Scientific Abstract Presentations: Research & Technology 1 Thursday, October 26 Thu, Oct 26 1:30-2:30 p.m. CDT

Introduction: The use of artificial intelligence (AI) in healthcare has the potential to improve patient outcomes and streamline processes, particularly in the field of complex wound care. One specific application of AI in this context is the use of chat-GPT (Generative Pre-trained Transformer) to assist with tasks such as wound diagnosis and treatment recommendations along with whole-person care recommendations.

Methods: In our study, we evaluated the use of chat-GPT in a sample of 80 patients presenting to our hospital-based advanced wound care center with complex wounds. All patients were evaluated in a traditional manner by the wound care provider who established both a diagnosis and treatment plan. Next, the chat-GPT system was used to provide personalized treatment and lifestyle recommendations to patients based on their medical history, wound characteristics, and other relevant information. Providers interacted with the chat-GPT system via a natural language interface, and the system provided recommendations based on the patient information provided.

Results: Our results showed that the chat-GPT system was able to accurately identify the most appropriate treatment for most of the patients in our sample with a greater than 90% correlation to the treatment plan proposed in the initial assessment. Furthermore, providers reported high levels of satisfaction with the system, with over 87% stating that they found the recommendations to be helpful in managing their complex wound patients. Additionally, 75% of the providers acknowledged that the chat-GPT whole-person care recommendations favorably augmented their care.

Discussion: Overall, our study demonstrates the potential for AI, specifically chat-GPT, to assist with complex wound care and possibly improve patient outcomes. Further research is needed to fully understand the capabilities and limitations of this technology in this specific context, as well as to determine the most effective methods for implementing and utilizing it. However, the promising results of our study suggest that chat-GPT and similar AI systems have the potential to significantly improve complex wound care.

References: 1. Iwaya, T., & Tani, Y. (2018). Artificial intelligence-based wound diagnosis and treatment support system using deep learning. Artificial Intelligence in Medicine, 87, 1-8. 2. Zeng, Z., Mardini, H., & Goertzel, B. (2019). AI-assisted wound diagnosis and treatment: A review. Artificial Intelligence in Medicine, 95, 44-54. 3. van der Meijden, O., & van der Meer, A. D. (2018). Artificial intelligence in wound care: A systematic review. Wound Repair and Regeneration, 26(4), 511-522. 4. Teixeira, F. S., & Mendes, R. D. (2018). Artificial intelligence in wound care: A review. Wound Medicine, 22, 7-12. 5. Zhang, Y., & Zhang, S. (2018). Artificial intelligence in wound care: A review. Artificial Intelligence in Medicine, 82, 38-45.

Objective: Historically, residency candidates are evaluated through the lens of academic plastic surgeons [1], despite most plastic surgeons not pursuing academic plastic surgery. [2] Considering this, we must question whether the metrics and qualities we select for align with our goal of creating a diverse specialty optimized for treating our diverse patients. This study aims to evaluate the impact of demographic factors on perceived strengths, weaknesses, opportunities, and threats to the specialty with the goal of deconstructing the excellent plastic surgeon.

Methods: An electronic survey study was created using SurveyMonkey and distributed via the American Council of Academic Plastic Surgeons' official email address on three occasions in the weeks leading up to the 2023 Winter Meeting. All participants responded to demographic questions followed by identification of the top three strengths, weaknesses, opportunities, and threats (SWOT analysis) for the specialty.[3] Subsequently, respondents were asked to identify the five most important qualities of an excellent plastic surgeon from a list generated by literature review and expert opinion. Analyses were conducted using either Chi-Square Goodness of Fit or Fisher's Exact test with adjusted standardized ratios for post-hoc testing.

Results: We received 187 responses, representing a response rate of 89.0% of meeting attendees. Our respondents were majority non-Hispanic (78.6%) white (66.8%) women (59.5%) in the role of faculty/independently practicing physicians (65.8%). Only 12.3% of respondents identified as Hispanic/Latino, followed by 8.6% Black/African American, and <1% each American Indian/Alaskan Native and Native Hawaiian Pacific Islander. Approximately one third of respondents were first-generation low-income (FGLI) and/or LGBTQIA+ identifying (34.2%). Half of respondents were a program Chief/Chair or Program Director, with 71% of respondents working in an academic setting.

Faculty were significantly less likely to deem lack of surgeon diversity as a top 3 weakness (χ^2=19.278, p<.001) compared to trainees, yet they did identify it as a top 3 threat to the specialty (χ^2=20.639, p<.001). Black/African American, FGLI, and LGTQIA+ respondents were significantly more likely to deem lack of surgeon diversity as a top 3 weakness (p<.001 for all) and a top 3 threat (p<.05 for all).

When comparing academic and non-academic surgeons, the former were more likely to deem collaboration with other specialists as a strength (χ^2=13.276, p<.001) and perception of plastic surgery by other specialties as a threat (χ^2=4.546, p=.035), while non-academic surgeons were more likely to identify improvement in patient quality of life (χ^2=7.325, p=.006) as a strength and scope of practice creep by non-physicians as a threat (χ^2=7.242, p=.011).

All three training levels identified technical ability and collaborative/team player among the top five most important qualities of an excellent plastic surgeon.

Conclusion: Demographic minority groups and trainees are more likely to identify lack of surgeon diversity as a weakness and a threat to the future of the specialty. Furthermore, different strengths and threats emerged based on practice in an academic or non-academic setting. Regardless of training level, technical competence and collaborative skill are highly valued.

Introduction: Artificial intelligence is an innovative, rapidly expanding field that transforms the ability to learn from complex data. Applications of artificial intelligence to plastic surgery have been limited despite its ability to advance practice, research, and education. The goal of this study was to create a validated machine learning model using three-dimensional images to identify and measure differences in facial feature curvature by gender.

Methods: Three-dimensional facial photos of 75 men and 75 women aged 20-29 were collected. Each photo was divided into 100 cross-sectional images: 60 were used as training/validation data and 15 were used as test data to build a convolutional neural network that could classify gender based on curvature analysis of each cross-section. The model underwent a five-part cross-validation, and Gradient-weighted Class Activation Mapping was implemented to measure the facial curvatures used by the model as the basis of inference when determining gender.

Results: The facial target area used by the model was bounded superiorly by the eyebrows, inferiorly by the upper vermillion, and laterally by the cheekbone apices. The model could classify gender with over 90% accuracy using any sagittal plane within the target area for females and any sagittal plane lateral to the oral commissures for males. Convexity of the nasal ala, protrusion of the supraorbital ridge and nasal dorsum, and a pointed nasal tip were all predictors of male gender in the model. Flat glabella and convexity of the cheeks and eyelids were determined to be feminine features.

Conclusion: The machine learning model identified novel features as determinants of gender that currently do not serve as areas of focus for gender-affirming surgery. The objective measurements of curvature provided by the model advance the sparse, subjective literature on facial feminization/masculinization, and this method of curvature analysis can be applied to pre-operative planning for all facial reconstruction or aesthetic procedures.

Background: Pressure ulcers are challenging for patients and the healthcare system. About 2.5 million pressure ulcers are treated annually in the United States alone, resulting in approximately $11 billion per year in healthcare costs.1-4 There is currently no universally reliable modality to prevent pressure ulcer formation. We have developed and completed initial animal studies on a wearable wireless sensor less than 2 centimeters in diameter which uses algorithms that measure pressure, time, temperature, and moisture at a given skin surface site to monitor at-risk sites for pressure ulcer prevention.5 Our system uses wireless sensors and proprietary algorithms to monitor these variables continuously. The purpose of this study was to complete initial clinical trials on volunteer healthy patients to establish preliminary clinical algorithms for decision making in the use of sensors in wider clinical deployment. Methods: Healthy adult volunteers (N=20; 10 females, 10 males) were recruited for the study. Subjects were placed on a table, supine. For reproducibility the posterior heel was selected as the sensor site over the most prominent portion of the calcaneus. The calf was off-loaded with a table gap in order to maximally load the heel. Twenty pound weights were placed on the ankle to increase heel pressure. Patients were studied continuously for 30 minutes measuring pressure. Pain and numbness in each heel were recorded every three minutes using a 1-10 Likert Scale. Subject feedback on the wear-ability of the sensor at the end of the testing period as well as photo documentation of the sensor site was recorded. Results: Twenty subjects were enrolled. The average age was 31.5±13 years and the average weight was 159.3±18.5. Ten males and ten females were enrolled. Average pressure recorded by the sensor was 205.5 mmHg. Average increase in pain from baseline was 3.8 and the average pain was 4.1 by the end of the trial. Average increase in numbness from baseline was 3.1 and the average numbness by the end of the trial was 3.6. All 20 subjects reported ease of wearing the sensor. Conclusions: The results of our preliminary study demonstrate that our wearable sensors can accurately measure pressure at localized sites at high risk. Subjects reported high wear-ability and comfort of the sensors. Our preliminary findings and algorithms suggest that current clinical guidelines for pressure ulcer prevention, such as patient turning every two hours, are excessively broadly directed and unfocused, suggesting substantial opportunity for more focused prevention and care.

References: 1. Ahmed AK, Goodwin CR, Sarabia-Estrada R, et al. A non-invasive method to produce pressure ulcers of varying severity in a spinal cord-injured rat model. Spinal Cord. 2016;54(12):1096-1104. 2. Alderden J, Rondinelli J, Pepper G, et al. Risk factors for pressure injuries among critical care patients: A systematic review. Int J Nurse Stud. 2017;71:97-114. 3. Bennett L, Kavner D, Lee BK, Trainor FA. Shear vs pressure as causative factors in skin blood flow occlusion. Arch Phys Med Rehabil. 1979;60(7):309-314. 4. Chan B, Cadarette S, Wodchis W, et al. Cost-of-illness studies in chronic ulcers: a systematic review. J Wound Care. 2017;26(4)S4-S14. 5. Hickle K, Slamin R, Baez A, et al. Wireless Pressure Ulcer Sensor: Validation in an Animal Model. Ann Plast Surg. 2019;82(4S Suppl 3):S215-S221. doi:10.1097/SAP.0000000000001882

Introduction. Artificial intelligence (AI) describes the field of computer learning that relies on extensive data sets and has accelerated in its capabilities through a transition to open source code. We reviewed widely used current AI softwares surrounding smart speech-to-text transcriptions,1 computer-generated images from text,2 and conversational Chat programs,3 to illustrate how open-sourced AI (OpenAI) can be incorporated into the plastic surgeon's toolkit now, and how it may change the field in the future.

Methods. We trialed the following OpenAI interfaces: (1) Stable Diffusion and (2) DALL-E, both text-to-image applications; (3) Clip, an image-to-text program, and (4) ChatGPT, a conversational Chat AI software. Inputs included operative descriptions from plastic surgeon operative notes, de-identified images from plastic surgery case reports, and verbal descriptions discussing operative techniques. AI-generated outputs were evaluated for accuracy to true anatomy and steps in the surgical procedure.

Results:
Current OpenAI systems can generate semi-accurate depictions of surgical procedures using phrases from operative reports. Stable Diffusion has the benefit of "image-to-image" mode, in which users can command the program to edit an existing image via a text prompt. This process allows for more accurate image adjusting via iteratively composing an image. To convert intraoperative images and videos into operative notes, we found that CLIP and ChatGPT can generate captions for inputted images.4 However, the program is pre-trained on public image databases which are currently limited in relevant content for surgical purposes. More development is needed before these programs can describe more specialized images. With a broader training database, OpenAI can be expanded to caption intraoperative images.

Conclusion. Open source, web-based applications can generate and edit accurate operative and anatomical figures. As researchers refine AI image and video generation, plastic surgeons can use AI to make operative reports more image- and video-based and accelerate the operative report writing process.

1. Whisper.ai I. Whisper.ai. Updated 2022. Accessed 02/25/23, https://whisper.ai/
2. Online SD. Stable Diffusion. Accessed 02/25/23, https://stablediffusionweb.com/
3. OpenAI. ChatGPT. Updated 02/13/23. Accessed 02/25/23, https://chat.openai.com/chat
4. Mokady R, Hertz A, Bermano AH. ClipCap: CLIP Prefix for Image Captioning. 2021;doi:10.48550/arxiv.2111.09734

The development of neuropathic pain will affect upwards of 70% of all amputees. Surgical techniques to reroute amputated nerves to create an appropriate signal for a prosthesis have been shown to also reduce amputation-related pain. One of these, targeted muscle reinnervation (TMR) reroutes nerves to motor branches. Previously we studied TMR in a rodent using the spared nerve injury model which preserved the hindlimb for standard pain behavior testing in the portion of the foot that remained innervated. Our aim is to expand to a full amputation model to study TMR and analgesia. This study evaluated a rat hindlimb amputation model to determine if it was effective to examine pain with and without TMR.

Ten male rats were randomly split into two cohorts: amputation with immediate TMR (iTMR) and amputation-only. For iTMR the common peroneal, tibial, and sural branches were transected and coapted to the motor branches of the semimembranosus and bicep femoris. For the amputation-only cohort, the nerves were ligated, and a 4-5mm distal segment was removed. A full below the knee amputation was then performed in both groups. Baseline behavior testing pre-amputation was obtained. At two weeks and five weeks post-amputation, von Frey (mechanical hypersensitivity) and pin responses (hyperalgesia) were tested. Conditioned place preference (CPP) with gabapentin was completed at three weeks post-amputation. At five weeks post-amputation, acetone cold hypersensitivity and guarding (spontaneous pain), and flinching (spontaneous) were assessed.

Hyperalgesia measured at two weeks post-amputation showed the iTMR rats had an average noxious response of 10%, and amputation-only had an average noxious response of 46% (p<0.05). At four weeks post-amputation, iTMR rats had an average noxious response of 36%, while amputation-only rats had 48% noxious response (p<0.05). For acetone cold hypersensitivity, iTMR rats exhibited 30% noxious responses, while amputation-only exhibited 100% noxious responses (p<0.05). Rats with iTMR demonstrated an average of 19.2 seconds of guarding over a two-minute interval, and amputation-only rats averaged 28.9 seconds (p<0.05). At seven weeks, rats underwent nerve stimulation, and three of the amputation-only rats exhibited successful reinnervation.

The pain behaviors of iTMR rats showed significantly greater analgesia, including reduced hyperalgesia, cold hypersensitivity, and spontaneous pain, compared to amputation only. However, the reinnervation of some of the ligated nerves in amputation-only rats may have resulted in the trend of improved pain behaviors over time. To better reflect clinical findings in humans, future studies will focus on preventing reinnervation in the amputation-only model and investigate sex-based differences. Additionally, our work will aim to assess differences in motor versus sensory regeneration compared to different interventions and provide a more thorough characterization of the regenerative microenvironment at the coaptation sites.

Background Augmented reality (AR) is the process of visually overlaying digital information on top of the physical world and can include interactions between the digital display and physical world. Within the field of plastic and reconstructive surgery (PRS), AR can utilize patient imaging to aid preoperative surgical planning, intraoperative image guidance, as well as patient and resident education. Previous reviews of AR in PRS have been limited in not addressing the underlying technological shortcomings of AR nor fully evaluating physician usability limitations. In this review, we discuss both current uses as well limitations that need to be addressed as development moves forward.

Methods A review of relevant literature was conducted. Electronic databases were screened using keywords including "augmented reality," "mixed reality," and "plastic surgery." Studies were individually assessed for quality using technological usability heuristics.

Results A total of 90 studies were reviewed. Several studies used the phrase "augmented reality" interchangeably with other imaging and guidance modalities, especially other types of extended reality, such as mixed reality (MR) or virtual reality (VR). Of the applied AR device studies, primary usage fell into one of three categories: preoperative surgical planning, intraoperative surgical guidance, or surgical education. A wide range of devices within the AR umbrella were utilized, with studies most commonly citing the HoloLens (Microsoft LLC, Redmond, Wash.). Across several plastic surgical subspecialties, AR has demonstrated practicality and success in surgical planning and education but minimal intraoperative usage due to device specific limitations or issues with physician usability. The most cited technical issues hampering widespread adoption were software difficulties distinguishing soft tissue and need to streamline artificial marker registration. Meanwhile, physician usability issues included limited field of vision, insufficient battery life and dim viewing conditions.

Conclusion The utility of augmented reality in plastic surgery is an exciting and nascent field of study. While there have been many initial attempts to develop relevant technology, significant limitations remain that constrain AR's ability to be used as an autonomous intraoperative guidance system. Promoting engineering-physician partnerships will allow for prioritization of key physician usability issues critical for AR's success in the operating room.

Background It has become increasingly apparent that a plastic surgery residency application with robust and lengthy research publications is the new standard when evaluating applicant competitiveness. Integrated residency match data from 2022 demonstrates a mean total publications of 7.21 and 1st author publications of 5.1 for matched medical students. This study aims to evaluate the research history of plastic surgery program faculty leadership to glean insight into the trends and evolution of research expectations.

Methods Faculty members who serve in leadership positions in the residency admissions process among all ACGME-accredited integrated and independent plastic surgery residency programs in the U.S. were reviewed. The following faculty positions were included in the study: Chairs, Chiefs, Associate Chiefs, Program Directors, and Associate Program Directors. Doximity and individual program websites were used to gather information on faculty. H-index as well as number of publications at the time of graduation from medical school, general surgery residency, and plastic surgery residency were collected through SCOPUS. Faculty were further classified based on whether they hold a leadership position at one of the top 20 research medical institutions in US News recordings. Student's t-test and Pearson's product-moment correlation were used to analyze data.

Results Two hundred thirty-six plastic surgeons held the title of Chair, Chief, Associate Chief, Program Director, or Associate Program Director and were considered in the analyses. Altogether, faculty held a median of 0 (IQR 0-1) publications by the end of medical school, 1 (IQR 0-3) publications by the end of general surgery residency (for those who attended), and 3 (IQR 1-8) publications by plastic surgery residency graduation. Number of publications at the end of medical school were greater in those who attended integrated programs compared to independent plastic surgery programs (1.67 ± 5.00 publications vs. 0.56 ± 1.48 publications; p=0.09). Integrated program residents graduated with significantly higher publications after plastic surgery residency compared to independent program residents (8.48 ± 10.53 publications vs. 4.89 ± 5.89 publications; p=0.01). More recent graduation year was weakly positively correlated with more publications obtained at the time of plastic surgery residency graduation (p<0.001). Faculty leaders at top 20 research medical institutions had significantly higher H-indices than those from other programs (24.2 ± 15.19 vs. 14.2 ± 10.87; p<0.001). Chiefs were found to have significantly higher H-indices compared to Program Directors and Associate positions (Chief: 23.7 ± 15.60, Program Director: 14.7 ± 11.05, Associate position: 12.2 ± 9.31; p<0.001).

Conclusions Based on current data, the majority of research publications in faculty leaders was performed after graduation from plastic surgery residency. However, it appears that research volume among plastic surgeons-in-training has increased more recently. This is especially evident among integrated pathway residents, who had a higher number of publications than their independent colleagues. Though historically it was not necessary to have a high pre-graduate research output to become an eventual leader, the current required research threshold to matriculate into a plastic surgery residency has profoundly increased.

Purpose: As many medical students continue to pursue research initiatives as a means of increasing their prospects in the plastic surgery (PS) match, presenting at national conferences has become a rite of passage for most applicants. Conferences can also serve as an additional opportunity for students with no home integrated PS residency program (NHP) to network. However, little is known about the backgrounds and match outcomes of student presenters in plastic surgery. Given the rapidly changing climate of the application process, an assessment of the trends and utility of students presenting research at national PS conferences is necessary.

Methods: Names of medical student presenters from the 2013-2020 ACAPS, AAPS, and ASPS annual meeting programs were obtained online after exclusion criteria were applied. Doximity, residency webpages, and LinkedIn were used to determine when and where each individual attended medical school and residency. Data was collected on the frequency of presentations, the presenters' medical schools, where their research was performed, and their match outcomes. Pearson's χ2 test was performed to assess differences between groups.

Results: In total, 369 students delivered a total of 600 presentations across three national PS conferences from 2013-2020. The number of student presentations in 2020 was five times greater than in 2013. Of this total population, 63% matriculated into an integrated PS residency program. Only 16% of student presenters were from medical schools with NHP, of which 53% of them presented research performed at other institutions with home programs (HP). Overall, NHP presenters had a 59% average PS match rate, while HP student presenters had a 64% match rate (p=0.57). NHP students who performed research externally had a 74% match rate, while NHP students who researched within their own institutions had a 43% match rate (p=0.014). Finally, students who delivered just one presentation had a 55% average PS match rate, while students who delivered multiple presentations had a 77% PS match rate (p<0.001).

Conclusions: Presenting frequently at conferences is positively correlated to improved PS match outcomes for medical students, including those without home integrated programs. The rising volume of student presentations at national conferences accurately mirrors the increasing competitiveness of the PS match in recent years. For NHP students, performing research externally at other HP institutions is associated with increased success in the PS match. Further discussion is warranted to determine whether the key benefit of presenting at conferences is the opportunity to network with faculty members, having a platform to demonstrate engagement in PS research, or a combination of the two. Nonetheless, the rising numbers across the years indicate that presenting at national conferences will remain an integral component of the PS application process.