Several hepatic steatosis formulae have been validated in various cohorts using ultrasonography. However, none of these studies has been validated in a community-based setting using the gold standard method. Thus, the aim of this study was to externally validate hepatic steatosis formulae in community-based settings using magnetic resonance imaging (MRI). A total of 1301 community-based health checkup subjects who underwent liver fat quantification with MRI were enrolled in this study. Diagnostic performance was assessed using the area under the receiver operating characteristic curve (AUROC). Non-alcoholic fatty liver disease (NAFLD) liver fat score showed the highest diagnostic performance with an AUROC of 0.72, followed by Framingham steatosis index (0.70), hepatic steatosis index (HSI, 0.69), ZJU index (0.69), and fatty liver index (FLI, 0.68). There were considerable gray zones in three fatty liver prediction models using two cutoffs (FLI, 28.9%; HSI, 48.9%; and ZJU index, 53.6%). The diagnostic performance of NAFLD liver fat score for detecting steatosis was comparable to that of ultrasonography. The diagnostic agreement was 72.7% between NAFLD liver fat score and 70.9% between ultrasound and MRI. In conclusion, the NAFLD liver fat score showed the best diagnostic performance for detecting hepatic steatosis. Its diagnostic performance was comparable to that of ultrasonography in a community-based setting.

Background/Objectives: This study investigated the diagnostic capabilities of two AI-based tools, M4CXR (research-only version) and ChatGPT-4o, in chest X-ray interpretation. M4CXR is a specialized cloud-based system using advanced large language models (LLMs) for generating comprehensive radiology reports, while ChatGPT, built on the GPT-4 architecture, offers potential in settings with limited radiological expertise. Methods: This study evaluated 826 anonymized chest X-ray images from Inha University Hospital. Two experienced radiologists independently assessed the performance of M4CXR and ChatGPT across multiple diagnostic parameters. The evaluation focused on diagnostic accuracy, false findings, location accuracy, count accuracy, and the presence of hallucinations. Interobserver agreement was quantified using Cohen’s kappa coefficient. Results: M4CXR consistently demonstrated superior performance compared to ChatGPT across all evaluation metrics. For diagnostic accuracy, M4CXR achieved approximately 60–62% acceptability ratings compared to ChatGPT’s 42–45%. Both systems showed high interobserver agreement rates, with M4CXR generally displaying stronger consistency. Notably, M4CXR showed better performance in anatomical localization (76–77.5% accuracy) compared to ChatGPT (36–36.5%) and demonstrated fewer instances of hallucination. Conclusions: The findings highlight the complementary potential of these AI technologies in medical diagnostics. While M4CXR shows stronger performance in specialized radiological analysis, the integration of both systems could potentially optimize diagnostic workflows. This study emphasizes the role of AI in augmenting human expertise rather than replacing it, suggesting that a combined approach leveraging both AI capabilities and clinical judgment could enhance patient care outcomes.

Background: Knee osteoarthritis (OA) is a prevalent degenerative joint disease significantly impacting global health. Early and accurate diagnosis is crucial for effective management, but traditional methods often rely on subjective assessments. This study evaluates the efficacy of a deep learning model implemented through a no-code AI platform for diagnosing and grading knee OA from plain radiographs. Methods: We utilized the Osteoarthritis Initiative (OAI) dataset, comprising knee X-ray data from 1526 patients. The data were split into training (47.0%), validation (26.5%), and test (26.5%) sets. We employed a ResNet101 model on the DEEP:PHI no-code AI platform for image analysis. The model was trained to classify knee OA into five grades (0–4) based on the Kellgren–Lawrence scale. Results: Our AI model demonstrated high accuracy in distinguishing between different OA grades, with particular strength in early-stage detection. The model achieved optimal performance at 20 epochs, suggesting efficient learning dynamics. Grad-CAM visualizations were used to enhance the interpretability of the model’s decision-making process. Conclusions: This study demonstrates the potential of AI, implemented through a no-code platform, to accurately diagnose and grade knee OA from radiographs. The use of a no-code AI platform such as DEEP:PHI represents a step towards democratizing AI in healthcare, enabling the rapid development and deployment of sophisticated medical AI applications without extensive coding expertise. This approach could significantly enhance the early detection and management of knee OA, potentially improving patient outcomes and streamlining clinical workflows.

Objectives: To assess whether virtual non-contrast (VNC) computed tomography (CT) images acquired from dual-energy CT (DECT) have sufficient image quality to replace true non-contrast (TNC) CT images in CT angiography (CTAG). Methods: This study enrolled 63 consecutive patients who underwent a CTAG that included a single-energy non-contrast scan, followed by a post-contrast DECT scan. Comparison of attenuation and noise between TNC and VNC images was made by drawing circular regions of interest (ROI) on a picture archiving and communication system (PACS) workstation within the aortic lumen at the levels of the renal arteries, the aortic bifurcation and right femoral artery. Mean attenuation and image noise (one standard deviation) were registered in Hounsfield units (HU). The VNC images were subjectively evaluated for artifacts such as subtraction of calcifications or architectural distortion based on TNC image as a standard of reference. Results: Most attenuations of the VNCs were higher than TNC, except right femoral artery of reader 2. Most image noises of TNC were higher than VNC, except abdominal aorta in reader 1. In qualitative image analysis, mean scores of VNC according to the 5-point scale were 3.68 and 3.63 (reader 1 and reader 2, respectively) which mean good to excellent to diagnose. Conclusion: HUs and VNC image noises are different from TNC images in CTAG. VNC images have sufficient image quality to replace TNC images in the diagnosis of calcific lesions. Advances in knowledge: VNC images acquired from DECT have image quality adequate to replace TNC images in the diagnosis of the calcific lesion on the CTAG. VNC images based on DECT can avoid excessive and unnecessary patient exposure to radiation during CTAG.

Background: This study aimed to compare the efficacy and safety of 70–150 μm doxorubicin drug-eluting bead (DEB) transarterial chemoembolization (TACE) with those of 100–300 μm DEB-TACE as first-line treatment in patients with hepatocellular carcinoma (HCC). Methods: We retrospectively investigated 72 patients who underwent TACE with 70–150 μm DEBs (n = 40) or 100–300 μm DEBs (n = 32) for HCC in a tertiary center between March 2013 and May 2019. Initial treatment response and adverse events were assessed using the modified Response Evaluation Criteria in Solid Tumors and the National Cancer Institute Common Terminology Criteria for Adverse Events version 5.0, respectively. Results: At the 2-month post-treatment assessment, the complete and objective response rates were 47.5% and 85.0%, respectively, for the 70–150 μm group and 34.4% and 81.3%, respectively, for the 100–300 μm group; however, the difference was not statistically significant (p > 0.05). In total, 65% patients in the 70–150 μm group and 59.4 % patients in the 100-300 μm group experienced at least one symptom of post-embolization syndrome after TACE; all symptoms were classified as grade 1 or 2. There was no significant difference between the two groups in terms of post-procedural laboratory changes such as changes in liver enzymes and bilirubin levels (p > 0.05). Laboratory toxicity of grade 3 occurred in three patients, all of which were transient elevation of liver enzyme levels. Hepatobiliary adverse events, such as bile duct injury, biloma, liver abscess, and hepatic infarction, were not observed in either treatment group. Conclusion: This study found no significant difference in tumor response between 70–150 μm and 100–300 μm DEB-TACE. Both groups showed favorable safety profiles, and the difference was not significant.