2:00 PM
|
The Research Quality Index (RQI): A New Method to Analyze Plastic Surgery Residency Applications
Purpose
In March 2024, the first class of integrated plastic surgery (IPS) applicants matched without numeric Step 1 scores. In a recent survey of IPS program directors (PD), 68.1% agreed that it was more difficult to differentiate candidates without these numeric scores (1). This same study showed an applicant's scholarly activity was consistently ranked as a top 3 most important metric to evaluate applicants (1). As expected, the research production of successfully matched applicants this past year has risen significantly (2). Concurrently, there are concerns regarding the incentives for medical students to avoid time-consuming, high-quality research in favor of producing high quantities of publications. These efforts have been termed "research pollution" and are seen as a way to game the application system (3). These concerns underscore a need to develop an objective metric in which PDs can account for both the quantity and quality of an applicant's scholarly work. Herein, we set forth to design a novel method to evaluate the publications of applicants while testing this new metric on the three most recently matched IPS residency classes.
Methods & Materials
First, we developed a new metric to analyze applicant publications that we call the "Research Quality Index" (RQI). During development, there were three factors we felt necessary to balance: publication quantity, quality, and student involvement in project. As such, we created the following equation to determine an applicant's RQI: RQI = (Total Publications X 2) + 1st Author Publications + (2nd Author Publications X 0.5) + Average Impact Factor + Weighted Relative Citation Ratio (RCR). Weighted RCR, an aggregate measure used to assess the influence of a scientific article, was collected through the NIH's iCite webpage (4).
To give PDs considering the use of this metric a relative idea as to the average RQI of an applicant, we collected the necessary data for the last three IPS residency classes and then conducted a one-way ANOVA to compare the mean RQI of each class (p=0.05).
Results
Pertinent data on over 4,000 articles authored by 503 applicants was collected. RQI's for each applicant were subsequently calculated. A yearly increase in average RQI was observed, with the most recently matched class recording an average RQI of 31.4. One-way ANOVA testing resulted in a p-value of 0.037 which revealed a significant difference in the mean RQIs. Of note, the average total number of publications across each class remained relatively unchanged (~7) indicating the rise in RQI was due to increases in factors related to authorship involvement and article quality.
Conclusion
This study demonstrates the Step 1 score change has not necessarily led to the adoption of a high-quantity, low-quality publication strategy for successful IPS applicants. Instead, the quality of and involvement in publications has increased, indicating successfully matched applicants are still participating in largely high-quality research. Further given the dynamic nature of the RQI, programs and even other specialties, can adjust the variables contained within this formula to best suite the priorities of their program.
References
Boyd LR, Lin LO, Janis JE. Matching Into Integrated Plastic Surgery: The Impact of USMLE Step 1 Transition to Pass/Fail: Did the 2019 Predictions Come True in 2024?. Plast Reconstr Surg Glob Open. 2025;13(1):e6417. Published 2025 Jan 16. doi:10.1097/GOX.0000000000006417
Charting OutcomesTM: Characteristics of U.S. MD Seniors Who Matched to Their Preferred Specialty: 2024 Main Residency Match®. NRMP. August 20, 2024. Accessed September 22, 2024. https://www.nrmp.org/match-data/2024/08/charting-outcomes-characteristics-of-u-s-md-seniors-who-matched-to-their-preferred-specialty-2024-main-residency-match/
Elliott B, Carmody JB. Publish or Perish: The Research Arms Race in Residency Selection. J Grad Med Educ. 2023;15(5):524-527. doi:10.4300/JGME-D-23-00262.1
iCite | New Analysis | NIH Office of Portfolio Analysis. Accessed February 25, 2025. https://icite.od.nih.gov/analysis
|
2:05 PM
|
Exploring the Scope of TriNetX in Plastic Surgery: Large-Scale Insights into Aesthetic and Reconstructive Procedures
Background: Large, multi-institutional databases have increasingly enabled robust and efficient research in plastic surgery by aggregating real-world patient data. TriNetX stands out as a powerful data and analytics platform that centralizes de-identified health records from numerous healthcare systems. This global federated real-world network allows investigators to perform comprehensive outcomes research with advanced statistical capabilities such as propensity score matching and logistic regression. Previous TriNetX-based studies in plastic surgery have compared surgical site occurrences in different breast reconstruction techniques, examined the impact of COVID-19 on procedural volumes, and investigated risk factors such as obesity for postoperative complications. These examples underscore the platform's versatility for analyzing trends, patient characteristics, and outcomes across diverse populations. This study aims to demonstrate the utility of TriNetX in plastic surgery research by analyzing sample sizes for common procedures and highlighting its potential for advanced statistical analyses.
Methods: We queried the TriNetX Global Collaborative Network to identify patient samples associated with several frequently performed plastic surgery procedures, spanning both aesthetic and reconstructive domains. Our primary goal was to assess the magnitude of available datasets rather than to perform an in-depth clinical comparison. We also reviewed TriNetX's in-platform analytical features-including propensity score matching, logistic regression, and time-to-event analysis-to evaluate their applicability in plastic surgery outcomes research.
Results: TriNetX queries revealed substantial cohorts, including over 77,000 patients for breast reduction and more than 20,000 for breast augmentation. Blepharoplasty and liposuction also had large sample sizes, reflecting the platform's breadth. A total of 233 TriNetX-related plastic surgery publications were found on Google Scholar and 34 on PubMed. Google Scholar listings rose from 8 in 2020 to 113 in 2024, while PubMed results increased from 3 in 2021 to 20 in 2024-likely due to Google Scholar's broader indexing criteria. These trends underscore TriNetX's growing prominence in the field, and a systematic review is pending to further evaluate the scope and impact of these studies.
Conclusions: TriNetX offers an unprecedented opportunity for large-scale, evidence-based research in plastic surgery by integrating diverse clinical data and advanced analytics in a single platform. Researchers can use TriNetX to efficiently examine trends in procedural volume, explore patient demographics and comorbidities, and compare techniques using robust propensity score methods. Future investigations leveraging TriNetX are poised to advance best practices in plastic surgery through improved identification of risk factors, evaluation of novel interventions, and ongoing assessment of surgical outcomes. By capitalizing on the platform's wide-ranging data and real-time analytic features, plastic surgeons and researchers can enhance patient care, refine clinical guidelines, and drive innovation throughout the specialty.
|
2:10 PM
|
A Novel Microsurgery-Specific Artificial Intelligence Large Language Model: MicrosurgeryLlama2
Background: Large language models (LLMs) like ChatGPT have demonstrated remarkable capabilities in processing vast datasets and generating coherent text. However, their application in specialized medical fields remains limited due to a lack of domain-specific knowledge. Aim of this study is to develop a novel domain-specific LLM for microsurgery, addressing the need for AI-driven tools in surgical education and clinical decision support.
Methods: To develop a domain-specific LLM for microsurgery, we curated a corpus of 9,086 microsurgery research abstracts from PubMed, spanning from 2010 to 2024. The open-source Llama-2-13b model served as our foundation. We first trained the model on the curated research abstracts dataset, then fine-tuned it using PyTorch and HuggingFace frameworks. We used Parameter-Efficient Fine-Tuning (PEFT) with Low-Rank Adaptation (LoRA) to enhance model adaptability and reduce computational overhead.
Results: The resulting model, named MicrosurgeryLlama2 demonstrated superior performance compared to the base Llama-2 model across multiple metrics. We observed improvements in BLEU (0.0371 vs. 0.0209), METEOR (0.2249 vs. 0.1168), ROUGE-1 (0.2459 vs. 0.2280), and ROUGE-L (0.1806 vs. 0.1740) scores. These enhancements indicate an improved capability in generating domain-specific, coherent text relevant to microsurgery.
Conclusion: This study introduces MicrosurgeryLlama2 which represents a significant advancement in applying AI to microsurgery education and practice. By leveraging a specialized corpus and advanced fine-tuning techniques, our model outperforms the native LLM, Llama2 in generating relevant, accurate content for the field. This innovation paves the way for AI-assisted clinical decision support and enhanced research capabilities in microsurgery.
|
2:15 PM
|
Impact of Research Publications on Match Success in the Post-Step 1 Pass / Fail Era: A Comparative Analysis of 2022 and 2024 NRMP Data
Background: USMLE Step 1 transitioned to pass/fail scoring in January 2022, eliminating a critical metric in residency selection. This study analyzes how research publications impact match success by comparing plastic surgery match data before and after this transition, focusing on the interplay with Step 2 CK scores. The 2024 match cycle represents the first cohort predominantly experiencing pass/fail Step 1 throughout their application.
Methods: Data from NRMP's Charting Outcomes™ report for 2022 (n=158) and 2024 (n=152) match cycles was analyzed for US MD seniors applying to integrated plastic surgery (1). We stratified by Step 2 CK scores and publication counts, then consolidated into broader categories (0-15 vs. ≥16 publications) for statistical analysis using Fisher's exact test, calculating odds ratios (OR) with 95% confidence intervals.
Results: In 2022, applicants with ≥16 publications had significantly higher match rates than those with 0-15 publications (74.4% vs. 48.7%, p=0.001, OR 3.04, 95% CI 1.49-6.33) across all Step 2 ranges. Within the Step 2 ≥250 subgroup, publications remained a significant predictor of match success (p=0.010, OR 3.19).
In 2024, overall match rates increased (80.2% vs. 61.3% for ≥16 vs. 0-15 publications), but statistical significance weakened (p=0.034) with confidence interval crossing 1.0 (95% CI 0.98-6.25). However, within the Step 2 240-249 range, publications had a pronounced effect (p=0.012, OR 16.10), indicating Step 2 performance became a more important determinant of outcomes.
Analysis of publication thresholds revealed a shift in optimal numbers. In 2022, match rates for Step 2 ≥250 peaked at 16-20 publications (85.7%) with no further benefit. By 2024, this point shifted upward: for Step 2 ≥250, match rates increased from 68.8% (16-20 publications) to 87.5% (21-25 publications), plateauing thereafter (86.2% with 26+). The 240-249 range showed a similar pattern.
Conclusions: As the first analysis of match outcomes in the post-Step 1 pass/fail era, this study provides valuable guidance for programs and applicants navigating this significant transition.
The shift to pass/fail Step 1 has altered how research influences match outcomes. In 2022, publications were broadly important regardless of Step 2 scores. By 2024, their impact became stratified-serving as a critical differentiator for mid-tier Step 2 scores but with diminishing importance for exceptional scores.
Our data indicates the point of diminishing returns for publications has shifted from approximately 16-20 publications in 2022 to 21-25 publications in 2024 (2).
These findings suggest three key implications: First, Step 2 CK has emerged as a crucial metric that programs rely on in the absence of numeric Step 1 scores. Second, substantial research provides significant advantage, especially for candidates with mid-tier Step 2 scores. Third, the data suggests potentially enhanced outcomes with increasing publications up to the 21-25 range, though individual circumstances may vary.
References:
1. NRMP. Charting Outcomes™: USMLE Step 2 CK Exam Baseline. 2020-2024. Available at https://www.nrmp.org/match-data/2024/08/charting-outcomes-usmle-step-2-ck-exam-baseline/. Accessed December 1, 2024.
2. Elemosho A, Sarac BA, Janis JE. The Law of Diminishing Returns in the Integrated Plastic Surgery Residency Match: A Deeper Look at the Numbers. Plast Reconstr Surg Glob Open. 2024;12(7):e5937. Published 2024 Jul 3. doi:10.1097/GOX.0000000000005937
|
2:20 PM
|
Metformin Topical Ointment Reduces Acute Damage in Irradiated Human Skin by Blocking TGFβ Signaling Through Modulation of AMPK and mTOR
INTRODUCTION: Radiation therapy has a pivotal role in modern cancer treatment, benefiting nearly half of all patients. Despite efforts to minimize its impact on healthy tissues, radiation-induced damage remains a significant concern. Skin, as the first-line defense, is particularly vulnerable to radiation injury. Acute radiation exposure can result in radiodermatitis, and prolonged exposure can lead to radiation-induced skin fibrosis (RISF), leading to functional and anatomical impairments. Ionizing radiation promotes fibrosis through the generation of reactive oxygen species (ROS), TGF-ß and mTOR signaling pathways. Metformin shows promise in mitigating fibrosis by targeting crucial proteins involved in these pathways, ultimately downregulating the fibrotic process. Previous research has explored the use of intraperitoneal metformin showcasing its potential to alleviate skin fibrosis in mice. Once established the potential of metformin as a therapeutic agent for RISF, we conducted a protocol to assess its capability to mitigate fibrosis on a human skin model.
METHODS: We devised a controlled experimental protocol using our ex vivo human skin perfusion system, where a human tissue was subjected to radiation-induced damage at a dosage of 15 Gy. The experimental intervention involved the topical application of metformin cream, control cream and no intervention. Skin biopsies were scheduled at specific time points to assess the histological and molecular changes.
RESULTS: Therapeutic metformin showed promising results in mitigating radiation damage to epithelial cells. Histological sections from the metformin-treated group consistently showed reduced cell damage, fibrosis, and epidermal-dermal disruption, when compared to the non-treated groups. Additionally, in the metformin-treated group, TUNEL results demonstrated the lowest percentage of cell death and γ-H2AX results showcased lower levels of DNA damage, compared to the other two groups. We observed increased elastogenesis in metformin treated skin. When looking at the gene expression, metformin treated skin showed a higher expression of: ECM remodeling, antioxidant, DNA repair, cellular proliferation and antiapoptotic markers, with a lessened expression of proinflammatory markers. We were able to observe metformin's capability to downregulate pro-fibrotic pathways, as showcased by a reduced activity of the proteins mTOR, p70S6K and TGF-ß1, as well as an increased activity of AMPK.
CONCLUSION: The results demonstrated that metformin decreases the activity of proteins involved in the oxidative stress and metabolic reprogramming pathways that drive fibroblast activation. By doing this, metformin helps maintain the structural integrity of skin after radiation injury and enhances cellular resilience under stress. Metformin also has a protective effect against DNA damage and enhances DNA repair mechanisms. These findings indicate that metformin's properties make it a promising therapeutic agent to mitigate radiation-induced damage and to enhance tissue repair and regeneration. This research has important clinical implications for preventing cutaneous functional and anatomical impairments in patients receiving radiation therapy and improving the quality of life of those affected by radiation-induced skin fibrosis.
|
2:25 PM
|
Purified Exosome Product (PEP) Enhances Skin Graft Donor Site Healing: A Phase 1b Clinical Trial
INTRODUCTION
Exosomes represent the latest cell-free strategy in wound care. However, a readily available clinical-grade exosome product is lacking. Purified Exosome Product (PEP) is a novel, off-the-shelf, clinical-grade exosome product derived from human platelets, with consistent batch quality and a shelf life of 24 months at room temperature. The objective of this study was to determine the safety of the first clinically used PEP in the United States for open wounds.
METHODS AND MATERIALS
This was a phase 1b, prospective, open-label, controlled, randomized, evaluator-blinded clinical trial of PEP in subjects with at least two equal sized split-thickness skin graft (STSG) donor site wounds with one donor site treated with standard of care (SOC) for wounds and the other treated with PEPs. Patients were then divided into low-dose PEP-only group, high-dose PEP-only group and high-dose PEP-TISSEEL group. TISSEEL is a commercially available fibrin based tissue sealant and served as a carrier for PEP in this study. Safety evaluations included occurrence of adverse events, laboratory values, vital signs, and physical examination findings. Wound healing was evaluated using the Vancouver Scar Scale (VSS) and the Photographic Wound Assessment Tool (PWAT) by a board-certified plastic surgeon and a board-certified dermatologist together. The trial is registered at ClinicalTrials.gov (NCT04664738).
RESULTS
Seven subjects were enrolled in the study, with a mean age of 49.3 years (range, 19 to 79). Three male patients and four female patients were included. There were no adverse events related to the treatments. By 4 weeks after treatment, all wounds reached 50% re-epithelialization and 6 (85.7%) of 7 subjects reached 100% re-epithelialization. The application of PEP significantly accelerated wound healing, with the PEP-treated sites reaching 100% re-epithelialization in an average of 21.8 days, compared to 60.6 days for the SOC sites (P < 0.05). The median VSS score was the lowest in the high-dose PEP-only group after 6 months of PEP application across all groups. The total PWAT score was significantly reduced in the high-dose PEP-only group compared to the low-dose PEP-only group at 3 months and 6 months after PEP administration (P<0.05). All donor site wounds in the high-dose PEP-only group, including those treated with PEP and those receiving SOC, were closed with intact overlying skin indicated by PWAT=0.
CONCLUSION
PEP is a clinical-grade exosome product that can benefit wound healing with immediate clinical translation. The results of this phase 1b trial showed that a single application of PEP at a dose up to 20% (2 vials) was safe and support additional clinical investigation of PEP in wound management. Wound healing was significantly accelerated in the PEP-treated wounds compared to SOC, with a dose-response effect. Guided by the FDA, this study delivers exosome-based therapies to patients to achieve advancements in wound management safely, conveniently and efficiently. Further clinical efficacy studies are warranted.
|
2:30 PM
|
Integrating Photogrammetry with Mixed Reality Platforms for Intraoperative Feedback during Cranial Vault Remodeling
Premature fusion of cranial sutures results in cranial and skullbase deformity. Surgical intervention for craniosynostosis simultaneously improves cosmetic development and reduces intracranial pressure. This is accomplished with fronto-orbital advancement for metopic and unilateral coronal synostoses. Essential intraoperative feedback on vault remodeling is currently limited to qualitative judgement.
An iPhone 15 acquired simultaneous video and LiDAR for 60 seconds after bony exposure and after remodeling. Scale-invariant features were extracted and matched between frames, enabling camera pose estimation and dense point cloud reconstruction of scene geometry (A). The pre- and post-remodeling point clouds were aligned using M-estimator sample consensus of fast point feature histogram descriptors, which identified and registered the intact biparietal convexity (C). These were similarly aligned to a skull mesh derived from the preoperative CT.
Intraoperative imaging was completed in 11 patients. Absolute error (mean 2.3mm) was measured between landmarks on the preoperative CT mesh and the pre-remodeling point cloud (D). Relative error (mean 1.7%) was measured using the affine scaling factor between pre- and post-remodeling point clouds. Cranial vault volume was calculated above a plane traversing the nasion and bilateral zygomatic roots (mean +90cc). Cranial vault symmetry highlighted areas requiring correction (E). These masks were projected onto the surgical scene through a mixed reality platform, enabled by the co-registration to the preoperative CT.
Pairing widely available camera technology with this fast and robust algorithmic pipeline delivers timely intraoperative feedback to guide the definitive correction of cranial vault deformity without radiation.
|
2:35 PM
|
Scientific Abstract Presentations: Research & Technology Session 4 - Discussion 1
|
2:45 PM
|
Can Artificial Intelligence Offload Some of the Faculty Burden of Resident Education?
Introduction
Artificial intelligence (AI) has been touted as a vehicle to improve resident education and lessen the strain of the teaching responsibilities of faculty. A study was designed to test the hypothesis by a challenge of an AI platform to generate faculty discussion guides on three distinct topics: breast reduction, frontal sinus fractures, flexor tendon.
Methods
A bibliography of readings was created and a faculty discussion guide composed from the readings. Both were vetted by an additional faculty member and senior resident for completeness of the bibliography and validity and comprehensiveness of the discussion guide. The same investigators developed a series of cardinal teaching points drawn from the composed faculty discussion guides.
The readings were subsequently programmed into the platform, Chatly®, and challenged to generate a faculty discussion guide for each of the three topics for application in a seminar setting. Repetitive iterative prompts demanded refinement of the output until the responses flattened in quality. Six residents, PGY1-6, compared the cardinal teaching points from the faculty-generated and AI-generated discussion guides for completeness, specificity, and absence of hallucinations on a scale of 1 (poor) to 10 (superb).
Results
The intra-class correlation coefficient for the refereed cardinal points for each of the three topics was 0.85 (high concordance). The composite resident assessment mean ranged from breast reduction (4.39), frontal sinus fractures (3.39) and flexor tendon repair (2.67). The curves are presented in Figure 1.
Conclusions
As complexity of topic increased from breast reduction to frontal sinus fractures to flexor tendon repair, AI performance deteriorated from distinctly mediocre to poor, indicating AI at present lacks the sophistication to facilitate resident education.
|
2:50 PM
|
Evolution and Efficacy of Virtual Training in Plastic Surgery Education in LMICs
Background: Limited access to specialized surgical care in Low- and Middle-Income Countries (LMICs) contributes to high morbidity and mortality from trauma, burns, and congenital anomalies-conditions frequently treated by plastic surgeons. Short-term surgical missions, while beneficial, are unsustainable due to high costs, lack of long-term integration, and minimal training for local providers. In recent years, virtual training has gained traction as a scalable approach to expanding plastic surgery education in LMICs and reducing reliance on external missions. This systematic review evaluates the evolving impact of virtual plastic surgery training in LMICs.
Methods: A PRISMA-guided systematic review searched PubMed, Embase, Scopus, and Cochrane Library (2009–2025) using terms like "virtual plastic surgery training," "tele-mentoring," and "LMICs." Studies assessing digital or remote surgical training with measurable outcomes in LMICs were included, while those exclusive to high-income countries or lacking defined interventions or outcomes were excluded. Data extraction focused on study design, country, training methods, effectiveness, and challenges.
Results: Nineteen studies from multiple LMICs were included. The most common training modalities were tele-mentoring/video conferencing (26.3%), digital simulation (26.3%), augmented/virtual reality (15.8%), mobile/social media-based education (15.8%), and hybrid models (15.8%). Virtual training demonstrated strong educational benefits, with 89.5% of studies reporting improved surgical knowledge, skills, or confidence, and 73.7% showing high satisfaction. However, 10.5% favored in-person training, particularly for skill-intensive procedures, with one study finding a significant preference for in-person engagement (p=0.04). Tele-mentoring and digital simulation had the highest satisfaction rates (84.5% ± 4.94). Virtual training significantly expanded access to education, reaching up to 148 countries-far exceeding the geographic limitations of in-person training. However, challenges persisted, with 94.7% of studies citing internet connectivity as a major barrier. Additional obstacles included high equipment/setup costs (42.1%), limited hands-on training (36.8%), ethical concerns (15.8%), and logistical constraints (15.8%). While 42.1% of studies were multi-country, only 21.1% examined regional differences in usage, and none directly compared training outcomes across regions. A majority of studies (57.9%) included Sub-Saharan/COSECSA region countries, reflecting a strong focus on this area for virtual plastic surgery training initiatives.
Conclusion: Virtual training has enhanced accessibility and skill development in plastic surgery education in LMICs where training is limited. Despite years of virtual outreach, internet connectivity remains the most persistent barrier pointing towards unresolved infrastructure challenges. The scalability of virtual training is one of its greatest strengths, allowing for broader global reach and enabling training for more individuals who would otherwise lack access. Still, virtual training alone does not fully replace the benefits of hands-on learning, suggesting a role for hybrid models that integrate virtual education for theoretical learning while reserving in-person training for skill-intensive procedures. Future research should assess whether certain virtual training approaches-such as AR versus simulation-yield better outcomes in specific regions or if effectiveness is consistent across settings. Additionally, longitudinal studies evaluating the impact of virtual training on reducing workforce gaps in LMICs are needed to determine its long-term sustainability. Virtual training is a transformative tool in surgical education– optimizing its role within hybrid models may provide the most effective and sustainable solution for LMICs.
|
2:55 PM
|
Burn Depth Diagnostic System for Healthcare Using Artificial Intelligence
Background. Burn injuries are complex and challenging in both diagnosis and management. In the U.S., 1.25 million individuals need medical attention for burns annually, with 40,000 hospitalized. Early and accurate assessment of burn depth is crucial to determine the need for surgery but challenging. Visually assessing burn depth is complex, with experienced surgeons achieving only 50-75% diagnostic accuracy. Burn depth is considered a predictor of pathological scarring that occurs in 30%-91% of burn injuries. Objective and non-invasive approaches are needed. This study presents an Artificial Intelligence (AI) system leveraging large-scale pretraining for determining burn depth using both pre-clinical and human burn dataset including non-invasive Tissue Doppler Elastography Imaging (TDI), Harmonic B-mode ultrasound and digital wound photography.
Methods. Different burn degrees were induced on the back of twelve anesthetized Yorkshire pigs (70-80 lbs.) using a standardized electrically heated burn device per IACUC protocol. Burns were created with a 150°C block for 1,10, and 60 sec, resulting in six (2" x 2") wounds per pig, with two unburnt control sites. Human data (n=30) was collected from patients treated at American Burn Association (ABA)-verified burn center from May 2022 to May 2023 per IRP protocol. Eligible subjects had thermal burn injury within 72 hours after burn, without prior surgical debridement or intervention, male or female age 18-89 years old, and Total Body Surface Area (TBSA) involvement ≤ 75%. Chemical, electrical or radiation burns were not included. TDI to measure tissue stiffness, B-mode images, and digital photos were collected for 42 days. Biopsies were obtained from pigs and from the human subjects during debridement at OR. The AI model was trained on the pig data and validated with human data.
Results. Thirty patients were enrolled with one withdrawal. The age was 47.6± 17.6 years old, BMI= 28.8 ± 5.3 kg/m². Our model demonstrated high accuracy in detecting surgical cases and this was validated by histology. The model identified 100% of surgical cases (third-degree burns) for pig subjects. On human subjects, the model demonstrated high accuracy, it identified 7 out of 7 surgical cases (100%) and 15 out of 16 cases (94%) for non-surgical cases. The precision was 87%, sensitivity 100%, and an F1-score of 93%. The AUROC was 0.97, with a 95% Confidence Interval (CI) from 0.89 to 1.0. Critically, our system presented a clear and human-readable output to understand the surface of burn wounds, allowing a high degree of explainability often required to interpret AI-produced results.
Conclusion. This work is the first to report a burn diagnostic system using non-invasive ultrasound TDI images and AI. Tissue stiffness measurements by TDI overcome the limitations of light-based technologies. Integrating AI to assist in TDI images interpretation results in a high degree of accuracy in predicting burn depth. Accurate early diagnosis of burn depth is expected to improve overall patient outcomes.
|
3:00 PM
|
Controlled Super-Zero Cooling After Volumetric Muscle Loss Injury Improves Functional Recovery in a Large Animal Model.
Purpose: Over 40,000 traumatic open fractures leading to volumetric muscle loss (VML) injuries are estimated to occur annually in the US (1,2). Even ~20% muscle loss causes disproportionate lasting strength deficits of 30 - 90% (3). The current gold standard, physical therapy, only achieves partial functional recovery of 17 - 58% (4, 5). Our prior murine model of VML showed that optimized post-injury treatment with controlled cooling fluid (CCF) above freezing improved recovery of baseline maximal muscle torque (MMT; +30%) and resistance to fatigue (RF; +66%). We hypothesized CCF would have similar results in a large animal model of VML.
Methods: Left Peroneus Tertius muscle (PTM) VML injury was surgically induced in female (38-42 kg) Yucatan pigs; animals received no treatment (n=10) or CCF of the PTM (n=4). Outcomes included post-injury day 35 (PID35) recovery of pre-injury MMT, baseline activity levels (step count), and baseline RF. PID35 strength in newton-centimeters per gram of PTM (ncm/g) and weekly gait assessments (Tarlov score) were also recorded.
Results: Compared to controls, CCF-treated animals had significantly improved recovery of pre-injury MMT (40±27% vs 18±14%; p=0.00012) and RF (37±26% vs 26±31%; p=0.0013) with nonsignificantly improved recovery of baseline activity level (127±194% vs 67±62%; p=0.17). CCF also significantly improved PID 35 PTM strength per gram of PTM (3.4±3.6 ncm/g vs 1.5±1.4 ncm/g; p=0.016) and did not significantly modify gait.
Conclusions: In a translational large animal model, CCF seems to promote functional recovery. These results should be validated in a larger cohort of animals with molecular analyses before a pilot clinical trial.
References:
1. Court-Brown CM, Bugler KE, Clement ND, Duckworth AD, McQueen MM. The epidemiology of open fractures in adults. A 15-year review. Injury. 2012 Jun;43(6):891-7. doi: 10.1016/j.injury.2011.12.007. Epub 2011 Dec 27. PMID: 22204774.
2. Population clock. https://www.census.gov/popclock/.
3. Garg K, Ward CL, Hurtgen BJ, Wilken JM, Stinner DJ, Wenke JC, Owens JG, Corona BT. Volumetric muscle loss: persistent functional deficits beyond frank loss of tissue. J Orthop Res. 2015 Jan;33(1):40-6. doi: 10.1002/jor.22730. Epub 2014 Sep 18. PMID: 25231205.
4. Corona, B.T. et al. (2013) 'Autologous minced muscle grafts: A tissue engineering therapy for the volumetric loss of skeletal muscle', American Journal of Physiology-Cell Physiology, 305(7). doi:10.1152/ajpcell.00189.2013.
5. Aurora, A. et al. (2014) 'Physical rehabilitation improves muscle function following volumetric muscle loss injury', BMC Sports Science, Medicine and Rehabilitation, 6(1). doi:10.1186/2052-1847-6-41.
|
3:05 PM
|
Revolutionizing Aesthetic Documentation: AI-Powered Photo Standardization for Optimizing Clinical Outcomes in Plastic Surgery
Background
High-quality, standardized before-and-after photography is essential in plastic surgery, directly impacting clinical decision-making, patient expectations, and outcome assessment and communication. Despite its importance, significant variability in lighting, positioning, and framing leads to documentation inconsistencies that compromise reliable progress tracking and limit objective outcome evaluation. This study presents ClinicOS, a novel AI-powered imaging platform that leverages computer vision and deep learning to standardize pre- and post-operative photography, ensuring reproducibility while enhancing clinical accuracy across diverse patient populations.
Methods
We developed and validated an AI-driven photo capture and management platform that integrates multiple machine-learning models to standardize patient photography. The system includes:
AI-Powered Image Alignment – Convolutional neural networks automatically detect anatomical landmarks, patient posture, and camera angles across all skin tones (Fitzpatrick I-VI), ensuring precise before-and-after comparisons regardless of patient demographics.
Intelligent Pose Guidance – Real-time feedback through augmented reality overlays guides both patients and clinical staff in capturing consistent photos, reducing human error and improving inter-operator reliability.
Automated Feature Recognition – A deep learning segmentation model (U-Net) identifies and selectively blurs distinguishing features such as tattoos and faces while preserving surgical landmarks. This ensures HIPAA compliance and patient privacy by anonymizing sensitive regions while maintaining critical medical details.
Workflow Integration with Outcomes Analysis – The software seamlessly integrates with EMRs, enabling automated photo organization, metadata tagging, and correlation with surgical outcomes data.
Results
In a prospective multi-center study across 12 plastic surgery practices (n=487 patients), our AI system demonstrated significant improvements in documentation quality and efficiency. Key findings include:
86.4% improvement in photo consistency metrics (p<0.001) compared to standard photography protocols
63.7% reduction in time spent on image capture and organization (average 5.8 minutes saved per patient)
92.3% accuracy in automated anatomical landmark detection
Conclusion
AI-driven photo standardization represents a significant advancement in plastic surgery documentation and outcomes assessment. By ensuring precision, efficiency, and consistent quality across diverse patients, ClinicOS empowers surgeons with reliable visual data for clinical decision-making while enhancing patient understanding and engagement. This technology not only improves current documentation standards but also establishes a foundation for data-driven, objective evaluation of aesthetic outcomes.
Significance
This work introduces a clinically-validated AI approach to plastic surgery imaging that addresses several challenges in aesthetic documentation: standardization across diverse patient populations, objective outcomes assessment, and workflow efficiency. Future applications include AI-powered predictive outcome visualization, automated surgical planning, quantitative scar analysis, and longitudinal outcome tracking to support evidence-based practice in aesthetic surgery.
|
3:10 PM
|
Addressing Nerve Size Mismatch in Targeted Muscle Reinnervation: Does Size Match Make a Difference for Analgesia?
Introduction
Targeted Muscle Reinnervation (TMR) is a surgical technique designed to enhance prosthetic control and alleviate post-amputation pain. A key but understudied feature of TMR is the size mismatch between
large, mixed motor-sensory nerves and relatively smaller motor nerve branches. Despite clinical evidence showing TMR's efficacy in pain reduction even with this mismatch, it is unknown if a better size match
would improve analgesia outcomes This study investigates how varying levels of nerve size discrepancy influence spontaneous and reflexive pain behaviors following tibial nerve injury (TNI).
Study Methods
A total of 24 male Sprague-Dawley rats were randomly assigned into three cohorts: (1) TNI-only, (2) TMR with the tibial nerve coapted to the motor branch to semimembranosus (large size mismatch, TMRsm),
and (3) TMR with the tibial nerve coapted to a near size-matched configuration (branches to semimembranosus, biceps femoris long head, biceps femoris short head; Multi-TMR). Behavioral
assessments were conducted at baseline, 2, 4, 6, 8, 10, 12, and 14 weeks, including von Frey (mechanical hypersensitivity), acetone (cold hypersensitivity), pin test (hyperalgesia), brush testing (allodynia),
guarding (spontaneous pain), and dynamic weight bearing. Anxiety and exploratory behaviors were evaluated at week 12 via the open field maze and elevated plus maze.
Results
In this all-male study, both TMR-sm and Multi-TMR significantly reduced mechanical (Von Frey), cold (acetone), brush, and pin hypersensitivity compared to TNI-only rats after the first two weeks postsurgery (p < 0.05). However, Multi-TMR and TMR-sm diverged in brush-evoked responses at weeks two (p = 0.003) and four (p = 0.025), with Multi-TMR showing higher noxious responses-differences that resolved by week six. No group experienced actual weight loss; nonetheless, TMR-sm weighed consistently less than Multi-TMR starting at week two (p < 0.05).
Conclusion
This study provides the first in vivo evaluation of nerve size mismatch in TMR, revealing that TMR sizematched procedures do not dramatically alter spontaneous or reflexive pain behaviors compared to TNI alone-although some variation emerged among TMR groups. Further analysis of axonal counts and neuroma formation will clarify these differences. Although counterintuitive to nerve repair dogma, this
study did not identify any advantage to improved size match.
References
- Roth E, Linehan A, Weihrauch D, Stucky C, Hogan Q, Hoben G. Targeted muscle reinnervation prevents and reverses rat pain behaviors after nerve transection. Pain. 2023;164(2):316 -324.
doi:10.1097/j.pain.0000000000002702
- Tanner N, Ayalon O. The neurobiology of targeted muscle reinnervation for post-amputation pain. Plast Aesthet Res. 2023;10:11. doi:10.20517/2347-9264.2022.95
- Fox IK, Brenner MJ, Johnson PJ, Hunter DA, Mackinnon SE. Axonal regeneration and motor neuron survival after microsurgical nerve reconstruction. Microsurgery. 2012;32(7):552-562. doi:10.1002/micr.22036
- Eberlin KR, Brown DA, Gaston RG, et al. A consensus approach for targeted muscle reinnervation in amputees. Plast Reconstr Surg Glob Open. 2023;11(4):e4928. doi:10.1097/GOX.0000000000004928
- Unal-Cevik I, Oaklander AL. Comparing partial and total tibial-nerve axotomy: long-term effects on prevalence and location of evoked pain behaviors. Pain Pract. 2011;11(2):109 -119. doi:10.1111/j.1533-2500.2010.00429.x
- Oliveira KMC, Pindur L, Han Z, Bhavsar MB, Barker JH, Leppik L. Time course of traumatic neuroma development. PLoS One. 2018;13(7):e0200548. Published 2018 Jul 16. doi:10.1371/journal.pone.0200548
- Berberoglu I, Sabbagh SW, Cederna PS, Kemp SWP. A novel animal model of symptomatic neuroma for assessing neuropathic pain. Neurosci Lett. 2024;836:137896. doi:10.1016/j.neulet.2024.137896
- Dehdashtian A, Timek JH, Svientek SR, et al. Sexually dimorphic pattern of pain mitigation following prophylactic regenerative peripheral nerve interface (RPNI) in a rat neuroma model.
Neurosurgery. 2023;93(5):1192-1201. doi:10.1227/neu.0000000000002548
|
3:15 PM
|
Hemoglobin A1C as a Predictor of Cancer-Related Lymphedema: A Laboratory-Based Predictive Model
Purpose: Lymphedema is a common and debilitating complication affecting 20-40% of breast cancer patients undergoing axillary lymph node dissection (ALND), with current interventions being largely palliative, relying on lifelong physiotherapy and compression. Despite evidence linking inflammation and insulin resistance to lymphatic dysfunction, the role of metabolic markers in predicting which patients will develop lymphedema remains unclear. This decade-long study aimed to identify laboratory markers associated with lymphedema development in patients undergoing ALND.
Methods: Data from 15,666 patients undergoing ALND at Yale Cancer Center (2013-2024) were analyzed. Preoperative Thyroid Stimulating Hormone (TSH), C-Reactive Protein (CRP), Low-Density Lipoprotein (LDL), total cholesterol, triglycerides, and Hemoglobin A1c (Hb A1C) were collected alongside demographic and clinical factors, including age, BMI, race, ethnicity, chemotherapy, and radiation. Univariate and multivariate regression models evaluated associations between laboratory markers, lymphedema development, and onset.
Results: Lymphedema developed in 2,345 patients (14.9%), with an average onset of 20.5 months post-ALND. Higher Hb A1C at the time of ALND was identified as an independent risk factor for lymphedema development (OR=1.114, 95% CI 1.001 to 1.237, p=0.0447), accounting for clinical and demographic variables. Chemotherapy (OR=2.296, 95% CI 1.522 to 3.539, p<0.0001) and BMI > 30 (OR=1.051, 95% CI 1.028 to 1.075, p<0.0001) were also significant predictors. On multivariate regression, elevated Hb A1C levels were associated with delayed lymphedema onset (β=3.708, 95% CI 1.244 to 6.171, p=0.0120) alongside BMI > 30 (β=12.18, 95% CI 3.432 to 20.92, p=0.0120). TSH, CRP, LDL, total cholesterol, or triglyceride levels at the time of ALND were not associated with lymphedema development.
Conclusions: This is the first study to propose a laboratory-based predictive model for cancer-related lymphedema after ALND, highlighting Hb A1C levels at the time of ALND as a novel predictive marker. Identifying metabolic markers at the time of ALND may enable personalized prevention strategies, including immediate lymphatic reconstruction and early physical therapy to reduce lymphedema incidence and severity.
-
Stav Brown, MD
Abstract Presenter
-
Siba Haykal, MD, PhD, FRCSC, FACS
Abstract Co-Author
-
Martin Kauke-Navarro, MD
Abstract Co-Author
-
Felix Klimitz
Abstract Co-Author
-
Alexzandra Mattia
Abstract Co-Author
-
Olivier Noel, MD, PhD
Abstract Co-Author
-
Bohdan Pomahac, MD
Abstract Co-Author
-
Yizhuo Shen
Abstract Co-Author
-
Luccie Wo, MD
Abstract Co-Author
|
3:20 PM
|
Scientific Abstract Presentations: Research & Technology Session 4 - Discussion 2
|