Scientific Abstract Presentations: Research & Technology Session 2 Thursday, September 26 Thu, Sep 26 3:00-4:00 p.m. PDT

BACKGROUND: Wound care is a critical aspect of healthcare, and advancements in therapeutic modalities continually seek to enhance the healing process. The novel transforming powder dressing (TPD), Altrazeal®, is composed of biologically compatible polymers (85% poly-2-hydroxyethylmethacrylate (pHEMA) and 15% poly-2-hydroxypropylmethacrylate (pHPMA)). It was engineered as a dressing for surgical wounds, traumatic wounds, chronic wounds, and burns. TPD is simple to use and painless to apply, as the powder is sprinkled over a moistened wound and has a wear time of up to 30 days. It has numerous purported benefits, including enhanced wound healing, pain reduction, cost reduction, and reduced care time. Multiple studies across a wide spectrum of wound types have reported on the efficacy of TPD, however, there is yet to be a systematic review that summarizes the results of these studies. This study aims to review the current literature to elucidate the impact of TPD on wound healing, pain management, and overall patient outcomes.

METHODS: We conducted a systematic review following PRISMA (Preferred Reporting Items of Systematic Reviews and Meta-Analyses) guidelines. Data including study characteristics, patient demographics, and wound outcomes were extracted. Articles were assessed for risk of bias and descriptive statistics were performed. The protocol for this study was registered in PROSPERO (ID: CRD42023485051).

RESULTS: Our systematic review included 26 articles representing 175 patients. Of these articles, 13 (50%) were case reports, 10 (38.5%) were case series, 2 (7.7%) were RCTs, and 1 (3.8%) was a cohort study. The most common wound type for which TPD was applied was venous ulcer (23.9%), followed by pressure sore (19.7%), burn (15.5%), skin graft (13.4%), diabetic foot ulcer (4.2%), Mohs defect (3.5%), abdominal wound (2.8%), trauma (2.8%), acne inversa (2.8%), chronic leg ulcer (2.8%), necrotizing fasciitis (2.8%), pilonidal cyst (2.1%), and other (3.5%). Complete re-epithelialization occurred in 90.1% of the wounds treated with TPD. A total of 19 studies (73%) discussed pain, each of which reported reduced pain with the use of TPD. No patients experienced complications related to their treatment with TPD.

CONCLUSION: The evaluated studies collectively suggest that TPD offers a promising re-epithelialization rate and analgesic effect across various wound types, encompassing surgical wounds, traumatic wounds, chronic wounds, and burns. Compared to the current standard-of-care dressing options, the long wear time and ease of application may also reduce the financial burden and care time associated with frequent dressing changes and thus should be the subject of future research.

PURPOSE Characterizing triple negative breast cancer (TNBC) cell behavior within personalized in vitro models of the breast milieu is critical to understanding disease progression and advancing therapeutics. Challenges to developing such platforms include co-culturing diverse primary cell types to recapitulate the complex 3D tumor microenvironment and establishing an analytic approach for reliable assessment of cell behavior overtime. It is also crucial to provide accurate measurement of TNBC and endothelial cell invasion and migration, particularly in detecting potently metastatic and angiogenic cell clusters. This study aims to: 1) develop a multilayer 3D platform capable of co-culturing TNBC cells, endothelial cells, and patient-derived breast adipocytes, stromal vascular fraction, and ductal epithelial organoids, and 2) employ a machine learning-based algorithm to quantify TNBC and endothelial cell migration and identify highly migratory subsets.

METHODS Breast tissue from healthy patients undergoing reduction mammoplasty (n=6, average age of 31.8, BMI of 29.1) was processed to isolate mature adipocytes, stromal vascular fraction, and ductal organoids and then suspended in 0.3% collagen. The final 3D construct contained three layers: 1) a base layer with RFP-tagged MDA-MB-231 tumor cells ("Tumor"), 2) a middle layer with either biomimetic collagen containing breast derived cellular constituents ("BM") or collagen alone without cells ("Collagen"), and 3) a top layer with GFP-tagged human umbilical vascular endothelial cells ("HUVEC"). Six separate combinations were fabricated: Tumor-BM-HUVEC, Tumor-Collagen-HUVEC, Tumor-Collagen, Tumor-BM, Collagen-HUVEC, and BM-HUVEC. All groups were plated in triplicate on a 96-well plate and cultured at 37°C under hypoxic (5% O2) or normoxic (21% O2) conditions. Confocal imaging was performed on days 0 and 5. A machine learning algorithm was developed in ImarisTM software to measure vertical cell movement along the z-axis, and a regression plane of cell location was reconstructed in RStudio. Migrated cells were defined as those exceeding one standard deviation away from the fitted plane.

RESULTS On days 0 and 5, confocal microscopy revealed ~800μm thick constructs with distinct TNBC and HUVEC cell layers. For migrated TNBC cells under normoxic condition, Tumor-BM-HUVEC group had the furthest migration of 125.3µm, followed by Tumor-BM at 103.2µm, with significantly less migration in groups without breast derived cell constituents: Tumor-Collagen-HUVEC at 62.4µm and Tumor-Collagen at 60.4µm (p<0.0001). Similar findings were observed under hypoxic conditions: Tumor-BM-HUVEC group migrated 88.4µm, Tumor-BM 84.8µm, Tumor-Collagen-HUVEC 59.1µm, and Tumor-Collagen 49.3µm. When focusing on migrated HUVEC cells, Tumor-BM-HUVEC group demonstrated significantly greater migration compared to groups without breast derived cell constituents and/or tumor cells (Tumor-Collagen-HUVEC, Collagen-HUVEC, and BM-HUVEC) under both normoxia (34.2µm vs 24.7µm, 22.2µm, and 28.5µm, respectively, p<0.01) and hypoxia (42.0µm vs 26.6µm, 22.9µm, 29.6µm, p<0.0001).

CONCLUSIONS This novel biomimetic construct allows for reliable, precise characterization of TNBC and HUVEC cell behavior within a biomimetic breast tumor microenvironment under both normoxic and hypoxic conditions. Increased TNBC and HUVEC migration seen in the biomimetic milieu suggests that the presence of breast derived cell constituents may contribute to enhanced cancer metastasis and angiogenesis. This unique tissue engineering platform holds significant promise as a personalized, high-throughput screening tool for investigating cancer cell behavior and potential therapeutics.

Introduction

An interdisciplinary approach benefits patient care and outcomes. Despite the benefits of collaboration, the extent to which plastic surgeons engage in interdisciplinary research has not been previously investigated. This study aims to quantify interdisciplinary collaboration patterns within plastic surgery.

Methods

PubMed was queried for all articles published from academic plastic surgery programs between 2018 and 2023. The resulting article list was merged with the Journal Citation Report to assess impact factor. An article was considered collaborative if the senior author was from an academic plastic surgery institution and at least one author was from a different department. A chi-squared test was performed to examine the association between collaboration and publication in a plastic surgery vs non-plastic surgery journal.

Results

21,068 articles were retrieved, 14,590 articles were reviewed after removal of duplicates, and 6686 articles qualified for inclusion. Less than half (44.8%; n=2992) of articles were collaborative, among these the majority (56.1%; n=1679) collaborated with another surgical department. The three most collaborated-with surgical specialties were general surgery (17.6%; n=522), otolaryngology (14.6%; n=437), and orthopedic surgery (11.8%; n=354). 38.1% (n=1141) of collaborations were with non-surgical specialties, most frequently, genetics and genomics (13.9%; n=416), internal medicine (8.5%; n=253); and radiology (5.1%; n=153). There were also a number of non-medical collaborations, including dentistry (4.9%; n=147) and engineering (7.6%; n=228).

Regionally, there were no significant differences in the proportion of collaborations in 15 programs in the Northeast (40.3%), 12 programs in the Midwest (42.0%), 9 programs in the South (45.8%), or 9 programs in the West (44.8%). Of note, there was no overlap between the three academic programs with the highest volume of publications (Stanford, n=490; New York University, n=380; and the University of Michigan, n=380) and three academic programs with the highest proportion of collaborative publications (Mayo Clinic, Jacksonville (79.2%; n=61); Mayo Clinic, Rochester (73.2%; n=101); University of California San Diego (61.5%; n=48)). Productivity at an institution and the proportion of their articles with collaboration were independent (R2=0.0003; p=0.9148).

Similarly, there was no overlap between plastic surgery journals with the highest volume of articles and those with the highest proportion of collaborative articles (Annals of Plastic Surgery (64.3%; n=142); Aesthetic Plastic Surgery (53.8%; n=56); Journal of Plastic, Reconstructive & Aesthetic Surgery (50.8%; n=97)). Most collaborative articles were published in non-plastic surgery journals (54.8%; n=1641), and most non-collaborative articles were published in plastic surgery journals (69.7%; n=2574) (p<0.0001).

Conclusion

This study defined the diverse affiliations that contribute to plastic surgery being a highly productive field by determining the prevalence of interdisciplinary collaboration in the literature. These results emphasize the importance of collaboration in research, offering valuable insights for both aspiring trainees seeking programs aligned with their research interests and academic institutions aiming to enhance the reach and impact of their research through strategic partnerships.

Introduction: Even though the alarm of women under-representation in surgery has been going off for decades, most approaches to measuring women under-representation have been sporadic and inconsistent.1 Our study employs an Artificial Intelligence (AI) big data approach to comprehensively and accurately document each state's gender disparity in research productivity and representation. We evaluate the impact of state family support policies, such as paid family leave (PFL) and reproductive rights, on women representation and productivity in plastic surgery research.

Methods: We followed Karamitros and Goulas in deploying AI technology by means of a web-scraping algorithm on PubMed to retrieve author names and affiliations for each publication on all PubMed-indexed plastic surgery journals on the SCImago classification from 2010 to 2022.2 We identified authors' gender using gender API. Using information by the Guttmacher Institute and Bipartisan Policy Center we classified each state by their reproductive rights (protective vs restrictive) and PFL rights (mandatory and active or not) respectively. We statistically compared the gender gap (i.e., difference between males and females) in representation (i.e., authorship) and productivity (i.e., number of publications) between states with and without family support policies.

Results: First, our AI approach identified 7,489 first authors, who produced a total of 16,019 publications. A significantly smaller gender publication gap was identified in states with mandatory PFL laws and protective reproductive rights as compared with those without, respectively (0.50 vs 0.68 with p<0.001 and 0.53 vs 0.66 with p<0.001). When progressively restricting the sample to authors with at least 2, 5, or 10 publications, the identified impact of PFL on the publication gap grew from 0.18 to 0.54 (p<0.001), 0.72 (p<0.001), and 2.83 (p<0.001), respectively. Similarly, the impact of reproductive rights grew from 0.13 to 0.25 (p<0.001), 0.88 (p<0.001), and 6.28 (p<0.001), respectively. When considering productivity of at least 2 and 5 publications, PFL laws are associated with an increased gender gap in representation (1.27 with p<0.001 and 2.17 with p<0.001 for the 2- and 5-publication cutoff, respectively). At the same time, reproductive rights are associated with a decreased gender gap in representation (1.80 with p<0.001 and 0.45 with p<0.001 for the 2- and 5-publication cutoff, respectively). This suggests that reproductive rights may be more effective than PFL laws in allowing women to penetrate through higher statuses of research productivity.

Conclusion: Our extensive data extraction via AI allowed us to investigate the direct impact of state-level protections for women on the plastic surgery labor market. Our policy evaluation suggests that state legislation can be effective in fostering gender equity and promoting research productivity parity. Our study serves as a signpost advocating for the importance of state level protections for early-career women plastic surgeons.

Citations: 1. Stephens EH, Heisler CA, Temkin SM, Miller P. The Current Status of Women in Surgery: How to Affect the Future. JAMA Surg. 2020;155(9):876–885. doi:10.1001/jamasurg.2020.0312 2. Goulas S, Karamitros G. Human capital and productivity in surgery research during a pandemic: an artificial intelligence approach. Br J Surg. 2023 Aug 11;110(9):1218-1220. doi: 10.1093/bjs/znad205. PMID: 37379481; PMCID: PMC10416675.

Background: Manual chart review is labor-intensive and error-prone. Integrating Large Language Models (LLMs), like GPT, may help to streamline research, reduce errors, and improve the reliability of plastic surgery outcome studies. This study compares the performance of GPT-3.5 Turbo and GPT-4 to manual review of patient operative reports detailing capsule revision in the second stage of implant-based breast reconstruction.

Methods: Filtering the STAnford Research Repository (STARR) using CPT codes, we randomly selected 101 adult patients who underwent second stage implant-based breast reconstruction from January 2018 to January 2021. Manual chart review of operative reports identified patient demographics and presence of capsule revisions, including capsulectomy, capsulorrhaphy, and 'extensive' capsulotomy, defined as those performed non-inferiorly. GPT-3.5 Turbo and GPT-4 first filtered all entries for each patient to identify a second stage breast reconstruction. Next, GPT recorded the occurrence of each capsule procedure. Discrepancies between manual review and each model were recorded.

Results: For the 101 patients, there were 1128 total entries; however, due to human error and inconsistencies in reporting, it was difficult to filter to only operative reports without manual review or the use of GPT. The GPT-3.5 Turbo model performed well at identifying operative reports where a second stage breast reconstruction occurred, incorrectly categorizing just one operative report. GPT-4 correctly identified all desired operative reports. When compared to the manual review, a naïve approach which determines only the presence of the word 'capsulectomy' and 'capsulorrhaphy' in the filtered operative reports for each patient yielded a 27% and 61% accuracy rate, respectively. Since we were interested in 'extensive' capsulotomies, as predefined, simply counting the presence of the term 'capsulotomy' would be incorrect, thereby requiring interpretation. Notably, many second stage operative reports did not explicitly use capsule procedure terms (e.g. 'capsulectomy'). In contrast, when instructed to determine the presence of capsulectomies, extensive capsulotomies and capsulorrhaphies during the second stage, the GPT-3.5 Turbo model correctly predicted 80.2%, 83.2%, and 82.2%, respectively. When considering this, GPT-3.5 performed well in determining the occurrence of capsule procedures with an overall success rate of approximately 81.8%. Notably, the overall recall was 0.92 with a lower precision of 0.73. The F-score was 0.814, which is a measure of model accuracy considering both precision and recall. Re-running with GPT-4 considerably improved performance, correctly classifying 89.1% of capsulectomies, 94.0% of extensive capsulotomies, and 92.1% of capsulorrhaphies. This yielded an overall success rate of 91.7%, overall recall of 0.99, and overall precision of 0.83. With GPT-4, the F-score was 0.903, indicating high efficacy at correctly identifying true positives while minimizing false positives and false negatives.

Conclusion: While GPT-4 demonstrated improved performance compared to GPT-3.5 Turbo, both models exhibit minimal limitations in precisely matching the manual chart review. These findings emphasize the potential for integrating LLMs into plastic surgery outcome studies, offering significant benefits in reducing human error and time spent. However, they also highlight the importance of cautious interpretation and the complementary role of manual data collection.

Purpose: Clinical outcome studies support use of incisional negative pressure wound therapy (iNPWT), with benefits including lower rates of surgical site infection, fewer wound complications, and improved scarring. Few studies, however, have convincingly explicated the efficacy of iNPWT. This systematic review examines reported mechanisms of iNPWT efficacy, synthesizing these findings into a cohesive model with particular attention to lymphangiogenesis and the ideal mechanical environment for wound healing.

Methods: This systematic review was conducted in accordance with PRISMA guidelines. PUBMED, OVID, and Web of Science were searched. All studies published before February 2024 were assessed for eligibility. Inclusion criteria were full text prospective or retrospective studies in English that reported on 1) in vitro mechanistic models, 2) biomechanical studies, 3) histologic analysis, and 4) macroscopic tissue changes induced by iNPWT. Reviews, opinions, case reports, and abstracts were excluded.

Results: Eighteen studies met inclusion criteria. All manuscripts were published between 1997 and 2023. Several studies suggest that iNPWT increases tissue perfusion and circulation of immunomodulatory cells, thereby decreasing rates of wound infection. There is contention about how these changes occur. Some studies propose that iNPWT initially induces local tissue ischemia and triggers release of angiogenic factors including VEGF and nitric oxide, while others suggest that suction forces stent open capillary beds for enhanced flow. These differences can, in part, be due to varying imaging modalities used to assess perfusion (ICG angiography vs. doppler flow vs. thermal imaging), which is a weakness of established literature.

From a biomechanical standpoint, like traditional vac therapy, iNPWT generates macrodeformational forces that support wound healing. iNPWT provides external suction that, one study hypothesizes, aligns deep dermal structures and allows for faster recovery of tissue tensile strength in the initial postoperative period. This suction force may also reduce local edema by encouraging lymphatic uptake of serous fluid, as postulated by three included studies. Like silicone taping, iNPWT has been modelled to reduce lateral incisional tension and realign force vectors to better approximate wound margins. These models are supported by qualitative outcomes reporting favourable scarring with use of iNPWT compared to conventional dressings.

Two studies propose the use of iNPWT as external splints in the contexts of orthopaedic incisions and hand burns, which stand out from commonly described applications for iNPWT. These applications suggest that the semirigid yet conforming nature of iNPWT is a fundamental reason why this technology is effective.

Conclusion: Existing literature suggests that iNPWT utilizes macrodeformational forces to better approximate incisions, increase blood flow to wound margins, and enhance lymphatic circulation. This coincides with use of iNPWT during initial stages of wound healing with hemostasis, inflammation, and proliferation, perhaps providing a superior mechanical environment for regeneration than conventional dressings. Additional factors including degree of negative suction (-75 vs. -125mmHg), optimal length of therapy, and use of contact layers have yet to be considered. Not discussed in the included studies are changes in patient behaviour with use of iNPWT, including early vs. delayed mobility, awareness of incisions, and wound care needs.

Purpose: Microsurgery is an intricate and specialized field within plastic surgery that has significant constraints due to the high resources needed for training and operation. One significant constraint is the unavailability of surgical microscopes within healthcare systems both in the United States and on the global scale. This leaves trainees with a lack of access to surgical microscopes to practice and develop their microsurgery competence. Despite prior investigations attempting to overcome this scarcity of microscopes using iPhones and virtual reality systems, there remains a gap in having high fidelity microscopes for training purposes. This investigation presents a novel microscope that utilizes low-cost materials to produce a high fidelity, highly adaptable, and travel friendly surgical microscope for microsurgery training.

Methods: Binocular objective lenses were modified for near field viewing and coupled with a double mirror array in series to produce a stereoscopic microscope that does not require prisms or calibration. This system was housed within a low-profile 3D printed chassis to produce the correct orientation for a top-down view of a microsurgical field. A microscope light source was built into the chassis that may be powered either with standard 120-240V outlets or battery powered in case of resource limited environments.

Results: The initial proof of concept and working prototype produce a fixed magnification of either 6.5X that enables trainees to complete end-to-end anastomoses on a 2mm vessel. Eight microscopes were created and transported to Rwanda, Ethiopia, and Vietnam for microsurgery training utilized by the SHARE plastic surgery organization and the author's home institution. This investigation remains ongoing to compare the utility of this microscope against a state-of-the art surgical microscope.

Conclusions: Microsurgical education and practice are often hampered by the high cost and inaccessibility of surgical microscopes. Our novel stereomicroscope aims to alleviate these barriers and promote better accessibility to residents and trainees on a global scale. By providing a low cost, portable solution that does not compromise the quality of the stereoscopic image, this innovation has the potential to transform the field of microsurgical education and practice.

Background: Plastic surgery (PS) remains one of the most competitive specialties in the NRMP Match, with the number of applicants increasing at a rate disproportional to the growth in number of positions. Numerous studies have been conducted on research productivity trends among successfully matched PS candidates; however, few have evaluated the subset of matriculants from medical schools without home integrated PS residency programs. This study aims to characterize the research productivity of PS matriculants in 2022 and 2023, comparing those with a home PS program to those without.

Methods: Websites and social media accounts of Accreditation Council for Graduate Medical Education (ACGME)-accredited integrated PS residency programs were queried to identify incoming first-year residents in 2022 and 2023. For each individual, PubMed was queried for peer-reviewed publications through the September prior to their matriculation and data was cross-verified using Scopus, Google Scholar, and ResearchGate. Additional study variables included number of PS-related publications, number of first-author publications, and author h-index. Data analysis was conducted in SPSS using Shapiro-Wilk and Wilcoxon-Mann-Whitney tests.

Results: From the 2022 match, 86 integrated PS programs and 194 first-year residents were identified and from the 2023 match, 88 integrated PS programs and 207 first-year residents were identified. After excluding eight individuals with outlier data points and 25 individuals with no identifiable data, the final study cohort included 368 residents. Of all matched applicants, 259 (70.4%) had a home program and 109 (29.6%) did not. The average number of PubMed-indexed publications was 5.96 ± 6.13, though matriculants with a home program had significantly more than those without a home program (6.46 ± 6.57 versus 4.77 ± 5.51; p = .003). Additionally, matriculants with a home program had a significantly greater number of first-author publications than those without a home program (2.00 ± 2.65 versus 1.42 ± 2.05; p = 0.023). No significant differences were found between cohorts when comparing author h-index and number of PS-related publications.

Conclusion: Similar findings between groups for h-index and number of PS-related publications indicates that those without a home program have a similarly competitive research background and impact when compared to their counterparts, though they comprise a minority of the matriculant pool. The disparity in volume of publications and first-author publications may be attributed to a lack of research-related resources at schools without home programs. Future directions of this work include determining how best to ameliorate the lack of research resources at schools without home programs, especially examining the rise of virtual research laboratories aimed at providing opportunities to students who may not otherwise have them.

Background/Purpose: Artificial Intelligence (AI) harbors significant potential for efficient capture and synthesis of information, enhancing its applicability across diverse fields to optimize the learning process. Notably, investigations into the generative capabilities of ChatGPT have demonstrated its proficiency in summarizing research articles and composing abstracts, albeit with some shortcomings (1). This prompts concerns about the extent to which AI can accurately integrate and summarize information from diverse sources and fields of study. The phenomenon of "artificial hallucination," wherein AI generates convincing yet unsubstantiated facts, has been recognized as a concern within the scientific community (2). Literature on the severity of artificial hallucination in the medical domain remains limited, with existing studies revealing a propensity for AI to fabricate data not supported by scientific literature (3). While previous studies shed light on the tendencies of ChatGPT, a broader examination of artificial hallucination across alternative AI interfaces and in varied domains is warranted. Our study seeks to address this gap by investigating artificial hallucination in five generative AI platforms, including ChatGPT, Perplexity, Claude, Microsoft Copilot, and Writesonic, within the context of plastic and reconstructive surgery, an unexplored domain concerning artificial hallucination.

Methods: We examined five different AI systems (Microsoft Copilot, Perplexity, Claude, ChatGPT 3.5, and Writesonic). These AI platforms were prompted to generate summaries on gender-affirming surgery and facial feminization surgery. They were then instructed to provide accompanying information on literature sources (title, author, PMID, and DOI). Verification through a PubMed web search identified instances of artificial hallucination, characterized by non-existent or incorrectly labeled results.

Results: Notably, ChatGPT 3.5 and WriteSonic exhibited 100% artificial hallucination across all categories, while Microsoft Copilot demonstrated the highest accuracy, correctly providing 80% of titles and 60% of authors accurately. Perplexity exhibited 40% accuracy in titles. Claude, however, consistently failed to provide qualifying information, marking DOI and PMID as "not available."

Conclusion: In conclusion, the findings underscore the need for continued improvements in AI accuracy, particularly in source citation. The study reveals varying degrees of artificial hallucination among the AI platforms, emphasizing the importance of ongoing screening and refinement to enhance their reliability and effectiveness in information synthesis. Authors relying upon AI to synthesize data for scientific purposes should understand the limitations and pitfalls of current AI tools.

References: 1. Salvagno M, Taccone FS, Gerli AG. Correction to: Can artificial intelligence help for scientific writing?. Crit Care. 2023;27(1):99. Published 2023 Mar 8. doi:10.1186/s13054-023-04390-0 2. Alkaissi H, McFarlane SI. Artificial Hallucinations in ChatGPT: Implications in Scientific Writing. Cureus. 2023;15(2):e35179. Published 2023 Feb 19. doi:10.7759/cureus.35179
3. Salvagno, M., Taccone, F.S. & Gerli, A.G. Artificial intelligence hallucinations. Crit Care 27, 180 (2023). https://doi.org/10.1186/s13054-023-04473-y

Introduction

Perforator location is commonly assessed preoperatively on CT via computer screen and correlated to intraoperative patient anatomy by surface landmarks. This process of translating 2D into 3D, however, can be imprecise and lead to uncertainty or difficult dissections. In particular, the Lateral Thigh Perforator (LTP) flap has highly variable anatomy that leads to multiple possible orientations. Even in more predictable flaps, such as the DIEP, improved anatomical understanding can improve decision making. Mixed reality (MR) allows three-dimensional renderings of CT data to be projected upon the donor site skin, allowing the surgeon to better visualize sub-surface anatomy. We aimed to assess the viability of this technology in both DIEP and LTP flap design.

Methods

The Medivis SurgicalAR system was used in conjunction with the Microsoft HoloLens 2, an AR/MR headset with a see-through visor. CT imaging was projected onto a patient undergoing a DIEP procedure, after registration using four fiducials – the bilateral anterior superior iliac spine, and superior and inferior points of the umbilicus. For the DIEP flap design, time and discrepancy of perforator location were assessed.  Although an LTP flap was not performed, the location and course of the LTP perforator was visualized and used to draw a mock flap design.

Results

Registration error was estimated to be 2mm in the rostral-caudal axis on the right ASIS, 4mm in the rostral-caudal axis on the left ASIS, and <1mm on both umbilicus points in all directions. Registering the hologram to the patient lasted 77 seconds. The course of the LTP pedicle vessels were clearly visualized, allowing the surgeon to easily draw an LTP flap design onto the skin with ideal location and axis orientation. Three candidate perforators were identified for the left LTP and two for the right. Flap outlining lasted 62 seconds on the left and 40 on the right. DIEP perforators were rapidly identified and traced in 21 seconds.

Discussion

In this first intraoperative evaluation of perforator identification for flap placement guidance, we found our process to be highly accurate and efficient. Despite concerns surrounding soft tissue deformation leading to differences between preoperative CT and patient anatomy in the OR, we did not find any differences that impacted the ability to localize perforators or outline flaps. We believe this process and technology have the potential to improve flap design and elevate standard of care.