Background Large language models (LLMs) hold substantial promise for medical imaging interpretation. However, there is a lack of studies on their feasibility in handling reasoning questions associated with medical diagnosis. Purpose To investigate the viability of leveraging three publicly available LLMs to enhance consistency and diagnostic accuracy in medical imaging based on standardized reporting, with pathology as the reference standard. Materials and Methods US images of thyroid nodules with pathologic results were retrospectively collected from a tertiary referral hospital between July 2022 and December 2022 and used to evaluate malignancy diagnoses generated by three LLMs-OpenAI's ChatGPT 3.5, ChatGPT 4.0, and Google's Bard. Inter- and intra-LLM agreement of diagnosis were evaluated. Then, diagnostic performance, including accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC), was evaluated and compared for the LLMs and three interactive approaches: human reader combined with LLMs, image-to-text model combined with LLMs, and an end-to-end convolutional neural network model. Results A total of 1161 US images of thyroid nodules (498 benign, 663 malignant) from 725 patients (mean age, 42.2 years ± 14.1 [SD]; 516 women) were evaluated. ChatGPT 4.0 and Bard displayed substantial to almost perfect intra-LLM agreement (κ range, 0.65-0.86 [95% CI: 0.64, 0.86]), while ChatGPT 3.5 showed fair to substantial agreement (κ range, 0.36-0.68 [95% CI: 0.36, 0.68]). ChatGPT 4.0 had an accuracy of 78%-86% (95% CI: 76%, 88%) and sensitivity of 86%-95% (95% CI: 83%, 96%), compared with 74%-86% (95% CI: 71%, 88%) and 74%-91% (95% CI: 71%, 93%), respectively, for Bard. Moreover, with ChatGPT 4.0, the image-to-text-LLM strategy exhibited an AUC (0.83 [95% CI: 0.80, 0.85]) and accuracy (84% [95% CI: 82%, 86%]) comparable to those of the human-LLM interaction strategy with two senior readers and one junior reader and exceeding those of the human-LLM interaction strategy with one junior reader. Conclusion LLMs, particularly integrated with image-to-text approaches, show potential in enhancing diagnostic medical imaging. ChatGPT 4.0 was optimal for consistency and diagnostic accuracy when compared with Bard and ChatGPT 3.5. © RSNA, 2024 Supplemental material is available for this article.

Objective To assess significant liver fibrosis by multiparametric ultrasomics data using machine learning. Materials and Methods This prospective study consisted of 144 patients with chronic hepatitis B. Ultrasomics—high-throughput quantitative data from ultrasound imaging of liver fibrosis—were generated using conventional radiomics, original radiofrequency (ORF) and contrast-enhanced micro-flow (CEMF) features. Three categories of features were explored using pairwise correlation and hierarchical clustering. Features were selected using diagnostic tests for fibrosis, activity and steatosis stage, with the histopathological results as the reference. The fibrosis staging performance of ultrasomics models with combinations of the selected features was evaluated with machine-learning algorithms by calculating the area under the receiver-operator characteristic curve (AUC). Results ORF and CEMF features had better predictive power than conventional radiomics for liver fibrosis stage (both p < 0.01). CEMF features exhibited the highest diagnostic value for activity stage (both p < 0.05), and ORF had the best diagnostic value for steatosis stage (both p < 0.01). The machine-learning classifiers of adaptive boosting, random forest and support vector machine were found to be optimal algorithms with better (all mean AUCs = 0.85) and more stable performance (coefficient of variation = 0.01–0.02) for fibrosis staging than decision tree, logistic regression and neural network (mean AUC = 0.61–0.72, CV = 0.07–0.08). The multiparametric ultrasomics model achieved much better performance (mean AUC values of 0.78–0.85) than the features from a single modality in discriminating significant fibrosis (≥ F2). Conclusion Machine-learning-based analysis of multiparametric ultrasomics can help improve the discrimination of significant fibrosis compared with mono or dual modalities. Key Points • Multiparametric ultrasomics has achieved much better performance in the discrimination of significant fibrosis (≥ F2) than the single modality of conventional radiomics, original radiofrequency and contrast-enhanced micro-flow. • Adaptive boosting, random forest and support vector machine are the optimal algorithms for machine learning.

Background and Aim: This study aims to construct a strategy that uses assistance from artificial intelligence (AI) to assist radiologists in the identification of malignant versus benign focal liver lesions (FLLs) using contrast‐enhanced ultrasound (CEUS).Methods: A training set (patients = 363) and a testing set (patients = 211) were collected from our institute. On four‐phase CEUS images in the training set, a composite deep learning architecture was trained and tuned for differentiating malignant and benign FLLs. In the test dataset, AI performance was evaluated by comparison with radiologists with varied levels of experience. Based on the comparison, an AI assistance strategy was constructed, and its usefulness in reducing CEUS interobserver heterogeneity was further tested.Results: In the test set, to identify malignant versus benign FLLs, AI achieved an area under the curve of 0.934 (95% CI 0.890–0.978) with an accuracy of 91.0%. Comparing with radiologists reviewing videos along with complementary patient information, AI outperformed residents (82.9–84.4%, P = 0.038) and matched the performance of experts (87.2–88.2%, P = 0.438). Due to the higher positive predictive value (PPV) (AI: 95.6% vs residents: 88.6–89.7%, P = 0.056), an AI strategy was defined to improve the malignant diagnosis. With the assistance of AI, radiologists exhibited a sensitivity improvement of 97.0–99.4% (P < 0.05) and an accuracy of 91.0–92.9% (P = 0.008–0.189), which was comparable with that of the experts (P = 0.904).Conclusions: The CEUS‐based AI strategy improved the performance of residents and reduced CEUS's interobserver heterogeneity in the differentiation of benign and malignant FLLs.

Background: Obstructive sleep apnea hypopnea syndrome (OSAHS) is a contributing factor for non-alcoholic fatty liver disease (NAFLD). Non-invasive algorithms including fatty liver index (FLI) and hepatic steatosis index (HSI) have been used as a screening test for NAFLD in epidemiologic studies. The aim of this study is to compare the diagnostic accuracy of FLI and HSI for NAFLD detection in adults with OSAHS. Methods: We enrolled consecutive adult subjects who were newly diagnosed with OSAHS from March 2016 to January 2018. NAFLD was diagnosed by ultrasonography. The accuracy and cut-off point of the FLI and HSI to detect NAFLD were assessed by analyzing the area under the receiver operating characteristic (AUROC) curve and the maximum Youden index analysis, respectively. Results: The 326 subjects were diagnosed as NAFLD according to ultrasound findings, while 105 subjects who had normal abdominal ultrasonography were grouped as controls. Both FLI and HSI values were significantly higher in patients with NAFLD compared with controls. The AUROC of FLI and HSI for predicting NAFLD was 0.802 (95% confidence interval [CI] 0.762–0.839) and 0.753 (95% CI 0.710–0.793), respectively. The AUROC of FLI was significantly higher than that of HSI (P = 0.0383). The optimal cut-off value of FLI and HSI was 60 (sensitivity 66% and specificity 80%) and 35 (sensitivity 81% and specificity 60%), respectively. Conclusions: Both FLI and HSI can serve as screening tools for NAFLD in OSAHS adults. The FLI shows better performance in diagnosing NAFLD than HSI. Trial registration: Chinese Clinical Trial Registry (No. ChiCTR-OOB-15007253), http://www.chictr.org.cn/showproj.aspx?proj=11606 .

By using the feature of tumor stiffness, the accuracy of preoperative staging for rectal tumors with shear-wave elastography was improved compared with endorectal US in the pilot study, but was not significantly different in the validation cohort, potentially due to small sample size. PurposeTo compare the value of endorectal ultrasonography (US) with shear-wave elastography (SWE) in staging rectal tumors before surgery.Materials and MethodsThis prospective study was approved by the institutional review board and written informed consent was obtained. In a pilot cohort from April 2015 to January 2016, 70 patients with rectal adenocarcinomas and/or adenomas confirmed with histopathologic examination underwent both endorectal US and SWE. Tumor stiffness and three regions of reference backgrounds, as well as tumor stiffness ratios (SRs) versus these backgrounds, were analyzed. The optimal staging feature was selected by using receiver operating characteristic analysis, and the concordance rate with pathologic stage was analyzed and compared with endorectal US. The results were validated in an independent cohort of 30 patients from February 2016 to July 2016.ResultsIn the pilot study, from rectal adenoma to stage T3 cancers, the tumor stiffness, stiffness of peritumoral fat, tumor SR versu...