Leibovich, Bradley C.

Medical researchers are increasingly utilizing advanced LLMs like ChatGPT-4 and Gemini to enhance diagnostic processes in the medical field. This research focuses on their ability to comprehend and apply complex medical classification systems for breast conditions, which can significantly aid plastic surgeons in making informed decisions for diagnosis and treatment, ultimately leading to improved patient outcomes. Fifty clinical scenarios were created to evaluate the classification accuracy of each LLM across five established breast-related classification systems. Scores from 0 to 2 were assigned to LLM responses to denote incorrect, partially correct, or completely correct classifications. Descriptive statistics were employed to compare the performances of ChatGPT-4 and Gemini. Gemini exhibited superior overall performance, achieving 98% accuracy compared to ChatGPT-4’s 71%. While both models performed well in the Baker classification for capsular contracture and UTSW classification for gynecomastia, Gemini consistently outperformed ChatGPT-4 in other systems, such as the Fischer Grade Classification for gender-affirming mastectomy, Kajava Classification for ectopic breast tissue, and Regnault Classification for breast ptosis. With further development, integrating LLMs into plastic surgery practice will likely enhance diagnostic support and decision making.

Since their release, the medical community has been actively exploring large language models’ (LLMs) capabilities, which show promise in providing accurate medical knowledge. One potential application is as a patient resource. This study analyzes and compares the ability of the currently available LLMs, ChatGPT-3.5, GPT-4, and Gemini, to provide postoperative care recommendations to plastic surgery patients. We presented each model with 32 questions addressing common patient concerns after surgical cosmetic procedures and evaluated the medical accuracy, readability, understandability, and actionability of the models’ responses. The three LLMs provided equally accurate information, with GPT-3.5 averaging the highest on the Likert scale (LS) (4.18 ± 0.93) (p = 0.849), while Gemini provided significantly more readable (p = 0.001) and understandable responses (p = 0.014; p = 0.001). There was no difference in the actionability of the models’ responses (p = 0.830). Although LLMs have shown their potential as adjunctive tools in postoperative patient care, further refinement and research are imperative to enable their evolution into comprehensive standalone resources.

In postoperative care, patient education and follow-up are pivotal for enhancing the quality of care and satisfaction. Artificial intelligence virtual assistants (AIVA) and large language models (LLMs) like Google BARD and ChatGPT-4 offer avenues for addressing patient queries using natural language processing (NLP) techniques. However, the accuracy and appropriateness of the information vary across these platforms, necessitating a comparative study to evaluate their efficacy in this domain. We conducted a study comparing AIVA (using Google Dialogflow) with ChatGPT-4 and Google BARD, assessing the accuracy, knowledge gap, and response appropriateness. AIVA demonstrated superior performance, with significantly higher accuracy (mean: 0.9) and lower knowledge gap (mean: 0.1) compared to BARD and ChatGPT-4. Additionally, AIVA’s responses received higher Likert scores for appropriateness. Our findings suggest that specialized AI tools like AIVA are more effective in delivering precise and contextually relevant information for postoperative care compared to general-purpose LLMs. While ChatGPT-4 shows promise, its performance varies, particularly in verbal interactions. This underscores the importance of tailored AI solutions in healthcare, where accuracy and clarity are paramount. Our study highlights the necessity for further research and the development of customized AI solutions to address specific medical contexts and improve patient outcomes.

With abundant information and interconnectedness among people, identifying knowledgeable individuals in specific domains has become crucial for organizations. Artificial intelligence (AI) algorithms have been employed to evaluate the knowledge and locate experts in specific areas, alleviating the manual burden of expert profiling and identification. However, there is a limited body of research exploring the application of AI algorithms for expert finding in the medical and biomedical fields. This study aims to conduct a scoping review of existing literature on utilizing AI algorithms for expert identification in medical domains. We systematically searched five platforms using a customized search string, and 21 studies were identified through other sources. The search spanned studies up to 2023, and study eligibility and selection adhered to the PRISMA 2020 statement. A total of 571 studies were assessed from the search. Out of these, we included six studies conducted between 2014 and 2020 that met our review criteria. Four studies used a machine learning algorithm as their model, while two utilized natural language processing. One study combined both approaches. All six studies demonstrated significant success in expert retrieval compared to baseline algorithms, as measured by various scoring metrics. AI enhances expert finding accuracy and effectiveness. However, more work is needed in intelligent medical expert retrieval.

458 Background: Immune checkpoint inhibitors (ICI) targeting PD-1/PD-L1 axis has revolutionized the therapeutic landscape of patients with advanced and locally advanced renal cell carcinoma (RCC). However, complete response rates remain low and there is a need for novel therapeutic options for ICI-resistant and refractory RCC. Our group has generated a novel humanized anti-PD-L1 antibody, called H1A, that induces PD-L1 degradation preventing not only its interaction with PD-1 but also inhibiting pro-tumorigenic intracellular signaling. In this study, we compared the antitumor activity of H1A with FDA-approved ICI using patient-derived immune cells and humanized mouse tumor models. Methods: In vivo antitumor activity of H1A and atezolizumab (ATZ) was compared in three humanized PD-1/PD-L1 mouse tumor models with different immunogenicity (E0771:high, MC38:moderate, B16-F10:poor). Peripheral blood mononuclear cells (PBMC) were isolated from 14 RCC patients undergoing with nephrectomy. PBMC were treated with H1A, nivolumab (NIV), atezolizumab (ATZ) and pembrolizumab (PEM) to evaluate their efficacy in inducing tumor cell death in an ex-vivo cytotoxicity assay. Mass cytometry was employed to determine the impact of H1A and atezolizumab on the PBMC profile. Results: MC38 tumors showed moderate response to both H1A and ATZ which was abrogated by CD8 T cell depletion. In E0771 tumors, higher percentage of complete responders was observed with H1A compared to ATZ and E0771 rechallenge revealed memory antitumor immunity with complete rejection of tumors. Patient-derived PBMC treated with H1A showed superior tumor cell killing compared to NIV, PEM and ATZ (p<0.01). While H1A treatment of PBMC induced expansion of effector NK and CD8 T cells (GZMB+ T-bet+), ATZ treatment led to enrichment of Tregs (CD25+,Foxp3+) expressing inhibitory markers PD-1 and LAG-3. Conclusions: H1A demonstrated superior antitumor activity compared to FDA-approved ICI which lays the ground for clinical testing of H1A as a next-generation immune checkpoint inhibitor for the treatment of renal cell carcinoma.

381 Background: Venous tumor thrombus (VTT) is associated with worse oncologic outcomes in clear cell renal cell carcinoma (ccRCC). It is unknown if VTT and pulmonary embolism (PE) predispose patients to pulmonary metastasis. Methods: We queried our institutional registry for ccRCC patients undergoing radical nephrectomy (1970-2019). Cox proportional hazards regression models, adjusting for factors associated with ccRCC progression, were used to determine if VTT and PE were associated with pulmonary metastasis. Results: Of 3,410 patients studied, 1,025 (30%) had VTT, and patients with VTT were more likely to present with pulmonary metastasis at nephrectomy (20% vs 7%, p<0.001). Within the VTT subset, level of VTT was not significantly associated with pulmonary metastasis at nephrectomy (p=0.3). For all patients who had metastatic disease at nephrectomy (n=643), pulmonary metastasis was more common in those with VTT versus those without VTT (68% vs 51%, p<0.001). Excluding patients with pulmonary metastasis at nephrectomy, on multivariable analysis, VTT remained associated with post-nephrectomy pulmonary metastasis (hazard ratio (HR) 1.31, p<0.001) without a notable difference in HRs between renal vein (1.27) and caval (1.38) VTT. Presence of PE in the setting of VTT was associated with increased pulmonary metastasis post-nephrectomy (HR 2.31, p<0.001). Conclusions: VTT is associated with disproportionately increased pulmonary metastasis at presentation and post-nephrectomy in ccRCC patients. Further, the presence of PE at nephrectomy for VTT was associated with increased pulmonary metastasis post-nephrectomy. These results support a metastatic predilection of ccRCC with VTT to the lungs.