Takayuki Takahashi

Background: Large language models (LLMs) are rapidly emerging in health care, offering opportunities in decision support, education, and research, but raising critical concerns about safety, reliability, and ethics. Although several guidelines for trustworthy AI exist in business and technology, few systematic reviews have applied them to medical contexts. Objective: This study aimed to conduct a systematic review of LLM research in health care, applying the AI Guidelines for Business as a framework across 11 domains, including safety, reliability, ethics, transparency, fairness, inclusiveness, privacy, security, robustness, data quality, and verifiability. Methods: Following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines (retrospectively registered on the Open Science Framework; DOI 10.17605/OSF.IO/P4KSB), the PubMed, Scopus, Web of Science, arXiv, and IEEE Xplore databases were searched on January 15, 2025. Records were screened in 2 stages by 3 reviewers (with records retained only upon unanimous agreement). A total of 247 studies were included, of which 211 (85.4%) contributed quantitative values. Eligible studies were classified across 11 trustworthy AI domains. Heterogeneous metrics were summarized within metric families; when multiple models were evaluated, the mean across models was used as the primary estimate, with best, median, and primary-model sensitivity analyses. The LLM-assisted categorization (GPT-5 mini) was validated by using an automated internal consistency check, and 95% CIs were estimated by using cluster bootstrap on study-level values. Results: Of the 25,156 records, 247 (1.0%) studies were included, and of these, 211 (85.4%) contributed quantitative values. Evaluation concentrated on accuracy (143/247, 57.9%) and fairness and inclusiveness (93/247, 37.7%), followed by data quality (47/247, 19.0%) and prevention of misinformation (44/247, 17.8%). Normalized performance was moderate to high (accuracy mean 0.73, 95% CI 0.7-0.76; data quality: 0.64; prevention of misinformation: 0.82). Selecting the best-performing model inflated domain means by up to 0.05. Privacy protection (2/247, 0.8%) and security assurance (0/247, 0.0%) were almost entirely absent. Domain assignments were recoverable from objective metric types in 98.9% of values (Cohen κ=0.985). Conclusions: Current evaluations emphasize accuracy while underreporting privacy, security, robustness, explainability, and verifiability. This finding reflects gaps in reporting rather than demonstrated poor performance, underscoring the need for comprehensive, guideline-based, multidomain evaluation before deployment in high-stakes clinical settings.

Objective: Chylous ascites (CA) is a rare yet clinically significant complication following gynecologic cancer surgery, with incidence rates of 0.17 % to 9%. We aimed to describe a case of CA with a delayed clinical presentation nearly 100 days postoperatively in a patient with advanced endometrial cancer and to review the management strategies. Methods: We retrospectively evaluated a 75‐year‐old patient who underwent radical hysterectomy, bilateral salpingo‐oophorectomy, extended lymphadenectomy (pelvic and para‐aortic), and partial omentectomy for stage IIIB endometrial cancer. Data collected included onset timing, ascitic fluid analysis, imaging findings, and treatment responses. Additionally, a narrative review identified 13 relevant studies discussing the onset, risk factors, diagnosis, and therapies for post‐operative CA in gynecologic oncology. Results: Although CA typically appears within 4 to 21 days, our patient developed CA at approximately post‐operative day 99. Diagnostic paracentesis confirmed triglyceride‐rich ascitic fluid, establishing the diagnosis of CA. Dietary modification (fasting followed by medium‐chain triglyceride diet), octreotide therapy, and computed tomography (CT)‐guided lymphangiography effectively controlled the chylous leakage without requiring surgery. Conservative measures—low‐fat or medium‐chain triglyceride diets, total parenteral nutrition, and somatostatin analogs—are considered first‐line, while lymphangiography/embolization and eventual surgical ligation may be needed for refractory cases. Conclusions: This case illustrates that CA with a delayed clinical presentation can be successfully treated with a stepwise conservative approach comprising dietary therapy, octreotide, and CT‐guided lymphangiography, even when presenting more than 3 months postoperatively. Moreover, our patient remained free of disease recurrence at 1 year and 8 months postoperatively, underscoring that timely management of CA can avoid delays in adjuvant therapy.

We investigated the clinical significance of positive surgical margins (PSMs) in index tumors following radical prostatectomy (RP), with particular attention to the tumor’s zonal origin. Among 1148 patients with localized prostate cancer who underwent RPs, 973 were included after excluding those who received perioperative therapy or had incomplete data. Index tumors were categorized by zonal origin: transition zone, peripheral zone, or central zone (CZ). Overall, PSMs were observed in 26.4% of index tumors. Although CZ index tumors were relatively uncommon (6.5%), they exhibited the highest PSM rate (42.9%) and showed the most aggressive pathological features. The 5-year biochemical recurrence (BCR)-free survival rate was significantly lower in patients with PSMs in index tumors than in those with negative surgical margins (45.6% vs. 86.8%, p < 0.0001). Notably, patients with PSMs in CZ index tumors had the worst outcomes, with a 5-year BCR-free survival rate of 22.0%. Multivariate analysis identified PSMs in index tumors as an independent predictor of BCR (HR: 3.4; 95% CI: 2.5–4.5), with a similar trend observed in early recurrence. These findings highlight the prognostic significance of PSMs in index tumors during RP, especially in CZ tumors, and emphasize the importance of securing local control in these cases.

Human papillomavirus (HPV) infection remains a principal cause of cervical cancer worldwide. Although large-scale vaccination efforts have substantially lowered HPV infection rates and precancerous lesions, not all regions have achieved high coverage. In Japan, proactive HPV vaccine recommendations were suspended from 2013 to 2022 due to concerns over alleged adverse events, causing vaccination rates to drop from over 70% to below 1%. This narrative review synthesized research published from 2014 to 2025 in PubMed, Cochrane Library, and Google Scholar, focusing on English-language studies. Inclusion criteria encompassed analyses of HPV vaccine efficacy or safety, policies related to vaccination in Japan or other countries, and investigations into vaccine hesitancy or media influences. Data were categorized into five thematic areas: historical and policy contexts, evidence of vaccine safety and efficacy, societal drivers of hesitancy, communication strategies, and administrative or clinical interventions. Evidence robustly confirms the HPV vaccine’s favorable safety profile, with severe adverse events appearing exceedingly rare. Nonetheless, media sensationalism and limited risk communication in Japan perpetuated mistrust, impeding vaccination uptake. Comparisons with Denmark and Ireland indicate that transparent, interactive risk communication can restore coverage to near-pre-suspension levels. Japan’s recent policy reforms, including reinstating proactive recommendations and catch-up initiatives, have begun to reverse vaccination hesitancy. Sustained policy support, evidence-based messaging, and empathetic engagement with communities are central to rebuilding trust in the HPV vaccine. Lessons from best international practices emphasize the importance of multifaceted interventions, collaborative stakeholder engagement, and transparent risk communication to reduce the burden of HPV-related malignancies.

Background: Artificial intelligence (AI) is increasingly applied to colposcopy to enhance the detection of cervical intraepithelial neoplasia (CIN) and cervical cancer. We conducted a systematic review to summarize the diagnostic performance achieved by AI‑based colposcopic systems. Methods: Following the PRISMA 2020 guidelines, the PubMed database was searched using the search terms ‘artificial intelligence’ and ‘colposcop*’ for articles published between 2019 and 2024. From the initial 43 articles retrieved, 19 studies were selected based on specific inclusion criteria: original research articles, written in the English language, and relevant to CIN or cervical cancer diagnosis. For each, we extracted the sample size, AI architecture (e.g., convolutional neural networks, U-Net/DeepLab V3 + segmentation models, multimodal fusion networks), reference standard, and reported metrics (sensitivity, specificity, accuracy, and area under the curve). Results: Across multiple studies, AI systems demonstrated superior diagnostic accuracy, sensitivity, and specificity, particularly for early detection of high-risk lesions and classification of cervical abnormalities. Deep-learning models, such as convolutional neural networks, consistently outperformed conventional methods by reducing diagnostic variability and offering robust performance even in low-resource settings. The review also highlights the potential of AI for real-time diagnostics and its capacity to support clinical decision-making via automated systems. Conclusion: AI has the potential to revolutionize cervical cancer diagnosis and management by enhancing the accuracy and efficiency of colposcopic evaluations. However, challenges remain, including the development of standardized datasets, validation in diverse populations, and ethical considerations surrounding data privacy and access to technology. Continued research and development are crucial to harness AI’s global potential to improve patient outcomes.

Prostate cancer (PCa) is a clinically heterogeneous disease. Predicting clinically significant PCa with low–intermediate prostate-specific antigen (PSA), which often includes aggressive cancers, is imperative. This study evaluated the predictive accuracy of deep learning analysis using multimodal medical data focused on clinically significant PCa in patients with PSA ≤ 20 ng/mL. Our cohort study included 178 consecutive patients who underwent ultrasound-guided prostate biopsy. Deep learning analyses were applied to predict clinically significant PCa. We generated receiver operating characteristic curves and calculated the corresponding area under the curve (AUC) to assess the prediction. The AUC of the integrated medical data using our multimodal deep learning approach was 0.878 (95% confidence interval [CI]: 0.772–0.984) in all patients without PSA restriction. Despite the reduced predictive ability of PSA when restricted to PSA ≤ 20 ng/mL (n = 122), the AUC was 0.862 (95% CI: 0.723–1.000), complemented by imaging data. In addition, we assessed clinical presentations and images belonging to representative false-negative and false-positive cases. Our multimodal deep learning approach assists physicians in determining treatment strategies by predicting clinically significant PCa in patients with PSA ≤ 20 ng/mL before biopsy, contributing to personalized medical workflows for PCa management.

Takahashi T, et al. J Gynecol Oncol. 2022 May;33:e57. https://doi.org/10.3802/jgo.2022.33.e57