stability of performance on repeat queries.

RESULTS

A total of 167 Surgical Council on Resident Education and 112 Data-B questions were presented to the ChatGPT interface. ChatGPT correctly answered 71.3% and 67.9% of multiple choice and 47.9% and 66.1% of open-ended questions for Surgical Council on Resident Education and Data-B, respectively. For both open-ended and multiple-choice questions, approximately two-thirds of ChatGPT responses contained nonobvious insights. Common reasons for incorrect responses included inaccurate information in a complex question (n = 16, 36.4%), inaccurate information in a fact-based question (n = 11, 25.0%), and accurate information with circumstantial discrepancy (n = 6, 13.6%). Upon repeat query, the answer selected by ChatGPT varied for 36.4% of questions answered incorrectly on the first query; the response accuracy changed for 6/16 (37.5%) questions.

CONCLUSION

Consistent with findings in other academic and professional domains, we demonstrate near or above human-level performance of ChatGPT on surgical knowledge questions from 2 widely used question banks. ChatGPT performed better on multiple-choice than open-ended questions, prompting questions regarding its potential for clinical application. Unique to this study, we demonstrate inconsistency in ChatGPT responses on repeat queries. This finding warrants future consideration including efforts at training large language models to provide the safe and consistent responses required for clinical application. Despite near or above human-level performance on question banks and given these observations, it is unclear whether large language models such as ChatGPT are able to safely assist clinicians in providing care.

Current applications and future potential of ChatGPT in radiology: A systematic review

This study aimed to comprehensively evaluate the current utilization and future potential of ChatGPT, an AI-based chat model, in the field of radiology. The primary focus is on its role in enhancing decision-making processes, optimizing workflow efficiency, and fostering interdisciplinary collaboration and teaching within healthcare. A systematic search was conducted in PubMed, EMBASE and Web of Science databases. Key aspects, such as its impact on complex decision-making, workflow enhancement and collaboration, were assessed. Limitations and challenges associated with ChatGPT implementation were also examined. Overall, six studies met the inclusion criteria and were included in our analysis. All studies were prospective in nature. A total of 551 chatGPT (version 3.0 to 4.0) assessment events were included in our analysis. Considering the generation of academic papers, ChatGPT was found to output data inaccuracies 80% of the time. When ChatGPT was asked questions regarding common interventional radiology procedures, it contained entirely incorrect information 45% of the time. ChatGPT was seen to better answer US board-style questions when lower order thinking was required (P = 0.002). Improvements were seen between chatGPT 3.5 and 4.0 in regard to imaging questions with accuracy rates of 61 versus 85%(P = 0.009). ChatGPT was observed to have an average translational ability score of 4.27/5 on the Likert scale regarding CT and MRI findings. ChatGPT demonstrates substantial potential to augment decision-making and optimizing workflow. While ChatGPT's promise is evident, thorough evaluation and validation are imperative before widespread adoption in the field of radiology.

Analysis of ChatGPT publications in radiology: Literature so far

OBJECTIVE

To perform a detailed qualitative and quantitative analysis of the published literature on ChatGPT and radiology in the nine months since its public release, detailing the scope of the work in the short timeframe.

METHODS

A systematic literature search was carried out of the MEDLINE, EMBASE databases through August 15, 2023 for articles that were focused on ChatGPT and imaging/radiology. Articles were classified into original research and reviews/perspectives. Quantitative analysis was carried out by two experienced radiologists using objective scoring systems for evaluating original and non-original research.

RESULTS

51 articles were published involving ChatGPT and radiology/imaging dating from 26 Jan 2023 to the last article published on 14 Aug 2023. 23 articles were original research while the rest included reviews/perspectives or brief communications. For quantitative analysis scored by two readers, we included 23 original research and 17 non-original research articles (after excluding 11 letters as responses to previous articles). Mean score for original research was 3.20 out of 5 (across five questions), while mean score for non-original research was 1.17 out of 2 (across six questions). Mean score grading performance of ChatGPT in original research was 3.20 out of five (across two questions).

DISCUSSION

While it is early days for ChatGPT and its impact in radiology, there has already been a plethora of articles talking about the multifaceted nature of the tool and how it can impact every aspect of radiology from patient education, pre-authorization, protocol selection, generating differentials, to structuring radiology reports. Most articles show impressive performance of ChatGPT which can only improve with more research and improvements in the tool itself. There have also been several articles which have highlighted the limitations of ChatGPT in its current iteration, which will allow radiologists and researchers to improve these areas.

Evaluating the Use of ChatGPT to Accurately Simplify Patient-centered Information about Breast Cancer Prevention and Screening

Purpose To evaluate the use of ChatGPT as a tool to simplify answers to common questions about breast cancer prevention and screening. Materials and Methods In this retrospective, exploratory study, ChatGPT was requested to simplify responses to 25 questions about breast cancer to a sixth-grade reading level in March and August 2023. Simplified responses were evaluated for clinical appropriateness. All original and simplified responses were assessed for reading ease on the Flesch Reading Ease Index and for readability on five scales: Flesch-Kincaid Grade Level, Gunning Fog Index, Coleman-Liau Index, Automated Readability Index, and the Simple Measure of Gobbledygook (ie, SMOG) Index. Mean reading ease, readability, and word count were compared between original and simplified responses using paired t tests. McNemar test was used to compare the proportion of responses with adequate reading ease (score of 60 or greater) and readability (sixth-grade level). Results ChatGPT improved mean reading ease (original responses, 46 vs simplified responses, 70; P < .001) and readability (original, grade 13 vs simplified, grade 8.9; P < .001) and decreased word count (original, 193 vs simplified, 173; P < .001). Ninety-two percent (23 of 25) of simplified responses were considered clinically appropriate. All 25 (100%) simplified responses met criteria for adequate reading ease, compared with only two of 25 original responses (P < .001). Two of the 25 simplified responses (8%) met criteria for adequate readability. Conclusion ChatGPT simplified answers to common breast cancer screening and prevention questions by improving the readability by four grade levels, though the potential to produce incorrect information necessitates physician oversight when using this tool. Keywords: Mammography, Screening, Informatics, Breast, Education, Health Policy and Practice, Oncology, Technology Assessment Supplemental material is available for this article. © RSNA, 2023.

ChatGPT encounters multiple opportunities and challenges in neurosurgery

BACKGROUND

ChatGPT, powered by the GPT model and Transformer architecture, has demonstrated remarkable performance in the domains of medicine and healthcare, providing customized and informative responses. In our study, we investigated the potential of ChatGPT in the field of neurosurgery, focusing on its applications at the patient, neurosurgery student/resident, and neurosurgeon levels.

METHOD

The authors conducted inquiries with ChatGPT from the viewpoints of patients, neurosurgery students/residents, and neurosurgeons, covering a range of topics, such as disease diagnosis, treatment options, prognosis, rehabilitation, and patient care. The authors also explored concepts related to neurosurgery, including fundamental principles and clinical aspects, as well as tools and techniques to enhance the skills of neurosurgery students/residents. Additionally, the authors examined disease-specific medical interventions and the decision-making processes involved in clinical practice.

RESULTS

The authors received individual responses from ChatGPT, but they tended to be shallow and repetitive, lacking depth and personalization. Furthermore, ChatGPT may struggle to discern a patient's emotional state, hindering the establishment of rapport and the delivery of appropriate care. The language used in the medical field is influenced by technical and cultural factors, and biases in the training data can result in skewed or inaccurate responses. Additionally, ChatGPT's limitations include the inability to conduct physical examinations or interpret diagnostic images, potentially overlooking complex details and individual nuances in each patient's case. Moreover, its absence in the surgical setting limits its practical utility.

CONCLUSION

Although ChatGPT is a powerful language model, it cannot substitute for the expertise and experience of trained medical professionals. It lacks the capability to perform physical examinations, make diagnoses, administer treatments, establish trust, provide emotional support, and assist in the recovery process. Moreover, the implementation of Artificial Intelligence in healthcare necessitates careful consideration of legal and ethical concerns. While recognizing the potential of ChatGPT, additional training with comprehensive data is necessary to fully maximize its capabilities.

Academic integrity and artificial intelligence: is ChatGPT hype, hero or heresy?

Academic integrity in both higher education and scientific writing has been challenged by developments in artificial intelligence. The limitations associated with algorithms have been largely overcome by the recently released ChatGPT; a chatbot powered by GPT-3.5 capable of producing accurate and human-like responses to questions in real-time. Despite the potential benefits, ChatGPT confronts significant limitations to its usefulness in nuclear medicine and radiology. Most notably, ChatGPT is prone to errors and fabrication of information which poses a risk to professionalism, ethics and integrity. These limitations simultaneously undermine the value of ChatGPT to the user by not producing outcomes at the expected standard. Nonetheless, there are a number of exciting applications of ChatGPT in nuclear medicine across education, clinical and research sectors. Assimilation of ChatGPT into practice requires redefining of norms, and re-engineering of information expectations.

Evaluating Capabilities of Large Language Models: Performance of GPT4 on Surgical Knowledge Assessments

Background

Artificial intelligence (AI) has the potential to dramatically alter healthcare by enhancing how we diagnosis and treat disease. One promising AI model is ChatGPT, a large general-purpose language model trained by OpenAI. The chat interface has shown robust, human-level performance on several professional and academic benchmarks. We sought to probe its performance and stability over time on surgical case questions.

Methods

We evaluated the performance of ChatGPT-4 on two surgical knowledge assessments: the Surgical Council on Resident Education (SCORE) and a second commonly used knowledge assessment, referred to as Data-B. Questions were entered in two formats: open-ended and multiple choice. ChatGPT output were assessed for accuracy and insights by surgeon evaluators. We categorized reasons for model errors and the stability of performance on repeat encounters.

Results

A total of 167 SCORE and 112 Data-B questions were presented to the ChatGPT interface. ChatGPT correctly answered 71% and 68% of multiple-choice SCORE and Data-B questions, respectively. For both open-ended and multiple-choice questions, approximately two-thirds of ChatGPT responses contained non-obvious insights. Common reasons for inaccurate responses included: inaccurate information in a complex question (n=16, 36.4%); inaccurate information in fact-based question (n=11, 25.0%); and accurate information with circumstantial discrepancy (n=6, 13.6%). Upon repeat query, the answer selected by ChatGPT varied for 36.4% of inaccurate questions; the response accuracy changed for 6/16 questions.

Conclusion

Consistent with prior findings, we demonstrate robust near or above human-level performance of ChatGPT within the surgical domain. Unique to this study, we demonstrate a substantial inconsistency in ChatGPT responses with repeat query. This finding warrants future consideration and presents an opportunity to further train these models to provide safe and consistent responses. Without mental and/or conceptual models, it is unclear whether language models such as ChatGPT would be able to safely assist clinicians in providing care.