Comparative analysis of GPT-4-based ChatGPT's diagnostic performance with radiologists using real-world radiology reports of brain tumors.

OBJECTIVES: Large language models like GPT-4 have demonstrated potential for diagnosis in radiology. Previous studies investigating this potential primarily utilized quizzes from academic journals. This study aimed to assess the diagnostic capabilities of GPT-4-based Chat Generative Pre-trained Transformer (ChatGPT) using actual clinical radiology reports of brain tumors and compare its performance with that of neuroradiologists and general radiologists. METHODS: We collected brain MRI reports written in Japanese from preoperative brain tumor patients at two institutions from January 2017 to December 2021. The MRI reports were translated into English by radiologists. GPT-4 and five radiologists were presented with the same textual findings from the reports and asked to suggest differential and final diagnoses. The pathological diagnosis of the excised tumor served as the ground truth. McNemar's test and Fisher's exact test were used for statistical analysis. RESULTS: In a study analyzing 150 radiological reports, GPT-4 achieved a final diagnostic accuracy of 73%, while radiologists' accuracy ranged from 65 to 79%. GPT-4's final diagnostic accuracy using reports from neuroradiologists was higher at 80%, compared to 60% using those from general radiologists. In the realm of differential diagnoses, GPT-4's accuracy was 94%, while radiologists' fell between 73 and 89%. Notably, for these differential diagnoses, GPT-4's accuracy remained consistent whether reports were from neuroradiologists or general radiologists. CONCLUSION: GPT-4 exhibited good diagnostic capability, comparable to neuroradiologists in differentiating brain tumors from MRI reports. GPT-4 can be a second opinion for neuroradiologists on final diagnoses and a guidance tool for general radiologists and residents. CLINICAL RELEVANCE STATEMENT: This study evaluated GPT-4-based ChatGPT's diagnostic capabilities using real-world clinical MRI reports from brain tumor cases, revealing that its accuracy in interpreting brain tumors from MRI findings is competitive with radiologists. KEY POINTS: We investigated the diagnostic accuracy of GPT-4 using real-world clinical MRI reports of brain tumors. GPT-4 achieved final and differential diagnostic accuracy that is comparable with neuroradiologists. GPT-4 has the potential to improve the diagnostic process in clinical radiology.

Immune Checkpoint Inhibitor Is Associated with Improved Survival in Advanced Non-small Cell Lung Cancer Occurring in Patients with Autoimmune Disease.

The use of immune checkpoint inhibitors (ICIs) has revolutionized the treatment of advanced non-small cell lung cancer (NSCLC). However, clinical trials often exclude those with a history of autoimmune diseases (ADs) because of concerns regarding immune-related adverse events. Therefore, the efficacy of ICIs in advanced NSCLC patients with ADs should be evaluated. This study used administrative claims data from advanced treatment centers in Japan and identified patients with advanced NSCLC who began chemotherapy between December 2016 and January 2023. The patients were divided into four groups based on the presence of ADs and types of chemotherapy received. The association between ICI therapy and overall survival in the subgroups with or without ADs, and the association between the presence of AD and overall survival in patients who received ICI therapy and conventional chemotherapy, were analyzed using Cox proportional hazard regression, including therapy and presence of ADs and their interaction as covariates. These results were obtained using the inverse probability of treatment weighting. ICI therapy had a hazard ratio (95% confidence interval) for death in the subgroup of AD and non-AD patients of 0.88 (0.84-0.92) and 0.83 (0.71-0.97), respectively (p = 0.459 for interaction). For some specific ADs, including type 1 diabetes mellitus, the association between ICI therapy and decreased mortality was not observed. In conclusion, our study showed comparable associations between ICI therapy and reduced mortality in AD and non-AD subgroups of patients with advanced NSCLC. However, therapy strategies tailored to each AD type and thorough discussions regarding the risk-benefit profile are crucial.

Incidence and potential risk factors of human cytomegalovirus infection in patients with severe and critical coronavirus disease 2019.

BACKGROUND: Human cytomegalovirus (HCMV) infection occurs in immunosuppressed individuals and is known to increase mortality. Patients with coronavirus disease 2019 (COVID-19) are often treated with steroids, require intensive care unit (ICU) treatment, and may therefore be at risk for HCMV infection. However, which factors predispose severely ill patients with COVID-19 to HCMV infection and the prognostic value of such infections remain largely unexplored. This study aimed to examine the incidence and potential risk factors of HCMV infection in patients with severe or critical COVID-19 and evaluate the relationship between HCMV infection and mortality. METHODS AND FINDINGS: We used administrative claims data from advanced treatment hospitals in Japan to identify and analyze patients with severe or critical COVID-19. We explored potential risk factors for HCMV infection using multivariable regression models and its contribution to mortality in patients with COVID-19. Overall, 33,151 patients who progressed to severe or critical COVID-19 illness were identified. The incidence of HCMV infection was 0.3-1.7 % depending on the definition of HCMV infection. Steroids, immunosuppressants, ICU admission, and blood transfusion were strongly associated with HCMV infection. Furthermore, HCMV infection was associated with patient mortality independent of the observed risk factors for death. CONCLUSIONS: HCMV infection is a notable complication in patients with severe or critical COVID-19 who are admitted to the ICU or receive steroids, immunosuppressants, and blood transfusion and can significantly increase mortality risk.

Artificial intelligence model to identify elderly patients with locomotive syndrome: A cross-section study.

BACKGROUND: Identifying elderly individuals with locomotive syndrome is important to prevent disability in this population. Although screening tools for locomotive syndrome are available, these require time commitment and are limited by an individual's ability to complete questionnaires independently. To improve on this limitation, we developed a screening tool that uses information on the distribution of pressure on the plantar surface of the foot with an artificial intelligence (AI)-based decision system to identify patients with locomotor syndrome. Herein, we describe our AI-based system and evaluate its performance. METHODS: This was a cross-sectional study of 409 participants (mean age, 73.5 years). A foot scan pressure system was used to record the planter pressure distribution during gait. In the image processing step, we developed a convolutional neural network (CNN) to return the logit of the probability of locomotive syndrome based on foot pressure images. In the logistic regression step of the AI model, we estimated the predictor coefficients, including age, sex, height, weight, and the output of the CNN, based on foot pressure images. RESULTS: The AI model improved the identification of locomotive syndrome among elderly individuals compared to clinical data, with an area under curve of 0.84 (95% confidence interval, 0.79-0.88) for the AI model compared to 0.80 (95% confidence interval, 0.75-0.85) for the clinical model. Including the footprint force distribution image significantly improved the prediction algorithm (the net reclassification improvement was 0.675 [95% confidence interval, 0.45-0.90] P < 0.01; the integrated discrimination improvement was 0.059 [95% confidence interval, 0.039-0.088] P < 0.01). CONCLUSIONS: The AI system, which includes force distribution over the plantar surface of the foot during gait, is an effective tool to screen for locomotive syndrome.

Association between time-dependent changes in cerebrovascular autoregulation after cardiac arrest and outcomes: A prospective cohort study.

This prospective observational single-center cohort study aimed to determine an association between cerebrovascular autoregulation (CVAR) and outcomes in hypoxic-ischemic brain injury post-cardiac arrest (CA), and assessed 100 consecutive post-CA patients in Japan between June 2017 and May 2020 who experienced a return of spontaneous circulation. Continuous monitoring was performed for 96 h to determine CVAR presence. A moving Pearson correlation coefficient was calculated from the mean arterial pressure and cerebral regional oxygen saturation. The association between CVAR and outcomes was evaluated using the Cox proportional hazard model; non-CVAR time percent was the time-dependent, age-adjusted covariate. The non-linear effect of target temperature management (TTM) was assessed using a restricted cubic spline. Of the 100 participants, CVAR was detected using the cerebral performance category (CPC) in all patients with a good neurological outcome (CPC 1-2) and in 65 patients (88%) with a poor outcome (CPC 3-5). Survival probability decreased significantly with increasing non-CVAR time percent. The TTM versus the non-TTM group had a significantly lower probability of a poor neurological outcome at 6 months with a non-CVAR time of 18%-37% (p < 0.05). Longer non-CVAR time may be associated with significantly increased mortality in hypoxic-ischemic brain injury post-CA.