Pub Date : 2024-05-07DOI: 10.1101/2024.05.06.24306947
Lee Wheless, Kai-Ping Liao, Siwei Zheng, Yao Li, Lydia Yao, Yaomin Xu, Christopher Madden, Jacqueline Ike, Isabelle T Smith, Dominique Mosley, Sarah Grossarth, Rebecca I Hartman, Otis Wilson, Adriana Hung, Mackenzie R Wehner
Importance Many patients will develop more than one skin cancer, however most research to date has examined only case status.
重要性 许多患者会罹患一种以上的皮肤癌,但迄今为止,大多数研究仅对病例状况进行了调查。
{"title":"Toward personalized skin cancer care: multiple skin cancer development in five cohorts","authors":"Lee Wheless, Kai-Ping Liao, Siwei Zheng, Yao Li, Lydia Yao, Yaomin Xu, Christopher Madden, Jacqueline Ike, Isabelle T Smith, Dominique Mosley, Sarah Grossarth, Rebecca I Hartman, Otis Wilson, Adriana Hung, Mackenzie R Wehner","doi":"10.1101/2024.05.06.24306947","DOIUrl":"https://doi.org/10.1101/2024.05.06.24306947","url":null,"abstract":"<strong>Importance</strong> Many patients will develop more than one skin cancer, however most research to date has examined only case status.","PeriodicalId":501385,"journal":{"name":"medRxiv - Dermatology","volume":"4 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-05-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140938315","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-04-12DOI: 10.1101/2024.04.09.24305289
Soham Gadgil, Alex J. DeGrave, Roxana Daneshjou, Su-In Lee
Recent advances in Artificial Intelligence (AI) have started disrupting the healthcare industry, especially medical imaging, and AI devices are increasingly being deployed into clinical practice. Such classifiers have previously demonstrated the ability to discern a range of protected demographic attributes (like race, age, sex) from medical images with unexpectedly high performance, a sensitive task which is difficult even for trained physicians. Focusing on the task of predicting sex from dermoscopic images of skin lesions, we are successfully able to train high-performing classifiers achieving a ROC-AUC score of ∼0.78. We highlight how incorrect use of these demographic shortcuts can have a detrimental effect on the performance of a clinically relevant downstream task like disease diagnosis under a domain shift. Further, we employ various explainable AI (XAI) techniques to identify specific signals which can be leveraged to predict sex. Finally, we introduce a technique to quantify how much a signal contributes to the classification performance. Using this technique and the signals identified, we are able to explain ∼44% of the total performance. This analysis not only underscores the importance of cautious AI application in healthcare but also opens avenues for improving the transparency and reliability of AI-driven diagnostic tools.
{"title":"Discovering mechanisms underlying medical AI prediction of protected attributes","authors":"Soham Gadgil, Alex J. DeGrave, Roxana Daneshjou, Su-In Lee","doi":"10.1101/2024.04.09.24305289","DOIUrl":"https://doi.org/10.1101/2024.04.09.24305289","url":null,"abstract":"Recent advances in Artificial Intelligence (AI) have started disrupting the healthcare industry, especially medical imaging, and AI devices are increasingly being deployed into clinical practice. Such classifiers have previously demonstrated the ability to discern a range of protected demographic attributes (like race, age, sex) from medical images with unexpectedly high performance, a sensitive task which is difficult even for trained physicians. Focusing on the task of predicting sex from dermoscopic images of skin lesions, we are successfully able to train high-performing classifiers achieving a ROC-AUC score of ∼0.78. We highlight how incorrect use of these demographic shortcuts can have a detrimental effect on the performance of a clinically relevant downstream task like disease diagnosis under a domain shift. Further, we employ various explainable AI (XAI) techniques to identify specific signals which can be leveraged to predict sex. Finally, we introduce a technique to quantify how much a signal contributes to the classification performance. Using this technique and the signals identified, we are able to explain ∼44% of the total performance. This analysis not only underscores the importance of cautious AI application in healthcare but also opens avenues for improving the transparency and reliability of AI-driven diagnostic tools.","PeriodicalId":501385,"journal":{"name":"medRxiv - Dermatology","volume":"42 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140573448","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-28DOI: 10.1101/2024.03.27.24304982
Dana Jolley, Varshita Chirumamilla, Abraham Korman
Background: The current clinical misdiagnosis rate among all medical specialties is approximately 10-15%, but diagnostic error within the field of dermatology has not been studied thoroughly1,2. As a field that relies heavily on visual perception, many physicians consider clinical intuition to be advantageous in diagnosing skin diseases and consider it to be a rapid and unconscious phenomenon7. Therefore, too much contemplation may lead to more incorrect diagnoses4. However, while clinical intuition is a valuable clinical tool, it is widely considered to be developed throughout medical training and only successfully employed by experienced attending physicians, perhaps due to experiential knowledge and associated confidence1,2,5. One may expect that self-reported confidence in diagnosis would correlate with diagnostic accuracy, but this is not supported in the literature9. The focus of our study is to examine the development and reliability of clinical intuition as well as associated self-reported confidence levels in diagnoses at different levels of medical training among dermatologists. Methods: Approximately 20 dermatologists who are PGY-2 or higher will be recruited for study participation via email. Participants will be sent a Qualtrics survey at two separate time points with a month waiting period in between. The survey will contain demographics questions, photos of 10 different dermatologic conditions for dermatologists to diagnose, and a self-reported confidence level for each diagnosis. The first survey will allow 5 seconds to evaluate a clinical photo prior to diagnosis, and this timeframe will be extended to 15 seconds in the second survey. The second survey will contain the same diagnoses, but with different pictures to avoid recall of specific photos. Following completion of all surveys, descriptive statistics will be completed with goal of publication. Discussion: This study has the potential to provide invaluable information regarding the development of clinical intuition among dermatologic physicians while also examining their confidence levels and likelihood of changing correct diagnoses when given more time to ruminate. It is possible that physicians are more likely to second guess original diagnoses based off of certain demographic factors, as one systematic review found that women in medicine perceive their clinical performance as deficient more often than men10. Therefore, this study may give insight to the ways that complicated societal factors contribute to clinical decision making. Data from this study may be used to aid dermatologists in understanding their thought processes when diagnosing patients, and may be useful in developing education curriculum. The protocol will hopefully serve as a blueprint for creation of studies in a multitude of fields, ultimately leading to better understanding of clinical decision making and, thus, improved patient care.
{"title":"Protocol: Trust Your Gut: An Analysis of Dermatologic Diagnostic Accuracy","authors":"Dana Jolley, Varshita Chirumamilla, Abraham Korman","doi":"10.1101/2024.03.27.24304982","DOIUrl":"https://doi.org/10.1101/2024.03.27.24304982","url":null,"abstract":"Background: The current clinical misdiagnosis rate among all medical specialties is approximately 10-15%, but diagnostic error within the field of dermatology has not been studied thoroughly1,2. As a field that relies heavily on visual perception, many physicians consider clinical intuition to be advantageous in diagnosing skin diseases and consider it to be a rapid and unconscious phenomenon7. Therefore, too much contemplation may lead to more incorrect diagnoses4. However, while clinical intuition is a valuable clinical tool, it is widely considered to be developed throughout medical training and only successfully employed by experienced attending physicians, perhaps due to experiential knowledge and associated confidence1,2,5. One may expect that self-reported confidence in diagnosis would correlate with diagnostic accuracy, but this is not supported in the literature9. The focus of our study is to examine the development and reliability of clinical intuition as well as associated self-reported confidence levels in diagnoses at different levels of medical training among dermatologists. Methods: Approximately 20 dermatologists who are PGY-2 or higher will be recruited for study participation via email. Participants will be sent a Qualtrics survey at two separate time points with a month waiting period in between. The survey will contain demographics questions, photos of 10 different dermatologic conditions for dermatologists to diagnose, and a self-reported confidence level for each diagnosis. The first survey will allow 5 seconds to evaluate a clinical photo prior to diagnosis, and this timeframe will be extended to 15 seconds in the second survey. The second survey will contain the same diagnoses, but with different pictures to avoid recall of specific photos. Following completion of all surveys, descriptive statistics will be completed with goal of publication. Discussion: This study has the potential to provide invaluable information regarding the development of clinical intuition among dermatologic physicians while also examining their confidence levels and likelihood of changing correct diagnoses when given more time to ruminate. It is possible that physicians are more likely to second guess original diagnoses based off of certain demographic factors, as one systematic review found that women in medicine perceive their clinical performance as deficient more often than men10. Therefore, this study may give insight to the ways that complicated societal factors contribute to clinical decision making. Data from this study may be used to aid dermatologists in understanding their thought processes when diagnosing patients, and may be useful in developing education curriculum. The protocol will hopefully serve as a blueprint for creation of studies in a multitude of fields, ultimately leading to better understanding of clinical decision making and, thus, improved patient care.","PeriodicalId":501385,"journal":{"name":"medRxiv - Dermatology","volume":"25 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140322195","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-03-28DOI: 10.1101/2024.03.27.24304771
Erinolaoluwa Araoye, Taylor Jamerson, Lu Yin, Kristen Lo Sicco, Crystal Aguh
Background Hairstyling practices are associated with the development and/or exacerbation of various forms of alopecia. Exposure to various hairstyling practices ranges but is often insufficient in current dermatologic textbooks and training curricula. We therefore conducted a survey to establish dermatologists understanding of hairstyling practices, particularly those that have been implicated in alopecia. Methods: A 34-item anonymous, electronic survey was distributed by email to 291 board-certified dermatologists and dermatology residents across the US between August 2020 and February 2021. Responses were rated on a 10-point scale to identify physician confidence in various styling practices Results: Black providers were more confident in both the knowledge and counseling of all hair practices (chemical straightening, heat styling, braiding, weaving, and wigs) compared to non-Black providers (p <0.001), with the exception of counseling patients on hair dyes for which no significant difference was found (p=0.337). Female providers were only more likely to indicate confidence in knowledge regarding different heat styling methods and hair dyes, and counseling of heat styling methods compared to male providers (OR 15.72, p<0.001; OR 2.47, p=0.022; OR 3.78, p=0.001 respectively) across all hair practices surveyed. Overall, 63.8% of providers reported that the majority of their knowledge on hair practices was from personal experience as opposed to formal training. Limitations: This survey is limited by its response rate and the inability to characterize non-responders due to anonymity. Conclusion: Our study highlights educational gaps in dermatologic training on hair practices, especially those more common among Black patients. Interestingly, the majority of provider knowledge came from personal experience rather than dermatologic training emphasizing the need for formalized curricula to enhance understanding among all dermatology providers.
{"title":"Highlighting Educational Gaps in Hairstyling Practices amongst Dermatologists and Trainees","authors":"Erinolaoluwa Araoye, Taylor Jamerson, Lu Yin, Kristen Lo Sicco, Crystal Aguh","doi":"10.1101/2024.03.27.24304771","DOIUrl":"https://doi.org/10.1101/2024.03.27.24304771","url":null,"abstract":"Background Hairstyling practices are associated with the development and/or exacerbation of various forms of alopecia. Exposure to various hairstyling practices ranges but is often insufficient in current dermatologic textbooks and training curricula. We therefore conducted a survey to establish dermatologists understanding of hairstyling practices, particularly those that have been implicated in alopecia. Methods: A 34-item anonymous, electronic survey was distributed by email to 291 board-certified dermatologists and dermatology residents across the US between August 2020 and February 2021. Responses were rated on a 10-point scale to identify physician confidence in various styling practices\u0000Results: Black providers were more confident in both the knowledge and counseling of all hair practices (chemical straightening, heat styling, braiding, weaving, and wigs) compared to non-Black providers (p <0.001), with the exception of counseling patients on hair dyes for which no significant difference was found (p=0.337). Female providers were only more likely to indicate confidence in knowledge regarding different heat styling methods and hair dyes, and counseling of heat styling methods compared to male providers (OR 15.72, p<0.001; OR 2.47, p=0.022; OR 3.78, p=0.001 respectively) across all hair practices surveyed. Overall, 63.8% of providers reported that the majority of their knowledge on hair practices was from personal experience as opposed to formal training. Limitations: This survey is limited by its response rate and the inability to characterize non-responders due to anonymity. Conclusion: Our study highlights educational gaps in dermatologic training on hair practices, especially those more common among Black patients. Interestingly, the majority of provider knowledge came from personal experience rather than dermatologic training emphasizing the need for formalized curricula to enhance understanding among all dermatology providers.","PeriodicalId":501385,"journal":{"name":"medRxiv - Dermatology","volume":"117 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140322356","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-27DOI: 10.1101/2024.02.26.24303359
Danique Berrevoet, Filip Van Nieuwerburgh, Dieter Deforce, Reinhart Speeckaert
An unbiased screening of which proteins are deregulated in vitiligo using proteomics can offer an enormous value. It could not only reveal robust biomarkers for detecting disease activity but can also identify which patients are most likely to respond to treatments. We performed a scoping review searching for all articles using proteomics in vitiligo. Eight manuscripts could be identified. Unfortunately, very limited overlap was found in the differentially expressed proteins between studies (15 out of 272; 5,51%) with variable degrees of the type of proteins and a substantial variety in the prevalence of acute phase proteins (range: 6-65%). Proteomics research has therefore brought little corroborating evidence on which proteins are differentially regulated between vitiligo patients and healthy controls or between active and stable vitiligo patients. While a limited patient size is an obvious weakness for several studies, an incomplete description of patient characteristics is an unfortunate and avoidable shortcoming. Additionally, the variations in the used methodology and analyses may further contribute to the overall observed variability. Nonetheless, more recent studies investigating the response to treatment seem to be more robust, as more differentially expressed proteins that have previously been confirmed to be involved in vitiligo were found. The further inclusion of proteomics analyses in clinical trials is recommended to increase insights into the pathogenic mechanisms in vitiligo and identify reliable biomarkers or promising drug targets. A harmonization in the study design, reporting and proteomics methodology could vastly improve the value of vitiligo proteomics research.
{"title":"Proteomics data in vitiligo: a scoping review","authors":"Danique Berrevoet, Filip Van Nieuwerburgh, Dieter Deforce, Reinhart Speeckaert","doi":"10.1101/2024.02.26.24303359","DOIUrl":"https://doi.org/10.1101/2024.02.26.24303359","url":null,"abstract":"An unbiased screening of which proteins are deregulated in vitiligo using proteomics can offer an enormous value. It could not only reveal robust biomarkers for detecting disease activity but can also identify which patients are most likely to respond to treatments. We performed a scoping review searching for all articles using proteomics in vitiligo. Eight manuscripts could be identified. Unfortunately, very limited overlap was found in the differentially expressed proteins between studies (15 out of 272; 5,51%) with variable degrees of the type of proteins and a substantial variety in the prevalence of acute phase proteins (range: 6-65%). Proteomics research has therefore brought little corroborating evidence on which proteins are differentially regulated between vitiligo patients and healthy controls or between active and stable vitiligo patients. While a limited patient size is an obvious weakness for several studies, an incomplete description of patient characteristics is an unfortunate and avoidable shortcoming. Additionally, the variations in the used methodology and analyses may further contribute to the overall observed variability. Nonetheless, more recent studies investigating the response to treatment seem to be more robust, as more differentially expressed proteins that have previously been confirmed to be involved in vitiligo were found. The further inclusion of proteomics analyses in clinical trials is recommended to increase insights into the pathogenic mechanisms in vitiligo and identify reliable biomarkers or promising drug targets. A harmonization in the study design, reporting and proteomics methodology could vastly improve the value of vitiligo proteomics research.","PeriodicalId":501385,"journal":{"name":"medRxiv - Dermatology","volume":"256 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139978294","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-02-06DOI: 10.1101/2024.02.05.24302298
Shams Nassir, Miranda Yousif, Xing Li, Kevin Severson, Alysia Hughes, Jacob Kechter, Angelina Hwang, Blake Boudreaux, Puneet Bhullar, Nan Zhang, Duke Butterfield, Tao Ma, Ewoma Ogbaudu, Collin M Costello, Steven Nelson, David J DiCaudo, Aleksandar Sekulic, Christian Baum, Mark Pittelkow, Aaron R Mangold
Cutaneous squamous cell carcinoma (cSCC) is one of the most common cancers in humans and kills as many people annually as melanoma. The mutational and transcriptional landscape of cSCC has identified driver mutations associated with disease progression as well as key pathway activation in the progression of pre-cancerous lesions. The understanding of the transcriptional changes with respect to high-risk clinical/histopathological features and outcome is poor. Here, we examine stage-matched, outcome-differentiated cSCC and associated clinicopathologic risk factors using whole exome and transcriptome sequencing on matched samples. Exome analysis identified key driver mutations including TP53, CDKN2A, NOTCH1, SHC4, MIIP, CNOT1, C17orf66, LPHN22, and TTC16 and pathway enrichment of driver mutations in replicative senescence, cellular response to UV, cell-cell adhesion, and cell cycle. Transcriptomic analysis identified pathway enrichment of immune signaling/inflammation, cell-cycle pathways, extracellular matrix function, and chromatin function. Our integrative analysis identified 183 critical genes in carcinogenesis and were used to develop a gene expression panel (GEP) model for cSCC. Three outcome-related gene clusters included those involved in keratinization, cell division, and metabolism. We found 16 genes were predictive of metastasis (Risk score ≥ 9 Met & Risk score < 9 NoMet). The Risk score has an AUC of 97.1% (95% CI: 93.5% - 100%), sensitivity 95.5%, specificity 85.7%, and overall accuracy of 90%. Eleven genes were chosen to generate the risk score for Overall Survival (OS). The Harrell’s C-statistic to predict OS is 80.8%. With each risk score increase, the risk of death increases by 2.47 (HR: 2.47, 95% CI: 1.64-3.74; p<0.001) after adjusting for age, immunosuppressant use, and metastasis status.
{"title":"Whole Exome and Transcriptome Sequencing of Stage-Matched, Outcome-Differentiated Cutaneous Squamous Cell Carcinoma Identifies Gene Expression Patterns Associated with Metastasis and Poor Outcomes","authors":"Shams Nassir, Miranda Yousif, Xing Li, Kevin Severson, Alysia Hughes, Jacob Kechter, Angelina Hwang, Blake Boudreaux, Puneet Bhullar, Nan Zhang, Duke Butterfield, Tao Ma, Ewoma Ogbaudu, Collin M Costello, Steven Nelson, David J DiCaudo, Aleksandar Sekulic, Christian Baum, Mark Pittelkow, Aaron R Mangold","doi":"10.1101/2024.02.05.24302298","DOIUrl":"https://doi.org/10.1101/2024.02.05.24302298","url":null,"abstract":"Cutaneous squamous cell carcinoma (cSCC) is one of the most common cancers in humans and kills as many people annually as melanoma. The mutational and transcriptional landscape of cSCC has identified driver mutations associated with disease progression as well as key pathway activation in the progression of pre-cancerous lesions. The understanding of the transcriptional changes with respect to high-risk clinical/histopathological features and outcome is poor. Here, we examine stage-matched, outcome-differentiated cSCC and associated clinicopathologic risk factors using whole exome and transcriptome sequencing on matched samples. Exome analysis identified key driver mutations including <em>TP53</em>, <em>CDKN2A</em>, <em>NOTCH1</em>, <em>SHC4</em>, <em>MIIP</em>, <em>CNOT1</em>, <em>C17orf66</em>, <em>LPHN22</em>, and <em>TTC16</em> and pathway enrichment of driver mutations in replicative senescence, cellular response to UV, cell-cell adhesion, and cell cycle. Transcriptomic analysis identified pathway enrichment of immune signaling/inflammation, cell-cycle pathways, extracellular matrix function, and chromatin function. Our integrative analysis identified 183 critical genes in carcinogenesis and were used to develop a gene expression panel (GEP) model for cSCC. Three outcome-related gene clusters included those involved in keratinization, cell division, and metabolism. We found 16 genes were predictive of metastasis (Risk score ≥ 9 Met & Risk score < 9 NoMet). The Risk score has an AUC of 97.1% (95% CI: 93.5% - 100%), sensitivity 95.5%, specificity 85.7%, and overall accuracy of 90%. Eleven genes were chosen to generate the risk score for Overall Survival (OS). The Harrell’s C-statistic to predict OS is 80.8%. With each risk score increase, the risk of death increases by 2.47 (HR: 2.47, 95% CI: 1.64-3.74; p<0.001) after adjusting for age, immunosuppressant use, and metastasis status.","PeriodicalId":501385,"journal":{"name":"medRxiv - Dermatology","volume":"41 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139755058","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Background: The integration of artificial intelligence (AI) in dermatology presents a promising frontier for enhancing diagnostic accuracy and treatment planning. However, general purpose AI models require rigorous evaluation before being applied to real-world medical cases. Objective: This project specifically evaluates GPT-4V's performance in accurately diagnosing and generating treatment plans for common dermatological conditions, comparing its assessment of textual versus image data and its performance with multimodal inputs. Beyond the immediate scope, this study contributes to the broader trajectory of integrating AI in healthcare, highlighting the limitations of these technologies, as well as their potential to enhance efficiency, and education within medical training and practice. Methods: A dataset of 102 images representing nine common dermatological conditions was compiled from open-access websites. Fifty-four images were ultimately selected by two board- certified dermatologists as being representative and typical of the common conditions. Additionally, nine clinical scenarios corresponding to these conditions were developed. GPT- 4V's diagnostic capabilities were assessed in three setups: Image Prompt (image-based), Scenario Prompt (text-based), and Image and Scenario Prompt (combining both modalities). The model's performance was evaluated based on diagnostic accuracy, differential diagnosis, and treatment recommendations. Results: In the Image Prompt setup, GPT-4V correctly identified the primary diagnosis for 29 of 54 images. The Scenario Prompt setup showed a higher accuracy rate of 89% in identifying the primary diagnosis. The multimodal Image and Scenario Prompt setup also achieved an 89% accuracy rate. However, a notable bias towards textual data over visual data was observed. Treatment recommendations were evaluated by the same two dermatologists, using a modified Entrustment Scale, showing competent but not expert-level performance. Conclusion: GPT-4V demonstrates promising capabilities in dermatological diagnosis and treatment recommendations, particularly in text-based scenarios. However, its performance in image-based diagnosis and integration of multimodal data highlights areas for improvement. The study underscores the potential of AI in augmenting dermatological practice, emphasizing the need for further development, and fine-tuning of such models to ensure their efficacy and reliability in clinical settings. Keywords: Artificial Intelligence; Dermatology, GPT-4V; Diagnostic Accuracy; Treatment Planning; Multimodal AI; Large Language Model.
{"title":"Evaluating the Diagnostic and Treatment Recommendation Capabilities of GPT-4 Vision in Dermatology","authors":"Abhinav Pillai, Sharon Parappally-Joseph, Jori Hardin","doi":"10.1101/2024.01.24.24301743","DOIUrl":"https://doi.org/10.1101/2024.01.24.24301743","url":null,"abstract":"Background: The integration of artificial intelligence (AI) in dermatology presents a promising frontier for enhancing diagnostic accuracy and treatment planning. However, general purpose AI models require rigorous evaluation before being applied to real-world medical cases.\u0000Objective: This project specifically evaluates GPT-4V's performance in accurately diagnosing and generating treatment plans for common dermatological conditions, comparing its assessment of textual versus image data and its performance with multimodal inputs. Beyond the immediate scope, this study contributes to the broader trajectory of integrating AI in healthcare, highlighting the limitations of these technologies, as well as their potential to enhance efficiency, and education within medical training and practice.\u0000Methods: A dataset of 102 images representing nine common dermatological conditions was compiled from open-access websites. Fifty-four images were ultimately selected by two board- certified dermatologists as being representative and typical of the common conditions. Additionally, nine clinical scenarios corresponding to these conditions were developed. GPT- 4V's diagnostic capabilities were assessed in three setups: Image Prompt (image-based), Scenario Prompt (text-based), and Image and Scenario Prompt (combining both modalities). The model's performance was evaluated based on diagnostic accuracy, differential diagnosis, and treatment recommendations.\u0000Results: In the Image Prompt setup, GPT-4V correctly identified the primary diagnosis for 29 of 54 images. The Scenario Prompt setup showed a higher accuracy rate of 89% in identifying the primary diagnosis. The multimodal Image and Scenario Prompt setup also achieved an 89% accuracy rate. However, a notable bias towards textual data over visual data was observed. Treatment recommendations were evaluated by the same two dermatologists, using a modified Entrustment Scale, showing competent but not expert-level performance.\u0000Conclusion: GPT-4V demonstrates promising capabilities in dermatological diagnosis and treatment recommendations, particularly in text-based scenarios. However, its performance in image-based diagnosis and integration of multimodal data highlights areas for improvement. The study underscores the potential of AI in augmenting dermatological practice, emphasizing the need for further development, and fine-tuning of such models to ensure their efficacy and reliability in clinical settings.\u0000Keywords: Artificial Intelligence; Dermatology, GPT-4V; Diagnostic Accuracy; Treatment Planning; Multimodal AI; Large Language Model.","PeriodicalId":501385,"journal":{"name":"medRxiv - Dermatology","volume":"180 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-01-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139585525","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-09DOI: 10.1101/2023.12.07.23299659
Sigrid Lundgren, Ganna Petruk, Karl Wallblom, José FP Cardoso, Ann-Charlotte Strömdahl, Fredrik Forsberg, Congyu Luo, Bo Nilson, Erik Hartman, Jane Fisher, Manoj Puthia, Karim Saleh, Artur Schmidtchen
The skin microbiome undergoes dynamic changes during different phases of wound healing, however the role of bacteria in the wound healing process remains poorly described. In this study, we aimed to determine how wound bacteria develop over time in epidermal wounds, and how they interact with inflammatory processes during wound healing. To this end, we analyzed wound fluid and swab samples collected from epidermal suction blister wounds in healthy volunteers. We found that bacterial numbers, measured in swabs and dressing fluid, increased rapidly after wounding and stabilized by day 8. The composition of bacterial species identified by MALDI-TOF mass spectrometry differed between wounds, but generally consisted primarily of commensal bacteria and remained largely stable over time. Inflammation and neutrophil activity, measured by quantification of cytokines and neutrophil proteins in dressing fluid, peaked on day 5. Exudation, measured by quantification of protein content in dressings, also peaked at this time and strongly correlated with cytokine and neutrophil protein levels. Inflammation, neutrophil activity, and exudation were not correlated with bacterial counts at any time, indicating that in normally healing wounds, these processes are primarily driven by the host and are independent of colonizing bacteria. Our analysis provides a comprehensive understanding of epidermal wound healing dynamics in the host and the role of the microbiome in healthy wound healing.
{"title":"Analysis of bacteria, inflammation, and exudation in epidermal suction blister wounds reveals dynamic changes during wound healing","authors":"Sigrid Lundgren, Ganna Petruk, Karl Wallblom, José FP Cardoso, Ann-Charlotte Strömdahl, Fredrik Forsberg, Congyu Luo, Bo Nilson, Erik Hartman, Jane Fisher, Manoj Puthia, Karim Saleh, Artur Schmidtchen","doi":"10.1101/2023.12.07.23299659","DOIUrl":"https://doi.org/10.1101/2023.12.07.23299659","url":null,"abstract":"The skin microbiome undergoes dynamic changes during different phases of wound healing, however the role of bacteria in the wound healing process remains poorly described. In this study, we aimed to determine how wound bacteria develop over time in epidermal wounds, and how they interact with inflammatory processes during wound healing. To this end, we analyzed wound fluid and swab samples collected from epidermal suction blister wounds in healthy volunteers. We found that bacterial numbers, measured in swabs and dressing fluid, increased rapidly after wounding and stabilized by day 8. The composition of bacterial species identified by MALDI-TOF mass spectrometry differed between wounds, but generally consisted primarily of commensal bacteria and remained largely stable over time. Inflammation and neutrophil activity, measured by quantification of cytokines and neutrophil proteins in dressing fluid, peaked on day 5. Exudation, measured by quantification of protein content in dressings, also peaked at this time and strongly correlated with cytokine and neutrophil protein levels. Inflammation, neutrophil activity, and exudation were not correlated with bacterial counts at any time, indicating that in normally healing wounds, these processes are primarily driven by the host and are independent of colonizing bacteria. Our analysis provides a comprehensive understanding of epidermal wound healing dynamics in the host and the role of the microbiome in healthy wound healing.","PeriodicalId":501385,"journal":{"name":"medRxiv - Dermatology","volume":"3 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138573398","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-07DOI: 10.1101/2023.12.05.23299283
Guy Fletcher, David K. Ryan, C B. Bunker
Introduction Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are a group of severe acute muco-cutaneous blistering disorders with a significant burden of morbidity and mortality. Drugs are commonly identified as potential precipitants of SJS/TEN, although it can be difficult to firmly identify causative agents. Ibuprofen has been proposed as a rare trigger for SJS/TEN and given the widespread use of this non-steroidal anti-inflammatory and significance of reaction, further pharmacovigilance analysis is warranted.
{"title":"Association between Stevens-Johnson syndrome and toxic epidermal necrolysis with ibuprofen: A pharmacovigilance study in the UK Yellow Card scheme and systematic review of case reports","authors":"Guy Fletcher, David K. Ryan, C B. Bunker","doi":"10.1101/2023.12.05.23299283","DOIUrl":"https://doi.org/10.1101/2023.12.05.23299283","url":null,"abstract":"<strong>Introduction</strong> Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are a group of severe acute muco-cutaneous blistering disorders with a significant burden of morbidity and mortality. Drugs are commonly identified as potential precipitants of SJS/TEN, although it can be difficult to firmly identify causative agents. Ibuprofen has been proposed as a rare trigger for SJS/TEN and given the widespread use of this non-steroidal anti-inflammatory and significance of reaction, further pharmacovigilance analysis is warranted.","PeriodicalId":501385,"journal":{"name":"medRxiv - Dermatology","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138562876","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-06DOI: 10.1101/2023.12.06.23299413
Rim Mhedbi, Peter Credico, Hannah O. Chan, Rakesh Joshi, Joshua N. Wong, Colin Hong
The escalating incidence of skin lesions, coupled with a scarcity of dermatologists and the intricate nature of diagnostic procedures, has resulted in prolonged waiting periods. Consequently, morbidity and mortality rates stemming from untreated cancerous skin lesions have witnessed and upward trend. To address this issue, we propose a skin lesion classification model that leverages EfficientNet B7 Convolutional Neural Network(CNN) architecture, enabling early screening of skin lesions based on camera images. The model is trained on a diverse dataset encompassing eight distinct skin lesion classes: Basal Cell Carcinoma(BCC), Squamous Cell Carcinoma(SCC), Melanoma(MEL), Dysplastic Nevus(DN), Benign Keratosis-Like lesions(BKL), Melanocytic Nevi(NV), and an 'Other' class. Through Multiple iterations of data preprocessing, as well as comprehensive error analysis, the model achieves a remarkable accuracy rate of 87%
{"title":"A Convolutional Neural Network based system for classifying malignant and benign skin lesions using mobile-device images","authors":"Rim Mhedbi, Peter Credico, Hannah O. Chan, Rakesh Joshi, Joshua N. Wong, Colin Hong","doi":"10.1101/2023.12.06.23299413","DOIUrl":"https://doi.org/10.1101/2023.12.06.23299413","url":null,"abstract":"The escalating incidence of skin lesions, coupled with a scarcity of dermatologists and the intricate nature of diagnostic procedures, has resulted in prolonged waiting periods. Consequently, morbidity and mortality rates stemming from untreated cancerous skin lesions have witnessed and upward trend. To address this issue, we propose a skin lesion classification model that leverages EfficientNet B7 Convolutional Neural Network(CNN) architecture, enabling early screening of skin lesions based on camera images. The model is trained on a diverse dataset encompassing eight distinct skin lesion classes: Basal Cell Carcinoma(BCC), Squamous Cell Carcinoma(SCC), Melanoma(MEL), Dysplastic Nevus(DN), Benign Keratosis-Like lesions(BKL), Melanocytic Nevi(NV), and an 'Other' class. Through Multiple iterations of data preprocessing, as well as comprehensive error analysis, the model achieves a remarkable accuracy rate of 87%","PeriodicalId":501385,"journal":{"name":"medRxiv - Dermatology","volume":"2 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138547392","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}