Introduction: Artificial intelligence (AI) is revolutionizing medical workflows, with self-learning systems like ChatGPT showing promise in therapy recommendations. Our study evaluated ChatGPT’s performance in suggesting treatments for 30 breast cancer cases. AI’s role in healthcare is expanding, particularly with tools like ChatGPT becoming accessible. However, understanding its limitations is vital for safe implementation. Material and Methods: We used 30 breast cancer cases from our medical board, assessing ChatGPT’s suggestions. The input was standardized, incorporating relevant patient details and treatment options. ChatGPT’s output was evaluated by oncologists based on a given questionnaire. Results: Treatment recommendations by ChatGPT were overall rated sufficient with minor limitations by the oncologists. The HER2 treatment category was the best-rated therapy option, with the most accurate recommendations. Primary cases received more accurate recommendations, especially regarding chemotherapy. Conclusions: While ChatGPT demonstrated potential, difficulties were shown in intricate cases and postoperative scenarios. Challenges arose in offering chronological treatment sequences and partially lacked precision. Refining inputs, addressing ethical intricacies, and ensuring chronological treatment suggestions are essential. Ongoing research is vital to improving AI’s accuracy, balancing AI-driven suggestions with expert insights and ensuring safe and reliable AI integration into patient care.

Primary or secondary immunodeficiencies are characterized by disruption of cellular and humoral immunity. Respiratory infections are a major cause of morbidity and mortality among immunodeficient or immunocompromised patients, with Staphylococcus aureus being a common offending organism. We propose here an adoptive macrophage transfer approach aiming to enhance impaired pulmonary immunity against S aureus. Our studies, using human-induced pluripotent stem cell-derived macrophages (iMφs), demonstrate efficient antimicrobial potential against methicillin-sensitive and methicillin-resistant clinical isolates of S aureus. Using an S aureus airway infection model in immunodeficient mice, we demonstrate that the adoptive transfer of iMφs is able to reduce the bacterial load more than 10-fold within 20 hours. This effect was associated with reduced granulocyte infiltration and less damage in lung tissue of transplanted animals. Whole transcriptome analysis of iMφs compared with monocyte-derived macrophages indicates a more profound upregulation of inflammatory genes early after infection and faster normalization 24 hours postinfection. Our data demonstrate high therapeutic efficacy of iMφ-based immunotherapy against S aureus infections and offer an alternative treatment strategy for immunodeficient or immunocompromised patients.

Purpose: This prospective multicenter study aimed to compare the decision-making abilities of board-certified colposcopists and two commercially available large language models (LLM), ChatGPT-4o and ChatGPT-5, in cervical dysplasia management. Methods: Twenty-three anonymized real-life patient cases with multiple-choice (MC) questions regarding treatment decisions were used to assess answer quality. Ten board-certified colposcopists and the two LLMs addressed the MC questions. The gold standard was defined by two guideline authors. LLMs were prompted to justify their responses. Concordance rates were calculated and compared across all questions and histopathological subgroups, including cervical intraepithelial neoplasia (CIN), unspecific histopathological results, and cervical cancer cases. Results: Clinicians and LLMs achieved similar overall concordance rates compared to the gold standard (69.6% for clinicians, 69.6% for ChatGPT-4o, and 65.2% for ChatGPT-5). ChatGPT-5 outperformed clinicians in precancerous lesions (81.8% vs. 66.4%), while clinicians excelled in complex cases with unspecific histopathology (86% vs. 60%). Clinicians showed a tendency to overtreat low-grade lesions (CIN I), opting for more intensive surveillance. ChatGPT-4o performed better than ChatGPT-5 in cervical cancer cases, though both models struggled with these scenarios. Conclusion: This study highlights the potential of LLMs as decision support tools in cervical dysplasia management, particularly for straightforward cases like precancerous lesions. However, clinicians remain superior in handling complex or ambiguous cases. The tendency of clinicians to overtreat low-grade lesions may offer the potential to test the implementation of a decision support tool for those cases. While LLMs show promise, exploring open-ended clinical scenarios and integrating retrieval-augmented generation could enhance their practical application.

Introduction: Medical decision-making is characterized by rapidly evolving evidence being represented by national and international guidelines that form the decision basis for treatment recommendations. Large language model-based clinical decision support systems (LLM-CDSS) have shown promising potential to serve as a support tool, e.g., in gynecologic oncology. Nevertheless, physicians’ acceptance of this technology in gynecologic oncology is currently not backed by evidence. Therefore, we present a pilot study evaluating the clinical technology acceptance of LLM-CDSS in gynecologic oncology. Methods: We designed a questionnaire based on the existing technology acceptance model (TAM) and the unified theory of acceptance and use of technology (UTAUT) characteristics, which was sent to clinicians employed at a clinic of gynecology and obstetrics of a large tertiary hospital. A total of 29 physicians spanning the complete range of clinical experience answered the survey. For identifying factors influencing the willingness to use a LLM-CDSS, Ordinary Least Squares (OLS) regression was applied based on answers grouped by question categories. Further, descriptive analysis was performed by measures of central tendency and correlation. Results: OLS regression did not identify any question categories significantly influencing willingness to use LLM-CDSS. Answer distribution was left skewed for the primary endpoint, indicating an acquiescence bias in the survey population. Still, descriptive findings indicate a generally positive but cautious attitude toward LLM-based therapy recommendation systems in gynecologic oncology. Acceptance appeared to depend strongly on transparency, integration of evidence, clinical validation, and the preservation of physician oversight and accountability. Discussion: The participants agreed on a possible positive influence of LLM-CDSS for clinical practice. However, the results emphasize that such tools should support rather than replace physicians to leave the final medical decision to the human. Participants were divided on whether patients must be informed about the usage of such tools whereas some participants reported concerns about their clinical autonomy and patients’ trust. The survey needs to be validated in a larger multicentered cohort in a shortened version.