Large language models' responses to liver cancer surveillance, diagnosis, and management questions: accuracy, reliability, readability.

PURPOSE: To assess the accuracy, reliability, and readability of publicly available large language models in answering fundamental questions on hepatocellular carcinoma diagnosis and management. METHODS: Twenty questions on liver cancer diagnosis and management were asked in triplicate to ChatGPT-3.5 (OpenAI), Gemini (Google), and Bing (Microsoft). Responses were assessed by six fellowship-trained physicians from three academic liver transplant centers who actively diagnose and/or treat liver cancer. Responses were categorized as accurate (score 1; all information is true and relevant), inadequate (score 0; all information is true, but does not fully answer the question or provides irrelevant information), or inaccurate (score - 1; any information is false). Means with standard deviations were recorded. Responses were considered as a whole accurate if mean score was > 0 and reliable if mean score was > 0 across all responses for the single question. Responses were also quantified for readability using the Flesch Reading Ease Score and Flesch-Kincaid Grade Level. Readability and accuracy across 60 responses were compared using one-way ANOVAs with Tukey's multiple comparison tests. RESULTS: Of the twenty questions, ChatGPT answered nine (45%), Gemini answered 12 (60%), and Bing answered six (30%) questions accurately; however, only six (30%), eight (40%), and three (15%), respectively, were both accurate and reliable. There were no significant differences in accuracy between any chatbot. ChatGPT responses were the least readable (mean Flesch Reading Ease Score 29; college graduate), followed by Gemini (30; college) and Bing (40; college; p < 0.001). CONCLUSION: Large language models provide complex responses to basic questions on hepatocellular carcinoma diagnosis and management that are seldomly accurate, reliable, or readable.

Differentiation of Hepatocellular Adenoma Subtypes and Focal Nodular Hyperplasia on Gadoxetate Disodium-Enhanced MRI: An Updated Diagnostic Algorithm.

This study evaluated an updated diagnostic algorithm for distinguishing HCA subtypes and FNH on gadoxetate disodium­enhanced MRI. The algorithm included a pathway recommending biopsy for indeterminate lesions that could represent HCA with beta-catenin mutations (which are at risk of malignant transformation) or I-HCA with an atypical MRI appearance.

Growth Kinetics of Pancreatic Neuroendocrine Neoplasms by Histopathologic Grade.

OBJECTIVES: The aims of the study are to describe the growth kinetics of pathologically proven, treatment-naive pancreatic neuroendocrine neoplasms (panNENs) at imaging surveillance and to determine their association with histopathologic grade and Ki-67. METHODS: This study included 100 panNENs from 95 patients who received pancreas protocol computed tomography or magnetic resonance imaging from January 2005 to July 2022. All masses were treatment-naive, had histopathologic correlation, and were imaged with at least 2 computed tomography or magnetic resonance imaging at least 90 days apart. Growth kinetics was assessed using linear and specific growth rate, stratified by grade and Ki-67. Masses were also assessed qualitatively to determine other possible imaging predictors of grade. RESULTS: There were 76 grade 1 masses, 17 grade 2 masses, and 7 grade 3 masses. Median (interquartile range) linear growth rates were 0.06 cm/y (0-0.20), 0.40 cm/y (0.22-1.06), and 2.70 cm/y (0.41-3.89) for grade 1, 2, and 3 masses, respectively (P < 0.001). Linear growth rate correlated with Ki-67 with r2 of 0.623 (P < 0.001). At multivariate analyses, linear growth rate was the only imaging feature significantly associated with grade (P = 0.009). CONCLUSIONS: Growth kinetics correlate with Ki-67 and grade. Grade 1 panNENs grow slowly versus grade 2-3 panNENs.

Accuracy of Information Provided by ChatGPT Regarding Liver Cancer Surveillance and Diagnosis.

ChatGPT did not reliably provide accurate information to 20 questions about liver cancer surveillance and diagnosis, as assessed by six physicians who actively diagnose and/or treat liver cancer. Answers deemed inaccurate commonly related to questions on specific LI-RADS categories and included contradictory or falsely reassuring, if not wrong, information.

Hepatocellular Adenoma Subtypes Based on 2017 Classification System: Exploratory Study of Gadoxetate Disodium-Enhanced MRI Features With Proposal of a Diagnostic Algorithm.

BACKGROUND. The classification of hepatocellular adenomas (HCAs) was updated in 2017 on the basis of genetic and molecular analysis. OBJECTIVE. The purpose of this article was to evaluate features on gadoxetate disodium-enhanced MRI of HCA subtypes on the basis of the 2017 classification and to propose a diagnostic algorithm for determining subtype using these features. METHODS. This retrospective study included 56 patients (49 women, seven men; mean age, 37 ± 13 [SD] years) with histologically confirmed HCA evaluated by gadoxetate disodium-enhanced MRI from January 2010 to January 2021. Subtypes were reclassified using 2017 criteria: hepatocyte nuclear factor-1α mutated HCA (HHCA), inflammatory HCA (IHCA), ß-catenin exon 3 activated HCA (ß-HCA), mixed inflammatory and ß-HCA (ß-IHCA), sonic hedgehog HCA (shHCA), and unclassified HCA (UHCA). Qualitative MRI features were assessed. Liver-to-lesion contrast enhancement ratios (LLCERs) were measured. Subtypes were compared, and a diagnostic algorithm was proposed. RESULTS. The analysis included 65 HCAs: 16 HHCAs, 31 IHCAs, six ß-HCA, four ß-IHCA, five shHCA, and three UHCAs. HHCAs showed homogeneous diffuse intralesional steatosis in 94%, whereas all other HCAs showed this finding in 0% (p < .001). IHCAs showed the "atoll" sign in 58%, whereas all other HCAs showed this finding in 12% (p < .001). IHCAs showed moderate T2 hyperintensity in 52%, whereas all other HCAs showed this finding in 12% (p < .001). The ß-HCAs and ß-IHCAs occurred in men in 63%, whereas all other HCAs occurred in men in 4% (p < .001). The ß-HCAs and ß-IHCAs had a mean size of 10.1 ± 6.8 cm, whereas all other HCAs had a mean size of 5.1 ± 2.9 cm (p = .03). The ß-HCAs and ß-IHCAs showed fluid components in 60%, whereas all other HCAs showed this finding in 5% (p < .001). Hepatobiliary phase iso- or hyperintensity was observed in 80% of ß-HCAs and ß-IHCAs versus 5% of all other HCAs (p < .001). Hepatobiliary phase LLCER was positive in nine HCAs (eight ß-HCAs and ß-IHCAs; one IHCA). The shHCA and UHCA did not show distinguishing features. The proposed diagnostic algorithm had accuracy of 98% for HHCAs, 83% for IHCAs, and 95% for ß-HCAs or ß-IHCAs. CONCLUSION. Findings on gadoxetate disodium-enhanced MRI, including hepatobiliary phase characteristics, were associated with HCA subtypes using the 2017 classification. CLINICAL IMPACT. The algorithm identified common HCA subtypes with high accuracy, including those with ß-catenin exon 3 mutations.