RESUMEN
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.
Asunto(s)
Nódulo Tiroideo , Humanos , Femenino , Adulto , Nódulo Tiroideo/diagnóstico por imagen , Estudios Retrospectivos , Lenguaje , Redes Neurales de la Computación , Curva ROCRESUMEN
Background Noninvasive tests can be used to screen patients with chronic liver disease for advanced liver fibrosis; however, the use of single tests may not be adequate. Purpose To construct sequential clinical algorithms that include a US deep learning (DL) model and compare their ability to predict advanced liver fibrosis with that of other noninvasive tests. Materials and Methods This retrospective study included adult patients with a history of chronic liver disease or unexplained abnormal liver function test results who underwent B-mode US of the liver between January 2014 and September 2022 at three health care facilities. A US-based DL network (FIB-Net) was trained on US images to predict whether the shear-wave elastography (SWE) value was 8.7 kPa or higher, indicative of advanced fibrosis. In the internal and external test sets, a two-step algorithm (Two-step#1) using the Fibrosis-4 Index (FIB-4) followed by FIB-Net and a three-step algorithm (Three-step#1) using FIB-4 followed by FIB-Net and SWE were used to simulate screening scenarios where liver stiffness measurements were not or were available, respectively. Measures of diagnostic accuracy were calculated using liver biopsy as the reference standard and compared between FIB-4, SWE, FIB-Net, and European Association for the Study of the Liver guidelines (ie, FIB-4 followed by SWE), along with sequential algorithms. Results The training, validation, and test data sets included 3067 (median age, 42 years [IQR, 33-53 years]; 2083 male), 1599 (median age, 41 years [IQR, 33-51 years]; 1124 male), and 1228 (median age, 44 years [IQR, 33-55 years]; 741 male) patients, respectively. FIB-Net obtained a noninferior specificity with a margin of 5% (P < .001) compared with SWE (80% vs 82%). The Two-step#1 algorithm showed higher specificity and positive predictive value (PPV) than FIB-4 (specificity, 79% vs 57%; PPV, 44% vs 32%) while reducing unnecessary referrals by 42%. The Three-step#1 algorithm had higher specificity and PPV compared with European Association for the Study of the Liver guidelines (specificity, 94% vs 88%; PPV, 73% vs 64%) while reducing unnecessary referrals by 35%. Conclusion A sequential algorithm combining FIB-4 and a US DL model showed higher diagnostic accuracy and improved referral management for all-cause advanced liver fibrosis compared with FIB-4 or the DL model alone. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Ghosh in this issue.
Asunto(s)
Algoritmos , Diagnóstico por Imagen de Elasticidad , Cirrosis Hepática , Humanos , Masculino , Cirrosis Hepática/diagnóstico por imagen , Persona de Mediana Edad , Femenino , Estudios Retrospectivos , Diagnóstico por Imagen de Elasticidad/métodos , Adulto , Aprendizaje Profundo , Hígado/diagnóstico por imagen , Hígado/patología , Anciano , Ultrasonografía/métodosRESUMEN
PURPOSES: To evaluate the postsurgical prognostic implication of contrast-enhanced ultrasound (CEUS) for combined hepatocellular-cholangiocarcinoma (CHC). To build a CEUS-based early recurrence prediction classifier for CHC, in comparison with tumor-node-metastasis (TNM) staging. METHODS: The CEUS features and clinicopathological findings of each case were analyzed, and the Liver Imaging Reporting and Data System categories were assigned. The recurrence-free survival associated factors were evaluated by Cox proportional hazard model. Incorporating the independent factors, nomograms were built to estimate the possibilities of 3-month, 6-month, and 1-year recurrence and whose prognostic value was determined by time-dependent receiver operating characteristics, calibration curves, and hazard layering efficiency validation, comparing with TNM staging system. RESULTS: In the multivariable analysis, the levels of carbohydrate antigen 19-9, prothrombin time and total bilirubin, and tumor shape, the Liver Imaging Reporting and Data System category were independent factors for recurrence-free survival. The LR-M category showed longer recurrence-free survival than did the LR-4/5 category. The 3-month, 6-month, and 1-year area under the curves of the CEUS-clinical nomogram, clinical nomogram, and TNM staging system were 0.518, 0.552, and 0.843 versus 0.354, 0.240, and 0.624 (P = .048, .049, and .471) vs. 0.562, 0.545, and 0.843 (P = .630, .564, and .007), respectively. The calibration curves of the CEUS-clinical model at different prediction time pionts were all close to the ideal line. The CEUS-clinical model effectively stratified patients into groups of high and low risk of recurrence in both training and validation set, while the TNM staging system only works on the training set. CONCLUSIONS: Our CEUS-clinical nomogram is a reliable early recurrence prediction tool for hepatocellular-cholangiocarcinoma and helps postoperative risk stratification.
Asunto(s)
Neoplasias de los Conductos Biliares , Carcinoma Hepatocelular , Colangiocarcinoma , Neoplasias Hepáticas , Nomogramas , Neoplasias de los Conductos Biliares/patología , Conductos Biliares Intrahepáticos/patología , Carcinoma Hepatocelular/diagnóstico por imagen , Carcinoma Hepatocelular/cirugía , Colangiocarcinoma/diagnóstico por imagen , Colangiocarcinoma/patología , Colangiocarcinoma/cirugía , Humanos , Neoplasias Hepáticas/diagnóstico por imagen , Neoplasias Hepáticas/cirugía , Estudios RetrospectivosRESUMEN
OBJECTIVES: To construct a preoperative model for survival prediction in intrahepatic cholangiocarcinoma (ICC) patients using ultrasound (US) based radiographic-radiomics signatures. METHODS: Between April 2010 and September 2015, 170 patients with ICC who underwent curative resection were retrospectively recruited. Overall survival (OS)-related radiographic signatures and radiomics signatures based on preoperative US were built and assessed through a time-dependent receiver operating characteristic curve analysis. A nomogram was developed based on the selected predictors from the radiographic-radiomics signatures and clinical and laboratory results of the training cohort (n = 127), validated in an independent testing cohort (n = 43) by the concordance index (C-index), and compared with the Tumor Node Metastasis (TNM) cancer staging system as well as the radiographic and radiomics nomograms. RESULTS: The median areas under the curve of the radiomics signature and radiographic signature were higher than that of the TNM staging system in the testing cohort, although the values were not significantly different (0.76-0.82 versus 0.62, P = .485 and .264). The preoperative nomogram with CA 19-9, sex, ascites, radiomics signature, and radiographic signature had C-indexes of 0.72 and 0.75 in the training and testing cohorts, respectively, and it had significantly higher predictive performance than the 8th TNM staging system in the testing cohort (C-index: 0.75 versus 0.67, P = .004) and a higher C-index than the radiomics nomograms (0.75 versus 0.68, P = .044). CONCLUSIONS: The preoperative nomogram integrated with the radiographic-radiomics signature demonstrated good predictive performance for OS in ICC and was superior to the 8th TNM staging system.
Asunto(s)
Neoplasias de los Conductos Biliares , Colangiocarcinoma , Neoplasias de los Conductos Biliares/diagnóstico por imagen , Neoplasias de los Conductos Biliares/cirugía , Conductos Biliares Intrahepáticos , Colangiocarcinoma/diagnóstico por imagen , Colangiocarcinoma/cirugía , Humanos , Nomogramas , Estudios RetrospectivosRESUMEN
Artificial intelligence (AI) transforms medical images into high-throughput mineable data. Machine learning algorithms, which can be designed for modeling for lesion detection, target segmentation, disease diagnosis, and prognosis prediction, have markedly promoted precision medicine for clinical decision support. There has been a dramatic increase in the number of articles, including articles on ultrasound with AI, published in only a few years. Given the unique properties of ultrasound that differentiate it from other imaging modalities, including real-time scanning, operator-dependence, and multi-modality, readers should pay additional attention to assessing studies that rely on ultrasound AI. This review offers the readers a targeted guide covering critical points that can be used to identify strong and underpowered ultrasound AI studies.