Objective: To evaluate the accuracy and reasoning capabilities of large multimodal language models compared with those of neuroradiology subspecialty-trained radiologists in neuroradiology case interpretation. Materials and Methods: This experimental study used custom-made 401 radiologic quizzes derived from articles published in RadioGraphics covering neuroradiology and head and neck topics (October 2020 to February 2024). We prompted the GPT-4 Turbo with Vision (GPT-4V), GPT-4 Omni, Gemini Flash, and Claude models to provide the top three differential diagnoses with a rationale and describe examination characteristics such as imaging modality, sequence, use of contrast, image plane, and body part. The temperature was adjusted to 0 and 1 (T1). Two neuroradiologists answered the same questions.The accuracies of the large language models (LLMs) and the neuroradiologists were compared using generalized estimating equations. Three neuroradiologists assessed the rationale provided by the LLMs for their differential diagnoses using four-point scales, separately for specific lesion locations and imaging findings, and evaluated the presence of hallucinations and the overall acceptability of the responses. Results: Top-3 accuracy (i.e., correct answers present among top-3 differential diagnoses) of LLMs ranged from 29.9% (120 of 401) to 49.4% (198 of 401, obtained with GPT-4V in the T1 setting), while radiologists achieved 80.3% (322 of 401) and 68.3% (274 of 401), respectively (P < 0.001). Regarding the rationale for differential diagnoses, GPT-4V (T1) accurately identified both the specific lesion location and imaging findings in 30.7% (123 of 401) and 12.9% (16 of 124) of cases without textual clinical history. Hallucinations occurred in 4.5% (18 of 401), and only 29.4% (118 of 401) of the LLM-generated analyses were deemed acceptable. GPT-4V (T1) demonstrated high accuracy in identifying the imaging modality (97.4% [800 of 821]) and scanned body parts (92.2% [756 of 820]). Conclusion LLMs remarkably underperformed compared with neuroradiologists and showed unsatisfactory reasoning for their differential diagnoses, with performance declining further in cases without textual input of clinical history. These findings highlight the limitations of current multimodal LLMs in neuroradiological interpretation and their reliance on text input.

Purpose: Postoperative complications following gastric cancer surgery significantly impact patient outcomes, yet standardized definitions for these events have not been consistently applied across institutions in Korea. This study aimed to develop a consensus-based, standardized complication classification system specific to gastrectomy for gastric cancer as part of the Korean Quality Improvement Platform in Surgery (K-QIPS) initiative. Methods: As part of K-QIPS, a dedicated task force team (TFT) was formed with surgical experts from fourteen high-volume hospitals across Korea. The TFT conducted ten formal meetings to review existing literature and international guidelines, and incorporated findings from randomized controlled trials. The final complication list was developed through expert consensus and structured into a standardized framework. A Data Entry Manual was created to support consistent data collection by surgical clinical reviewers. Results: The TFT defined specific postoperative complications following gastrectomy for gastric cancer, including anastomotic leakage, duodenal stump leakage, pancreatic fistula, intra-abdominal and luminal bleeding, delayed gastric emptying, and internal hernia. Notably, internal hernia was described in standardized form for the first time. General complications were developed first and overlapped in part with the gastric cancer-specific list. The task force also produced a Data Entry Manual that provides practical instructions to ensure consistency and accuracy in complication reporting. Conclusion This nationwide consensus initiative established the first standardized complication classification system for gastric cancer surgery in Korea. The proposed definitions and data entry system are expected to improve complication reporting, enable multicenter research, support surgical quality benchmarking, and ultimately enhance patient outcomes.

Purpose: The male predominance in the incidence of gastric cancer (GC) is established; however, sex differences in the prognosis of GC remain controversial. As such, this study analyzed the prognosis of patients with GC based on age and sex. Materials and Methods: Data from 14,739 patients diagnosed with GC at Seoul National University Bundang Hospital between 2003 and 2023 were analyzed. Baseline characteristics, histological types of GC, overall and GC-specific survival rates (age and stage stratification), and associated risk factors were analyzed. Results: Females were significantly younger (p < 0.001) and exhibited more gastric body cancers (p < 0.001) and tumors with diffuse-type or poorly differentiated histology (p < 0.001) than males. Females exhibited an advantage over males in terms of overall survival (p=0.004), but not in GC-specific survival. However, age stratification revealed significant sex differences, that females < 50 years of age exhibited survival disadvantages (p < 0.001); however, this trend was reversed with age, and females > 60 years exhibited survival advantages (p < 0.001) for both overall and GC-specific survival. This may be explained by the lower ratio of diffuse-type GC as females age. Furthermore, in the analysis according to stage, females with stage IV disease exhibited significant survival disadvantages, with significantly younger age and a higher proportion of diffuse-type GC which exhibits aggressive features, resulting in poorer survival than in males. Conclusion Age and stage stratification revealed significant differences in survival between the sexes, which can be helpful for public health strategies.

Purpose: This study evaluates the Korean Cancer Study Group Geriatric Score-7 (KG-7) frailty screening tool’s effectiveness in elderly multiple myeloma (MM) patients to prevent under and overtreatment. Materials and Methods: This prospective pilot cohort study included 100 elderly patients aged 70 and older with newly diagnosed MM who had not undergone transplantation from August 2020 to January 2022. Results: The median age was 77 years, and 73.0% of patients were classified at International Staging System stages 2 or 3. Using a 5-point cutoff on the KG-7 index (non-frail, score ≥ 5; frail, score < 5), 31% were categorized as frail. After a median follow-up of 26.8 months, the 3-year overall survival rate was 73.0%. There was no statistically significant association between any frailty index and the risk of death. However, frail patients defined by the simplified frailty index (hazard ratio [HR], 2.49; 95% confidence interval [CI], 1.09 to 5.95; p=0.030) and by KG-7 (HR, 2.43; 95% CI, 1.03 to 5.86; p=0.043) had a significantly higher risk of grade 3-4 non-hematologic toxicity, whereas the International Myeloma Working Group definition did not. Over a 24-month tracking period, vulnerability as measured by KG-7 either improved or deteriorated. Conclusion The pilot study, which had a limited number of participants, did not demonstrate KG-7’s effectiveness in predicting survival; however, it successfully predicted severe non-hematologic toxicities. We plan to conduct larger studies in elderly MM patients to determine whether KG-7 can help tailor their treatment regimens.

Objective: To classify life stages for companion dogs and cats by identifying clusters in age-specific disease proportions derived from medical records, providing a data-driven foundation for health examination programs. Methods: We collected 505,667 medical records from 82 veterinary facilities in South Korea between 2020 and 2023. Diagnoses were standardized using GPT-4o and S-BioBERT. Following preprocessing, data from 27 facilities yielded 222,706 canine and 39,910 feline records for the final analysis. Principal component analysis and K-means clustering (K = 4) were applied to age-specific disease proportions to identify life stages.The 10 most highest-proportion diagnoses diseases were determined for each cluster. Results: Canine life stages were classified as ≤ 1 year, 2–5 years, 6–10 years, and 11–15+ years.Feline life stages were 1–2 years, 3–8 years, 9–12 years, and 13–15+ years. In dogs, developmental diseases were common in the youngest age group, while chronic diseases were more prevalent in older groups. In cats, oral and urinary diseases were high-ranking, conjunctivitis was most common in the early stage, and chronic diseases increased with age. Conclusions and Relevance: Age-specific diagnosis patterns support four practical life stages for dogs and cats in South Korea. These boundaries can inform evidence-based preventive examination schedules, animal health policy, and pet insurance product design.

Background/Aims: The incidence of steatotic liver disease (SLD) is increasing across all age groups as the incidence of obesity increases worldwide. The existing noninvasive prediction models for SLD require laboratory tests or imaging and perform poorly in the early diagnosis of infrequently screened populations such as young adults and individuals with healthcare disparities. We developed a machine learning-based point-of-care prediction model for SLD that is readily available to the broader population with the aim of facilitating early detection and timely intervention and ultimately reducing the burden of SLD. Methods: We retrospectively analyzed the clinical data of 28,506 adults who had routine health check-ups in South Korea from January to December 2022. A total of 229,162 individuals were included in the external validation study. Data were analyzed and predictions were made using a logistic regression model with machine learning algorithms. Results: A total of 20,094 individuals were categorized into SLD and non-SLD groups on the basis of the presence of fatty liver disease. We developed three prediction models: SLD model 1, which included age and body mass index (BMI); SLD model 2, which included BMI and body fat per muscle mass; and SLD model 3, which included BMI and visceral fat per muscle mass. In the derivation cohort, the area under the receiver operating characteristic curve (AUROC) was 0.817 for model 1, 0.821 for model 2, and 0.820 for model 3. In the internal validation cohort, 86.9% of individuals were correctly classified by the SLD models. The external validation study revealed an AUROC above 0.84 for all the models. Conclusions As our three novel SLD prediction models are cost-effective, noninvasive, and accessible, they could serve as validated clinical tools for mass screening of SLD.

To investigate the relationship between depression and AD, water avoidance stress (WAS) was induced for 10 days in both Tg2576 mice and wild-type (WT) mice. After WAS, memory function and depressive-like behavior were investigated in Tg2576 mice. Tg2576 WAS mice exhibited more depressive-like behaviors than WT WAS and Tg2576 control (CON) mice. Strikingly, Tg2576 CON mice showed more depressive-like behaviors than WT mice. Moreover, corticosterone and phospho-glucocorticoid receptor (p-GR) levels were also higher in Tg2576 WAS mice in comparison to Tg2576 CON mice. Spinster homologue 2 (SPNS2) is a member of non-ATP-dependent transporter. The role of SPNS2 was widely known as a sphingosine-1-phosphate (S1P) transporter, which export intracellular S1P from cells. Using GEO database to analyze SPNS2 gene expression changes in patients with AD and depression, we show that SPNS2 gene expression correlates with AD and depression. Interestingly, Tg2576 WAS mice displayed significantly increased levels of SPNS2 w1hen compared to Tg2576 CON counterparts. SPNS2 levels were also higher in Tg2576 CON mice in comparison with WT CON mice. Remarkably, we found a decrease in S1P brain levels and an increase in S1P serum levels of Tg2576 WAS mice in comparison with Tg2576 CON mice. Accordingly, WAS induced group further decreased S1P levels in the brains. However, the level in the serum further increased in comparison with non-induced group. Therefore, these results suggest that AD and depression could be associated, and that Tg2576 transgenic mice are more susceptible to stress-induced depression through the release of S1P by SPNS2 up-regulation.

Background/Aims: Elevated troponin levels predict in-hospital mortality and influence decisions regarding thrombolytic therapy in patients with acute pulmonary embolism (PE). However, the usefulness of high-sensitivity troponin T (hsTnT) regarding PE remains uncertain. We aimed to establish the optimal cut-off level and compare its performance for precise risk stratification. Methods: 374 patients diagnosed with acute PE were reviewed. PE-related adverse outcomes, a composite of PE-related deaths, cardiopulmonary resuscitation incidents, systolic blood pressure < 90 mmHg, and all-cause mortality within 30 days were evaluated. The optimal hsTnT cut-off for all-cause mortality, and the net reclassification index (NRI) was used to assess the incremental value in risk stratification. Results: Among 343 normotensive patients, 17 (5.0%) experienced all-cause mortality, while 40 (10.7%) had PE-related adverse outcomes. An optimal hsTnT cut-off value of 60 ng/L for all-cause mortality (AUC 0.74, 95% CI 0.61–0.85, p < 0.001) was identified, which was significantly associated with PE-related adverse outcomes (OR 4.07, 95% CI 2.06–8.06, p < 0.001). Patients with hsTnT ≥ 60 ng/L were older, hypotensive, had higher creatinine levels, and right ventricular dysfunction signs. Combining hsTnT ≥ 60 ng/L with simplified pulmonary embolism severity index ≥1 provided additional prognostic information. Reclassification analysis showed a significant shift in risk categories, with an NRI of 1.016 ± 0.201 (p < 0.001). Conclusions We refined troponin’s predictive value in patients with acute PE, proposing a new cut-off value of hsTnT ≥ 60 ng/L. Validation through large-scale studies is essential to offer clinically useful guidance for managing patient population.

Background/Aims: Elevated troponin levels predict in-hospital mortality and influence decisions regarding thrombolytic therapy in patients with acute pulmonary embolism (PE). However, the usefulness of high-sensitivity troponin T (hsTnT) regarding PE remains uncertain. We aimed to establish the optimal cut-off level and compare its performance for precise risk stratification. Methods: 374 patients diagnosed with acute PE were reviewed. PE-related adverse outcomes, a composite of PE-related deaths, cardiopulmonary resuscitation incidents, systolic blood pressure < 90 mmHg, and all-cause mortality within 30 days were evaluated. The optimal hsTnT cut-off for all-cause mortality, and the net reclassification index (NRI) was used to assess the incremental value in risk stratification. Results: Among 343 normotensive patients, 17 (5.0%) experienced all-cause mortality, while 40 (10.7%) had PE-related adverse outcomes. An optimal hsTnT cut-off value of 60 ng/L for all-cause mortality (AUC 0.74, 95% CI 0.61–0.85, p < 0.001) was identified, which was significantly associated with PE-related adverse outcomes (OR 4.07, 95% CI 2.06–8.06, p < 0.001). Patients with hsTnT ≥ 60 ng/L were older, hypotensive, had higher creatinine levels, and right ventricular dysfunction signs. Combining hsTnT ≥ 60 ng/L with simplified pulmonary embolism severity index ≥1 provided additional prognostic information. Reclassification analysis showed a significant shift in risk categories, with an NRI of 1.016 ± 0.201 (p < 0.001). Conclusions We refined troponin’s predictive value in patients with acute PE, proposing a new cut-off value of hsTnT ≥ 60 ng/L. Validation through large-scale studies is essential to offer clinically useful guidance for managing patient population.