Conference History

Breast cancer–related lymphedema (BCRL) is a long-term and often life-altering consequence of breast cancer therapy, frequently occurring after procedures such as mastectomy, axillary lymph node dissection, and radiation treatment. While clinical and demographic factors contributing to BCRL have been well described, growing research suggests that genetic variation may also influence an individual's susceptibility. There is evidence that a number of genes are involved in lymphatic vessel development, inflammatory pathways, immune regulation, and lymphatic function. These have been proposed as potential contributors. Despite these insights, progress in the field is constrained by the limited number of studies, small cohorts, and inconsistent research designs. Continued investigation may ultimately support improved risk stratification and the development of tailored approaches to reduce the occurrence and impact of secondary lymphedema.

The ABO gene on chromosome 9 is highly polymorphic and clinically important, with over 300 variants tied to blood type, cardiovascular disease, COVID-19 severity, and the gut microbiome. To help students explore personal genomics, MiRcore created the Sequencing Your Genome (SYG) program, a three-week course where high schoolers extracted DNA, performed whole-exome sequencing, and analyzed their own ABO variants. With guided instruction, students processed Illumina sequencing data (Cutadapt, Bowtie2, SAMtools), visualized the ABO locus in IGV, and compared variants to ClinVar and UCSC Genome Browser. They successfully identified common and rare variants and reproduced known allele frequencies. Surveys showed increased genomic literacy, confidence, and STEM interest. The pilot highlights the value of citizen-science education, enabling students to own and interpret their genomic data while contributing to population genetics research.

Major cardiovascular events (MACEs) remain a leading cause of death and hospitalization in the United States. In Kentucky alone, 12.4% of the population, a total of 558,723 people, have been diagnosed with atherosclerotic cardiovascular disease (ASCVD). Additionally, the prevalence of heart failure in the US is approximately 1.9-2.6% with an additional 33% at risk of heart failure or already in a stage of pre-heart failure. Our aim was to develop a non-invasive, effective monitoring device for patients at risk of developing heart failure, decompensating, or experiencing a MACE, improving outcomes by alerting physicians of critical changes early. We identified brachial ankle pulse wave velocity (baPWV), as an early indicator used widely in Asia, but underutilized in the US. We identified photoplethysmography (PPG) sensors as an effective way to measure baPWV and additional metrics of interest, heart rate variability (HRV), ankle-brachial index (ABI), and reflection index (RI). We designed and built a prototype which utilizes PPG sensors coupled to an Arduino computer and attached to three flexible arm and leg bands. To process the signal output, we developed a random forest regressor machine learning model. We verified our prototype against standard clinical measurement techniques and reference literature values for healthy participants. Our current prototype measures baPWV within ≤ 2.0 m/s error, ABI within ≤ 0.10 error, and RI within ≤ 15% of reference measurement at a signal-to-noise ratio (SNR) of ≥ 40 dB under still conditions. We were able to obtain HRV reported as Standard Deviation of Normal-to-Normal (SDNN) heartbeat intervals within normal physiologic limits, but it did not meet our goal of ±2 ms error compared to Fitbit ECG-derived measurements. To improve accuracy of SDNN and functionality in moving conditions, future steps include retesting with a clinical ECG, adding an ECG to the wrist sensor, and incorporating accelerometers to all three sensors. Overall, our ATriFlow prototype accurately measured our metrics of interest, demonstrating strong feasibility as a non-invasive system for monitoring cardiovascular disease and reducing hospitalizations, morbidity, and deaths from MACEs.

Clean Intermittent Self-Catheterization is a preferred and cost-effective method for urine management but is inaccessible for patients with limited hand dexterity or tremor. We sought to optimize clean intermittent self-catheter insertion to be more accessible to persons with hand tremors and limited hand dexterity by adding innovative features to an existing lap-mounted catheter-insertion product marketed for adult males with spinal cord injuries. The weighted handle is a notable feature of our prototype that has not been included in catheter technologies to date. Our device offers universal intermittent catheter compatibility at low-cost with endless reusability of the adaptive device thereby offering patient's independence in catheter-assisted intermittent voiding.

The Micro/Nano Technology Center - Bioelectron Microscopy Center (MNTC-bioEM) is a core facility accessable for all researchers. Information regarding the core facility's equipment and services will be available on the following items. -Biological Sample Preparation -Cryo EM Sample Preparation -Ultramicrotomy -Scanning Electron Microscopy (SEM) -Transmission Electron Microscopy (TEM) -Large Data Analysis and Storage -3D Volumetric Model Building -Access to the core facility equipment, services and rates. -Training workshops -Research Voucher Opportunities

Foreign body retrieval using snare devices is a common but technically challenging task across several specialties such as radiology, cardiology, urology, and more. Existing snare devices rely on simple loop-based designs which limit control, stability, and visualization when manipulating three-dimensional objects under two-dimensional imaging. These limitations often result in prolonged procedure time, increased radiation exposure, and higher complication rates. Our project, the Bulb Snare System, is a novel retrieval device designed to address limitations of current models. Through a unique multi-plane, bulb-shaped capture head and ergonomic torque-control handle, the device improves visualization, grasping, and precision. Bench testing in simulated intraluminal environments demonstrated successful capture of a variety of foreign objects, with reproducible reductions in time-to-capture compared to common existing snares. By enhancing control, efficiency, and reliability, the Bulb Snare System has the potential to reduce procedural variability, minimize tissue trauma, and lower overall healthcare costs. This technology represents a feasible and scalable advancement in foreign body retrieval within existing regulatory and manufacturing frameworks.

Introduction: Diabetic nephropathy (DN) is a microvascular complication of Type 1 diabetes that results in kidney damage and associated with prolonged exposure to high blood glucose levels. N6-methyl-adenosine (m6A) is an epitranscriptomic RNA modification mediated by two writers, METTL3 and METTL14, which form in a complex to methylate position six of adenine. This modification has a regulatory role in oxidative stress, inflammation, and fibrosis, which are all related to various stages in the pathogenesis of DN. The role of m6A and its potential regulatory mechanism in DN remain largely unknown, so its study will broaden our understanding of m6A in diabetic nephropathy leading to novel biomarkers and therapeutic targets for treatment. Aim/Objective: Our aim is to develop and use methods to determine the relative abundance and cellular localization of METTL3, METTL14, and m6A cultured in human tubular epithelial (HK11) cells cultured in a media designed to represent the diabetic milieu. The goal was to test the hypothesis that the spatial localization and relative abundance of METTL3, METTL14 and m6A will vary in HK11 cells exposed to hyperglycemic culture conditions compared to low glucose media. Methods: HK11 cells were plated into 8-well coverslips in normal media + 5% Fetal Bovine Serum. Once desired confluency was reached, cells were serum starved for 24 hours to allow for cell cycle arrest. HK11 cells were exposed to three treatment conditions including low glucose (LG, 10 mM glucose), high glucose (HG, 30 mM glucose) and low glucose + mannitol (Mann, 10 mM glucose + 20 mM mannitol). Primary antibodies for METTL3, METTL14 and m6A were added to the wells. The secondary antibody, labeled with Alexa Fluor 546, was added along with DAPI to stain the nuclei. Confocal microscopy and ImageJ (https://imagej.net/ij/) analysis was used to quantify the total, nuclear, and cytoplasmic protein and modification abundance. Differences in abundance of means between treatment conditions and spatial localization was determined using two-way ANOVA and post-hoc t-tests were used to determine pair-wise statistical significance in results. An unadjusted p-value<0.05 was considered significant. Results: METTL3 and METTL14 were significantly enriched in the nuclei versus the cytoplasm. HG induced a significant increase in cellular fraction of nuclear METTL3 and significant decrease in cytoplasmic METTL3. No significant differences were observed for HG on METTL14 abundance or localization. m6A was significantly enriched in the cytoplasm. HG did not significantly alter m6A distribution or abundance over NG. Mann significantly increased the nuclear and decreased the cytoplasmic m6A's fractional abundance. Conclusions: Epitranscriptomics is an evolving and exciting area of research into disease mechanisms for diabetic nephropathy. Our study addressed the impact of hyperglycemic conditions on the tubular epithelial cells m6A, METTL3, and METTL14 localization and abundance. Significant findings from our study identify the role for osmolality in m6A abundance. Thus suggested a role for epigenetics in the control of solute concentration mechanisms in the tubular epithelium. These findings open a new window into the epigenetic mechanisms regulating kidney function and DN. These gaps in knowledge support the need for further exploration to define their roles in the progression of the disease. Future studies include obtaining m6A, METTL3, and METTL14 quantification data via confocal images of HK11 cells cultured in media more representative of normal conditions (5 mM glucose) and the diabetic milieu (25 mM glucose) through an automated image analysis protocol in ImageJ. Additionally, total nuclear RNA samples of cells treated in these conditions have been submitted for analysis to incorporate RNA-immunoprecipitation sequencing to identify how the diabetic environment alters m6A markers in a transcript specific manner.

Background Healthcare systems are rapidly integrating artificial intelligence into clinical workflows, yet medical training continues to depend on resource-intensive in-person simulations and faculty-dependent assessments. These traditional approaches may be difficult to sustain in rural institutions, resource-limited programs, and low-income settings. AI-driven virtual patient consultations modeled after OSCE/C3DO encounters represent a potential innovation capable of expanding access to high-quality clinical training across diverse educational environments. Objectives This study evaluated whether repeated exposure to AI-simulated clinical encounters improves learner performance and explored the broader potential of such systems to support scalable and accessible clinical education. Methods A pilot implementation involving 120 fourth-year medical students incorporated 16 AI-simulated virtual consultations over four weeks. Encounters replicated standardized clinical scenarios and were scored using predefined performance criteria. Linear mixed-effects models assessed performance change across consultations while accounting for individual learner variation. Results Performance scores improved substantially across repeated encounters, increasing from a mean of 56.8 to 79.3 (p < 0.001). Modeling demonstrated significant improvement between first and final consultations and a positive association between consultation number and performance score (1.14-point increase per encounter, p < 0.001). Conclusions AI-simulated clinical encounters may represent a scalable innovation in medical education with implications beyond traditional academic settings. By reducing reliance on standardized patients, simulation centers, and faculty-intensive assessment models, virtual patient platforms may expand access to structured clinical training in rural institutions, underserved regions, and low-resource countries. Integrating such tools into medical curricula may help bridge disparities in training infrastructure while preparing future clinicians for technology-enabled healthcare systems.

Purpose: To compare the scientific rigor of BSS use in generalized JH studies (healthy subject injury risk/rate, physiological or kinesiological function determination and joint-specific or arthroscopy JH studies; to identify the most commonly used BSS score thresholds; and to describe ways to improve BSS score use for improved surgical and clinical decision-making. Methods: Following the Preferred Reporting Items for Systematic Reviews and Meta-analyses guidelines, the PubMed, EBSCO Host, and Web of Science databases were searched using "Beighton score" and "sports injury outcome" terms. Study purpose, publication year, female or male subject number, age and group type, measurement tools, BSS criteria, results and conclusions data were extracted. Results: Twenty-eight generalized JH studies (43.8%, 28/64) involving 12,138 subjects (6512 females) and 36 joint-specific or arthroscopy JH studies (56.3%, 36/64) involving 7351 subjects (3441 females) were identified. Overall, most studies reported that BSS scores influenced most/all (54.7%, n = 35/64) or some (26.6%, 17/64) study outcomes and most were evidence level II (57.8%, n = 37/64) or III (35.9%, n = 23/64). Most generalized JH studies used a BSS score ≥ 4 (n = 13, 46.4%) or BSS ≥ 5 (n = 7, 25%) while most joint-specific or arthroscopy JH studies used a BSS score ≥ 4 (n = 17, 47.2%), a BSS score ≥ 5 (n = 9, 25%, or the full BSS score scale (n = 7, 19.4%). Joint-specific and arthroscopy JH studies were more recently published (2020.2 ± 3.5 vs. 2014.2 ± 6.8, p < 0.001). Generalized JH studies more frequently reported separate subject sex and age data (53.6%, n = 15/28) while joint-specific or arthroscopy JH studies more often combined this information (92.7%, 33/36)(p = 0.001). Most generalized JH studies were evidence level II (85.7%, 24/28) while most joint-specific or arthroscopy JH studies were evidence level III (52.8%, 19/36)(p < 0.001). Group study quality and bias risk was comparable, however, generalized JH studies had more prospective research designs (96.4%, 27/28 vs. 58.3%, 21/36 (p < 0.001). Conclusion: Generalized JH studies had more prospective research designs, had higher evidence levels, and more frequently reported separate subject age and sex details. Greater use of these characteristics in joint-specific or arthroscopy JH studies may strengthen surgical and clinical decision-making and patient outcome prediction validity.

Objective/Purpose: Burnout is one of the most pressing issues affecting physicians today. One of the leading contributors is increased patient message volume and time spent with EHRs (Holmgren AJ et al., 2025). Burnout decreases physician wellbeing and negatively impacts patient care. Therefore, there is a pressing need to optimize how physicians respond to patient messages. To address this, an AI-powered system was created that classifies messages based on urgency and generates patient responses to streamline workflow and reduce time spent responding to messages. A secondary goal was to create a low-cost solution for small healthcare systems. Methods: An AI-powered, end-to-end patient messaging system was developed for this study with assistance from Claude Sonnet. This is a physician-in-the-loop design, so the physician role remains necessary to ensure patient safety. The system automatically classifies patient messages based on clinical urgency while generating clinical documentation and patient-facing responses. GPT-4o-mini was used as the large language model (LLM) provider. This system incorporates safeguards to reduce risks of depending solely on LLM reasoning. Because LLMs are non-deterministic and prone to hallucinations, the platform integrates a self-consistency layer. Each message is processed through three independent calls to GPT-4o-mini, and the system uses majority consensus classification. This design minimizes error and outlier responses. Another novel feature is that it provides justification for each classification to increase physician trust. To improve efficiency, the system incorporates deterministic pre-filtering that automatically detects "LOW" urgency routine task messages. Conversely, "red flag" emergency symptoms automatically escalate emergency messages. The response generator writes patient-facing responses with human language reflecting the determined urgency. These components work together to provide physicians with a streamlined messaging workflow that augments, rather than replaces, their efforts. Data/Results: This study is undergoing IRB approval for validation. It will use 40 AI-generated mock patient messages pre-approved by independent physicians not participating in the evaluation. Outcome measures will include physician-AI agreement rate, comparative analysis of time spent per message, and qualitative feedback assessing physician comfort with AI, ease of use, and concerns regarding AI integration in clinical workflows. Conclusions: This AI-powered messaging system overcomes limitations of existing solutions by addressing all aspects of messaging workflow. It integrates LLM reasoning within a multi-layered architecture incorporating safety mechanisms and physician oversight to improve reliability and efficiency. This system has the potential to be a low-cost solution for small or rural practices. References: 1.Holmgren AJ;Apathy NC;Sinsky CA;Adler-Milstein J;Bates DW;Rotenstein L; Trends in physician electronic health record time and message volume. JAMA internal medicine. February 2025. Accessed November 8, 2025. https://pubmed.ncbi.nlm.nih.gov/39992635/.

Background: Middle ear disease is a common reason for primary care visits and a major contributor to preventable hearing loss, especially in the pediatric patient population. In many primary care and low-resource settings, diagnosis relies on otoscopic visualization alone, which is subjective and can miss effusion or reduced tympanic membrane mobility. Tympanometry objectively assesses middle ear function but is underused outside ENT and audiology due to device cost, bulky equipment, and the need for trained interpretation. Purpose: To develop an accessible and affordable accessory for otoscopes to enable tympanometry to be performed in point-of-care settings without requiring specialist interpretation, called the OtoTympact. Methods: Medical students and engineering students at the University of Louisville's Bluegrass Biodesign program collaborated with otolaryngology and engineering faculty. Clinical needs assessment and provider interviews were performed to identify barriers limiting tympanometry use in frontline settings, particularly cost and interpretation complexity. A Jerger Classification software was developed and validated using datasets from the National Health and Nutrition Examination Survey. A 3-D printed device housing was then created, along with an interface with a microcontroller linked to the software. Results: The software correctly classifies dataset results using Jerger classification and then provides simplified color-coordinated LED outputs on the 3-D printed housing designed for non-specialist clinicians. Type A indicates normal middle ear function and appears as a green light. Type B suggests a likely middle ear effusion or tympanic membrane perforation and appears as a blue light. Type C indicates negative middle ear pressure consistent with diseases like Eustachian tube dysfunction and appears as a red light. Conclusions: Early testing supports the feasibility of a combined otoscope-tympanometry attachment tool for rapid, point-of-care screening. Future development will integrate the classification software into an otoscopic-compatible attachment that can perform a real-time tympanogram, allowing for a simple color-coded interpretation of the tympanogram at the point-of-care. With continued development, the OtoTympact could enhance diagnostic accuracy, reduce unnecessary referrals, and expand access to objective middle ear assessment in a variety of care settings. Funding: We would like to thank the Bluegrass Biodesign and University of Louisville Engineering for funding and support.

Background: People living with Parkinson's disease (PD) face not only motor and non-motor symptoms, but also significant psychosocial and educational challenges—particularly at the time of diagnosis. Feelings of isolation, uncertainty, and difficulty navigating medical systems are common. At the same time, medical trainees often have limited exposure to the lived experience of Parkinson's outside of clinical settings, creating gaps in empathy, communication, and patient-centered care. Objective: The Parkinson's Buddy Program was developed by the National Parkinson's Institute (NPI) in Louisville, Kentucky, to address these dual needs by pairing individuals with Parkinson's disease with medical students in a sustained, relationship-based peer connection model. The program aims to reduce social isolation for participants with PD, enhance medical students' understanding of Parkinson's beyond textbooks, and support the development of a more compassionate and well-prepared movement disorders workforce. A core goal of the program is national expansion to help address the growing shortage of movement disorder specialists across the United States. Methods: Participants with Parkinson's and volunteer medical students are matched based on interests, communication preferences, and availability. Pairs connect regularly through phone calls, text messages, shared activities, and community-based events facilitated by NPI. The program operates throughout the academic year, with structured check-ins and opportunities for engagement that allow relationships to develop organically while remaining supported. Results: Since its inception, the Parkinson's Buddy Program has served over 100 individuals with Parkinson's and more than 100 medical students, generating over 3,000 hours of direct peer engagement. Participants with Parkinson's report reduced isolation, improved emotional well-being, and greater confidence navigating their care. Medical students report increased empathy, stronger communication skills, and deeper understanding of the lived experience of Parkinson's. Importantly, six former student participants have gone on to become movement disorder specialists, highlighting the program's potential to shape future workforce pathways. Conclusion: The Parkinson's Buddy Program demonstrates that sustained, community-based peer connection can meaningfully improve quality of life for people with Parkinson's while transforming medical education. Expanding this model nationally offers a scalable, low-cost strategy to strengthen patient support, improve clinician training, and help build the next generation of movement disorder specialists needed to meet the growing demand for Parkinson's care.

Optical coherence tomography (OCT) for staging age-related macular degeneration (AMD) is an important tool for retina clinic triage, treatment planning, and follow up. However, this automated tool often collapses the disease spectrum into broad groups and offers limited support for clinician review. We propose an interpretable OCT-based deep learning framework that grades a single macular B scan into six clinically meaningful categories spanning normal retina, dry AMD sub-stages, geographic atrophy, and inactive versus active wet AMD, producing visual explanation maps to highlight regions that influence the staging decision. Method: The method was evaluated retrospectively on a multi-center dataset acquired across commercial OCT platforms with specialist provided reference labels, and reproducibility was assessed using repeated runs with different random seeds. There were a total of 954 patients including 123 with normal findings, 127 with early AMD, 178 with intermediate AMD, 38 with Geographic atrophy, 251 with inactive AMD, and 237 with active AMD. Results: Across the six classes, geographic atrophy showed perfect detection with sensitivity of 1.000, most classes maintained strong balanced performance with F1 scores exceeding 0.890, and receiver operating characteristic analysis demonstrated excellent separability with class AUCs ranging from 0.906 to 1.000 and a macro average AUC of 0.953. The explanation maps qualitatively localized areas consistent with stage defining retinal abnormalities, supporting transparent review alongside predicted stage. Discussion: These findings suggest that an interpretable OCT based staging approach can provide clinically aligned grading across the AMD spectrum and may be useful for screening, triage, and longitudinal monitoring, while prospective evaluation remains necessary before clinical use.

Background: Generative artificial intelligence (AI), including large language models (LLMs), is increasingly used in clinical medicine to support guideline retrieval and decision-making. In surgical oncology, however, management decisions are complex, multidisciplinary, and often controversial, particularly in settings without access to specialized tumor boards. Objective: To evaluate the ability of ChatGPT to generate clinically relevant recommendations for complex and debated surgical oncology scenarios and to define its strengths and limitations across key dimensions of oncologic decision-making. Methods: Twelve clinical vignettes representing commonly debated surgical oncology scenarios across multiple disease sites were developed. Each prompt was entered into ChatGPT (version 4.1) to request concise management recommendations. Responses were independently evaluated by four fellowship-trained surgical oncologists and one surgical oncology fellow using a 10-domain Likert-scale rubric assessing accuracy, relevance, risks, benefits, alternatives, uncertainty, and overall quality. Descriptive statistics were calculated, and a linear mixed-effects model was used to assess systematic performance differences across domains while accounting for rater and scenario variability. Results: ChatGPT performed best in relevance (mean ± SD: 4.0 ± 0.9) and accuracy (3.8 ± 1.0), with moderate alignment with standard of care (3.0 ± 0.8). Performance was significantly lower for articulation of risks (2.1 ± 1.2), alternatives (2.2 ± 1.1), benefits (2.7 ± 1.2), and uncertainty (2.5 ± 1.1). Mixed-effects modeling confirmed significantly higher scores for relevance and accuracy and considerably lower scores for risks, alternatives, benefits, and uncertainty (all p ≤ 0.03). Performance was strongest in guideline-driven scenarios and weakest in clinically ambiguous cases. Conclusions: ChatGPT can reliably synthesize guideline-aligned oncologic information but demonstrates consistent deficits in higher-order clinical reasoning essential for individualized surgical oncology decision-making. These findings support its use as an educational or preliminary decision-support tool, while underscoring the need for expert oversight and caution before clinical integration.

Accurate identification of the epidural space is critical for safe and effective neuraxial anesthesia, yet current techniques rely heavily on the clinician's subjective perception of loss of resistance, contributing to variability in success rates and prolonged training curves. We present the Epidural Pressure Alert Located (Epidural PAL), a compact, reusable device designed to transform the subjective "feel" of epidural placement into an objective, real-time pressure-based alert. The Epidural PAL connects inline with a standard epidural syringe and continuously measures fluid pressure as the needle is advanced. Upon entry into the epidural space, the characteristic sudden drop in resistance produces a distinct pressure-drop signature, which the device immediately detects and communicates through both visual and audible alerts, prompting the clinician to stop advancement. The Epidural PAL is designed to integrate seamlessly into existing workflows without altering established technique, while providing real-time confirmation of epidural space entry. By offering objective feedback at a critical procedural moment, the Epidural PAL has the potential to improve first-pass success, enhance procedural safety, and reduce training time for clinicians learning epidural placement.

Length of hospital stay (LOS) is a critical metric in orthopedic care. While factors influencing LOS in hip fractures are known, research on isolated femoral shaft fractures is limited. This study's main purpose was to identify the demographic, clinical, and treatment-related factors that prolong LOS specifically in this patient population. This retrospective study analyzed data from 551 patients with isolated femoral shaft fractures. We calculated descriptive statistics for patient demographics and clinical values. Negative binomial log-linked generalized linear models were employed to analyze the impact of variables, including age, sex, Body Mass Index (BMI), transfusion status, hemoglobin levels, and pre-operative lactic acid, on LOS. The average LOS was 4.22 days. A BMI greater than 40 was significantly associated with an approximately 39% increase in expected LOS (P < .05) compared to patients with a normal BMI. In non-transfused patients, each unit decrease in hemoglobin at discharge was associated with an 11.3% increase in expected LOS (P < .05). Additionally, a secondary analysis found that female patients had higher odds of a prolonged LOS (P < .05). Age, comorbidities, and pre-operative lactic acid levels were not significant predictors of LOS. A BMI greater than 40, lower hemoglobin levels at discharge, receiving transfusion during admission, and female sex are key factors associated with a longer hospital stay following an isolated femoral shaft fracture. These findings suggest that emphasizing earlier mobility and more intensive physical therapy for obese patients, as well as maintaining adequate postoperative hemoglobin levels, may optimize patient pathways, reduce LOS, and improve the quality of care.

Abstract A Biologically Constrained Computational Platform for Mechanistic Modeling of Cancer Evolution and Therapeutic Response Background: Cancer therapy development remains limited by an incomplete ability to model tumor behavior as a dynamic biological system. Current computational approaches often rely on correlation-based machine learning or simplified biological assumptions, which limits interpretability and predictive reliability, particularly for solid tumors such as pancreatic ductal adenocarcinoma. Methods: We are developing a computational platform that represents cancer as a population of interacting biological state models, integrating genetic, signaling, metabolic, phenotypic, and microenvironmental factors. The system simulates tumor evolution under biologically constrained transition rules, with an artificial intelligence component generating interpretable therapeutic hypotheses. Preliminary Results: Initial simulations demonstrate feasibility of modeling heterogeneous tumor populations and capturing qualitative patterns of adaptive response to therapeutic perturbations. Conclusion: This work presents a biologically grounded approach to cancer modeling that integrates constrained artificial intelligence with mechanistic biological simulation to support rational therapeutic diagnoses

Background: The 21st Century Cures Act grants patients immediate electronic access to clinical records, often exposing them to complex diagnostic information before clinician counseling. Increasingly, patients turn to large language models (LLMs) rather than traditional search engines for interpretation. How LLMs frame and communicate medical information in these early encounters has implications for health literacy, equity, and patient safety. Methods: We conducted an in silico, simulation-based, cross-sectional study using GPT-4 on a cloud-based platform. A total of 2,550 responses were generated from 225 expert-selected gastrointestinal pathology diagnoses. Prompts varied by identity ("I am a patient" vs "I am a physician"), thematic domain (oncologic classification, clinical management, diagnostic associations), and language complexity (simplified vs complex). Readability was assessed using Flesch Reading Ease (FRE) and Flesch–Kincaid Grade Level (FK). Multivariable linear regressions evaluated predictors of readability, and references to healthcare professionals were quantified. Results: Physician-initiated prompts consistently produced responses with lower FRE scores and higher FK grade levels compared with patient-initiated prompts across all domains (p<.002). Increased prompt language complexity independently increased response complexity. For clinical management queries, malignant and neither benign nor malignant diagnoses were associated with higher FK scores, while malignant diagnoses in diagnostic association queries were associated with greater readability, potentially reflecting empathetic framing. Patient-initiated prompts more frequently elicited professional-identifying language and safety disclaimers. Conclusions: LLM responses are sensitive to prompt identity, linguistic input, and diagnostic context, often producing patient-facing explanations that exceed recommended readability levels. These findings underscore the need for AI systems that align medical explanations with average health literacy, allow users to guide informational versus emotional framing, and support equitable patient understanding in the era of open clinical notes.

Minimally invasive catheterizations are the diagnostic and therapeutic standard of care for children with arrhythmias, congenital heart defects, and other cardiac pathologies. Pediatric cardiologists operate in a world where accuracy, precision, and specificity all intrinsically correlate with interventional outcomes. The current landscape of procedural imaging techniques—including coronary angiography, X-ray fluoroscopy, and echocardiography—fails to provide direct in vivo visualization of pertinent anatomy. Thus, minimally invasive procedures, while far safer than the open-heart surgeries of yesterday, still confer risks for iatrogenic injury due to present imaging limitations. It was this environment that inspired the development of Visi-Cath. This multifunctional catheter provides real-time feedback by allowing the proceduralist to visualize intracardiac and intravascular structures during ablations, catheterizations, and other techniques performed in a catheterization lab. The Visi-Cath employs a fiber-optic, multi-lumen design with a saline flush, allowing its user to directly visualize the anatomy they are assessing. This revolutionary tool will significantly reduce opportunity costs for hospitals, help electrophysiologists and interventionalists perform more effectively, and markedly reduce barriers to care in Kentucky and beyond. Our prototype device was extremely successful during initial testing, and its combination of financial practicality and clinical utility will make it a staple of pediatric cardiology for years to come.

Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition with a strong genetic component, affecting millions worldwide. Early diagnosis is critical for effective intervention, yet the heterogeneity of ASD poses significant challenges. Several studies have been investigated the role of genomic markers and artificial intelligence (AI) in early diagnosis of ASD with no definitive conclusion. Therefore, this study aims to explore several research questions to fill the gap and open the door for future researchers. First, what are the optimal genomic markers associated with ASD? Second, which AI tools are commonly used and can provide effective diagnosis for ASD? Last, does the combination of genomics and AI offers an added diagnostic value? Hence, we conducted a systematic review to find all the related studies published in the last decade. After searching well known databases including, scholar Google, PubMed, Scopus, IEEE explore, and Research Gate, using Keywords such as "genomics," "AI," "Autism", and "machine learning" to identify relevant studies, 46 studies focusing on genomics alone or in combination withAI in ASD diagnosis were identified. A PRISMA chart was employed to ensure transparency, and inclusion/exclusion criteria were applied to obtain these studies. The analysis showed that several genomic markers, including NBEA, HOXB3, HERC1, NR2F2, and MID2, are strongly associated with ASD. These markers are linked to chromatin remodeling, neuronal signaling, and synaptic function, highlighting their potential as diagnostic biomarkers. In terms of AI tools, methods like convolutional neural networks (CNNs), gradient boosted trees, and random forests demonstrated high diagnostic accuracy, with some models achieving AUC-ROC values up to 0.955 and accuracy rates exceeding 88%. Studies utilizing AI in combination with genomics consistently outperformed those relying only on genomics without using AI. The integration of both approaches offers a more robust and accurate diagnostic tool, paving the way for personalized and early intervention strategies. Future research should explore radio genomics as a non-invasive diagnostic direction, leveraging imaging and genomic data to further refine ASD diagnosis and understanding.

1 University of Louisville, Louisville, KY; 2 University of Hawai'i, Honolulu, HI; 3 Louisiana State University, Baton Rouge, LA; and 4 University of Toledo, Toledo, OH. Background and Purpose: Hexavalent chromium [Cr(VI)] is a ubiquitous environmental pollutant and potent toxicant, currently ranked as the number one chemical hazard in U.S. public drinking water. To date, most mechanistic data describing Cr(VI) toxicity come from lung, liver, and kidney tissues. While these studies have yielded extensive datasets, there remains comparatively little information on neurotoxicity. Even less is known about how pathways of Cr(VI) neurotoxicity may vary by sex. This study assessed how chronic Cr(VI) exposure induces distinct metabolic signatures in the hippocampus that reflect oxidative stress, altered energy metabolism, and neurotransmitter disruption, and how these signatures differ between male and female guinea pigs. Methods: Hartley guinea pigs were given either tap water or 5 mg/L Cr(VI) in drinking water for 90 days. Hippocampal tissues were collected and processed for untargeted metabolomics and metallomics. Inductively coupled plasma–mass spectrometry was conducted to determine hippocampal Cr accumulation. Statistical analyses used the Mann–Whitney test with significance set at p < 0.05. Results: Hippocampal metabolomics from Hartley guinea pigs exposed to 5 mg/L Cr(VI) in drinking water for 90 days revealed distinct metabolite profiles that differed markedly between males and females. In males, Cr(VI) exposure was associated with a broad loss of metabolic reserve capacity. Creatine levels were lower and glyceric acid levels were reduced, suggesting a diminished ability to buffer energy demand. There were reductions in several amino acids that support neurotransmitter synthesis, including lower GABA, glutamate, and histidine, suggesting a weakened ability to sustain synaptic signaling. Males displayed decreased membrane-building precursor molecules such as O-phosphorylethanolamine and neuraminic acid, implying stress on membrane synthesis. Levels of antioxidant-linked intermediates such as γ-glutamyl-2-aminobutyric acid and phenyl-lactate were lower, consistent with a reduced redox response. N1-acetylspermidine decreased, indicating reduced polyamine turnover, which is important for stress resilience and synaptic plasticity. Cr(VI) exposure in females was accompanied by increased dipeptides and increased glutamate, lysine, O-phosphorylethanolamine, and neuraminic acid, indicating amino acid and membrane-related support pools are increased in response to Cr(VI) exposure. Females displayed coordinated decreases in major myelin-associated lipid classes, including lower levels of ceramides, sphingomyelin, and galactocerebrosides. This pattern of lipid loss is consistent with vulnerability of white-matter-related lipid structures and may indicate stress to myelin integrity. Conclusions: Chronic Cr(VI) exposure induced distinct hippocampal metabolomic fingerprints between sexes. Male guinea pigs exhibited broad depletion of metabolic substrates that normally support energy buffering, neurotransmitter cycling, antioxidant defense, and membrane synthesis. Meanwhile, females exhibited increases in amino acid metabolites but coordinated losses in myelin-associated lipids. Together, these results demonstrate chronic Cr(VI) exposure produced distinct metabolic response patterns in the hippocampus between sexes. The hippocampus is a key structure for learning and memory, making these metabolic shifts relevant to functional neurotoxicity. These metabolic shifts align with prior Cr(VI) findings, including disrupted GABA/glutamate pathways and evidence of myelin-related vulnerability. Further, these results underscore the importance of evaluating sex as a biological variable for brain susceptibility to environmental toxicants. Support from R35ES032876 (JPWSr.), "la Caixa" Foundation LCF/BQ/EU22/11930060 (IM), T32ES011564 (STV, MD, GCG), R21ES033327 (JPW), and USDA Multistate Research Project W5045 (JTFW).

Drug-induced liver injury (DILI) remains a diagnostic challenge, particularly when associated with over-the-counter herbal supplements that are often underreported by patients. The increasing use of alternative therapies has contributed to a rising incidence of supplement-associated hepatotoxicity, which may present with nonspecific symptoms and overlap with other hepatobiliary conditions. We report the case of a 44-year-old Hispanic female with a history of cholecystectomy and gastroesophageal reflux disease who presented with recurrent right upper quadrant abdominal pain, nausea, vomiting, and persistent cholestatic liver enzyme elevation despite prior endoscopic intervention for suspected sphincter of Oddi dysfunction. Extensive evaluation, including cross-sectional imaging, serologic testing, endoscopic retrograde cholangiopancreatography, and ultimately liver biopsy, revealed macrovesicular steatosis with portal lymphoplasmacytic inflammation and mild interface activity, raising concern for drug-induced liver injury. Subsequent detailed history uncovered chronic ingestion of oregano oil and black seed oil supplements over several months, which were discontinued. This case highlights the diagnostic complexity of DILI in the setting of metabolic liver disease and overlapping gastrointestinal pathology. It underscores the importance of obtaining a thorough supplement history and maintaining a high index of suspicion for herbal supplement–associated hepatotoxicity, even when initial work-up suggests alternative etiologies.

Background: Pediatric emergency care presents distinct clinical and ethical challenges, including time-sensitive decision-making and developmental variability. As artificial intelligence (AI) enters clinical workflows, its real-time applications in pediatric emergency departments (EDs) require evaluation. Objectives: To examine how AI has been implemented in real-time pediatric emergency care and summarize its clinical applications, outcomes, and associated ethical or cognitive considerations. Methods: A scoping review was conducted in accordance with PRISMA-ScR guidelines. MEDLINE, Embase, CINAHL, Web of Science, and the Cochrane Library were searched for English-language studies published between 2020–2025. Eligible studies evaluated AI-based tools applied to real-time clinical care of children (0–18 years) in ED settings. Results: Six studies met inclusion criteria. AI applications primarily supported diagnostic interpretation and clinical decision-making, with emerging exploration of psychosocial and workflow-related functions. Reported outcomes included improved diagnostic accuracy, increased workflow efficiency, and enhanced patient experience. All AI systems operated with clinician oversight. Ethical considerations including consent, trust, and cognitive workload, were acknowledged but inconsistently evaluated. Conclusion: AI in pediatric emergency care is currently positioned as a clinician-augmenting tool, particularly in diagnostic and supportive roles. While early findings are promising, limited study volume and external validation constrain generalizability. Future research should address ethical safeguards, clinician cognition, and implementation equity.

Background: Gamified learning is transforming medical education by enhancing engagement and memory retention. Inspired by Duke-NUS Singapore's TBL model and global studies, we explore how gamification—across TBL, PBL, and AI-assisted learning—can activate the limbic system and reduce the forgetting curve. Aim: We aim to identify practical gamification strategies that foster long-term learning retention in medical education. Methods: Our approach is a robust synthesis of literature. We critically analyze existing TBL/PBL gamification models—such as those at Duke-NUS—and gamified learning studies, connecting them to memory theory (emotional anchoring and the forgetting curve). Rather than direct experimentation, we extract evidence-based principles for immediate classroom application. Results: Preliminary findings suggest that gamification, when tied to emotional engagement and competition, enhances learner participation and memory retention. Conclusion: Gamified learning, grounded in evidence and emotion, offers a powerful path to resilient, joyful medical education. We propose actionable strategies to implement today, while paving the way for future systems of learning.

Introduction With the increasing use of generative AI in every facet of work, the use of such models continues to pose efficiency in the medical landscape. Every year, medical school admissions teams are tasked with evaluating thousands of applicants on numerous attributes. While there is growing investigation into the use of AI to evaluate personal statements, interviews, and other aspects of the medical school admissions process, there is a gap in knowledge as to how common AI models evaluate medical school applications on core characteristics. Because of this, it is vital that the trends in their ability to evaluate applicants be studied. Methods We provided 5 commonly used and publicly accessible forms of generative AI (ChatGPT, Google Gemini, Microsoft Copilot, Perplexity, and Claude) with a common prompt in selecting a group of 7 medical students from a pool of 100 applicants (representing allopathic medical school acceptance rates). The prompt was engineered to avoid bias while allowing each form of AI to make its own choices. We then provided each form with a randomized pool of theoretical applicants of varying characteristics (MCAT scores, GPA, demographics, common flags [first generation medical students, active military], extracurricular flags [athletics, leadership, service, research]). After recording applicant choices, each model was asked to provide rationale for its choices. This was recorded and coded for themes of influence. Results Data collected in the form of applicants chosen was compared between the 5 AI models used. Each rationale was coded for common themes, and those themes were then compared. Data collection is ongoing. Conclusion With the increasing commonality of Generative AI, these tools continue to grow as everyday assistants. Evaluating how these platforms make decisions with tasks like screening large volumes of applicants is key to understanding the benefits and risks that are posed in their use.

Introduction ReSpectacle, a 501(c)(3) nonprofit organization, was established in 2011 with a mission to distribute repurposed prescription eyeglasses to individuals lacking access. It is dependent on a national network of student-led chapters that coordinate the processing and shipping of donated glasses to the main headquarters in Hudson, WI. The ReSpectacle Louisville Chapter was founded in 2022 as a student-led initiative at the ULSOM. Within a year, the chapter initiated a partnership with Goodwill Industries of Kentucky, creating a community-based model for sustainable eyeglass collection and redistribution. Building upon the chapter's early activities, this study evaluates the effectiveness of this first and only ReSpectacle-Goodwill collaboration in increasing access to vision correction through repurposed eyeglasses. Methods Medical student volunteers from the University of Louisville met monthly to process donated eyeglasses, categorizing them as prescription, reading, sunglasses, blue-light, or damaged. Prescription glasses were cleaned, photographed, and uploaded to ReSpectacle's online redistribution database. Before partnering with Goodwill, the chapter independently funded the shipment of processed eyeglasses to ReSpectacle headquarters, with costs ranging from approximately $75 – $128 per 50-lb box. Afterwards, Goodwill assumed responsibility for all shipments to ReSpectacle headquarters for final quality assurance and distribution. Program records from 2022 to 2025 were reviewed to quantify eyeglass collections, volunteer participation, and geographic reach. Retrospective data were analyzed to evaluate program growth and operational efficiency. Results Prior to the Goodwill partnership, fewer than 1,000 pairs of eyeglasses were processed, reaching only one to two nearby states. Following, annual collections increased to 6,925 pairs in 2023 and 11,626 pairs in 2024, with 19,583 total pairs collected by 2025. To date, eyeglasses have expanded to 65 cities across more than 30 states, reflecting the expanding national reach and impact of this community collaboration. This partnership vastly helped to streamline the utilization of Goodwill's embedded interstate trucking routes that eliminated shipping costs incurred by the local ULSOM chapter, allowing better use of student resources towards volunteer training and community outreach. Conclusions Three years after its founding, the unique collaboration between the ReSpectacle Louisville Chapter and Goodwill Industries of Kentucky demonstrates how student-led community partnerships can play a meaningful role in expanding access to vision correction. Due to its success, the collaboration has been extended to 2028. Moving forward, efforts will focus on maintaining this momentum, refining processes, and further expanding access.

Abdominoplasty remains one of the most commonly performed cosmetic surgical procedures, with postoperative satisfaction closely tied to the aesthetic appearance of the umbilicus. Despite its lack of functional role in adulthood, the umbilicus serves as the central focal point of the abdomen, and postoperative shrinkage or stenosis may significantly detract from overall surgical outcomes. Narrowing of the umbilicus is primarily attributable to cicatricial contracture and ongoing scar remodeling, and may be exacerbated by vascular compromise after flap elevation. Multiple surgical techniques have been described to address postoperative umbilical contraction; however, many focus primarily on structural revision without directly addressing the biologic mechanisms contributing to ongoing fibrosis. We describe an adjunctive technique combining Mercedes-pattern scar release, topical silver nitrate application, and temporary foam earplug stenting to restore and maintain umbilical patency. This approach relieves circumferential tension while supporting controlled wound remodeling and stabilizing the umbilical aperture during healing. The adjunctive use of silver nitrate with temporary stenting represents a simple and reproducible method to preserve umbilical contour and optimize aesthetic stability following abdominoplasty.

Idea Bringing medical, engineering, and business students together fosters teamwork, communication, and innovation to address unmet clinical needs. Need/Rationale The current healthcare landscape is experiencing unprecedented growth in biomedical markets as rapid technological evolutions and changes in disease states create an ever-increasing demand for innovative solutions. As medical care becomes more specialized, new challenges emerge that require equally specialized innovations. Current medical education curricula often overlook structured opportunities for students to explore biodesign and the entrepreneurial skills required for it. Harnessing the curiosity and creativity of interdisciplinary students with an interest in biodesign provides an untapped resource for identifying clinical inefficiencies and generating novel solutions. By introducing learners to innovation principles early in their training, we cultivate future clinicians and collaborators who are comfortable working across disciplines. Our program leverages the collective expertise of medical, engineering, and business students to address unmet clinical needs, foster teamwork, and accelerate the translation of ideas into impactful devices. This model not only addresses the growing demand for medical technology but also prepares the next generation of physicians to be innovators and advocates for advancing patient care. Methods Bluegrass Biodesign is a 9-month interdisciplinary program founded in 2021 by medical students to expose learners to healthcare innovation. Target learners include medical and engineering students, with future expansion to business and law students, residents, and fellows. Our program's goal is to train participants to identify unmet clinical needs and develop viable solutions through collaboration. Learners are grouped into interdisciplinary teams aligned with specialties such as ophthalmology, cardiothoracic surgery, neurosurgery, otolaryngology, radiology, pediatric cardiology, pulmonology and critical care, and urology. Under guidance from physician and engineering faculty, teams progress through two structured components: (1) a standardized curriculum covering medical entrepreneurship, intellectual property, and commercialization, including lectures from faculty, device companies, and the NSF I-Corp LaunchIt program; and (2) a biodesign pipeline guiding teams from needs assessment to prototype development, with opportunities for patents or commercialization. Training occurs at the medical school with partnerships across colleges, spanning the academic year. Interactive methods emphasize teamwork, mentorship, and iterative design, with students actively engaged in identifying, inventing, and implementing solutions to real-world problems. Program Evaluation Plan Program effectiveness is monitored through a structured evaluation process built around our standardized curriculum. All participants must complete required modules to receive a certificate of completion. The curriculum is overseen by an executive board composed of student alumni and faculty mentors, ensuring peer-driven accountability and faculty oversight. Feedback is continuously collected from participants, mentors, and stakeholders, with results reviewed at the end of each cohort cycle. This feedback directly informs curriculum revision, allowing us to refine content and improve delivery for future cohorts. Additionally, learner progress is tracked through milestones such as needs identification, ideation, prototype development, and final presentation. Success is measured not only by course completion but also by the ability of teams to generate tangible outcomes such as prototypes, patent applications, or partnerships. Longitudinal tracking of alumni participation in research, innovation, or entrepreneurial activities will provide insight into sustained impact. This dynamic evaluation system ensures our program remains responsive to learner needs and the evolving landscape of medical innovation. Potential Impact on the Field or Beyond Integrating bioinnovation training into medical education empowers future physicians to identify inefficiencies and create practical solutions. Expanding such programs nationwide could drive healthcare innovation and cultivate interdisciplinary leaders prepared for evolving clinical challenges.

Background: Using machine learning and artificial intelligence (AI) techniques, we have developed a computer vision tool for the automated interpretation of electrophoresis gel scans. Serum protein electrophoresis (SPE) is a well-established routine laboratory diagnostic technique used to quantify different types of blood proteins, whose composition can be altered by various pathologies. Serum immunofixation electrophoresis (IFE), a type of SPE, can help determine disease type, severity, prognosis, and is useful for disease monitoring. Serum IFE is primarily indicated for patients suspected of having plasma cell dyscrasias/disorders, such as multiple myeloma, a type of blood cell cancer. Anti-sera is added into distinct gel lanes for identification of antibody heavy chains (IgG, IgA, IgM) and light chains (kappa and lambda) which correspond to specific monoclonal immunoglobulins and have clinical significance. Interpretation of IFE gels requires significant time and effort from highly skilled experts in the field of clinical pathology and/or laboratory medicine. Methods: Over 4,000 archived serum IFE scan images from UofL Health have been collected and processed for model development, with 45% positive for monoclonal proteins and the remainder negative. Model parameters were tuned using 70% of the data for model training, with the remainder reserved as validation data. Additional data was collected as internal (N = 906) and external (N = 120) test data, both of which were excluded from the training process. Training, validation, and internal test scans were labeled according to routine clinical laboratory reports which were interpreted by board-certified clinical pathologists. External test images were labeled by consensus agreement among 3 independent visual interpreters. Results: When predicting if a band corresponding to a monoclonal protein was present or not present in IFE gel scans, the model achieved a sensitivity of 0.96, specificity of 0.98, and AUROC of 0.99, on internal test data, and sensitivity of 0.89, specificity of 1.00, and AUROC of 0.98, on external test data, compared to routine visual interpretation. AUROC was 0.95 and 0.96 for identifying IgG heavy chains and 0.98 and 0.99 for kappa light chains within internal and external test data, respectively. IgM heavy chains had an AUROC of 0.95 and 0.73 within internal and external test data, respectively. Further, heavy chain and light chain combinations were evaluated for monoclonal protein identification. For instance, IgG-kappa monoclonal proteins were detected with a sensitivity of 1.0 in both internal and external test scans. The model pipeline is also practical to implement, with an estimated average computation time of 16 seconds per sample using personal computer class desktop hardware and less than 0.5 seconds per sample with professional workstation class hardware. Conclusion: The computer vision tool is poised to be integrated with routine clinical laboratory workflows where it can assist the interpretive efforts of medical staff by triaging patients with potential plasma cell disorders.

Lambert Eaton Myasthenic Syndrome (LEMS) is a rare autoimmune neuromuscular junction disorder or paraneoplastic syndrome. Patients often present with a clinical triad of gradual proximal muscle weakness most commonly in the legs, autonomic dysfunction like dry mouth, and areflexia. Because of its rarity, national data describing its outcomes are limited. Data from the 2022 Healthcare Cost and Utilization Program- National Inpatient Sample were analyzed for cases of Lambert Eaton Myasthenic Syndrome. Within the cases found, the presence of hypertension, obesity, and neoplasm status was distinguished in addition to race/ethnicity, sex, and hospital outcome. 705 cases were identified. 9.9% were associated with neoplasm. 8.5% were obese, 63.1% had hypertension. 8.5% of patients died during hospitalization. The results from this study are useful contextualizing the prevalence of comorbidities in those diagnosed with LEMS as well as providing insight into outcomes associated with LEMS.

Extracellular vesicles (EVs) are nanoscale vesicles, containing proteins, lipids and nucleic acids, released for cellular communication and environmental modulation. Cancer cells utilize EVs to facilitate tumor survival, including interference with anti-tumor immune responses. Hepatocellular carcinoma (HCC) is the fifth most common cancer worldwide and is notable for generating a highly immunosuppressive tumor microenvironment. Normally, EVs from HCC would cause treated cells to display an immunosuppressed phenotype. To address this immunosuppression, we treated HCC derived EVs with melittin, a peptide component of bee venom significant for its ability to induce proinflammatory macrophage immune signaling. Engineered melittin containing nanoparticles have been developed previously, however, cell derived EVs containing melittin are a novel treatment for cancer. Hypothesis: Melittin treated HepG2 EVs modulate inflammatory changes in human THP-1 monocytes. Methods: HepG2 EVs were isolated via differential ultracentrifugation. Melittin treated EVs were loaded by incubation with melittin in conditioned media at a concentration of 50 μg/mL melittin and allowed to rest at room temperature for 1 hour before washing away unincorporated melittin, and isolating the melittinized EVs at 110,000xg by ultracentrifugation. Zetaview nanoparticle tracking analysis (NTA) was used to measure EV size and zeta potential of melittin loaded and unmodified EVs. RNA isolated from THP-1 monocytes treated with melittinized HepG2 EVs, control HepG2 EVs, or untreated cells was prepared for RT-qPCR. Human Cancer Inflammation and Immunity Crosstalk arrays were used to detect differences in the upregulation of immune relevant mRNAs of EV treated THP-1 monocytes. Results: Melittinized HepG2 EVs had a median size of 276.9 +/- 186.9 nm and unmodified HepG2 EVs had a median size of 281.7 +/- 180.5 nm. Melittinized HepG2 EVs were stable and intact according to NTA. THP-1 cells treated with HepG2 control vesicles resulted in a greater than 3-fold upregulation of C-C chemokine receptor type 7 (CCR7) mRNA. No upregulation in CCR7 was observed using melittinized HepG2 EVs. Conclusions: CCR7 is involved in tumor immune suppression. Therefore, melittin's ability to inhibit HepG2 EV mediated induction of CCR7 in THP-1 monocytes, when loaded into the HepG2 EVs, supports further evaluation of melittin's ability to abrogate tumor EV mediated immunosuppressive properties for HCC and other cancer types.