Accuracy of Store-and-Forward Teledermatology for the Diagnosis of Skin Cancer: The Nouvelle-Aquitaine Experience

Background In Nouvelle-Aquitaine (a French region with a population of almost 6 million), the density of dermatologists is less than 3.8/100,000 inhabitants. This lack of dermatological care is delaying diagnosis and management, especially for skin cancer. The SmartDerm Project is a store-and-forward (SAF) teledermatology platform for primary care in Nouvelle-Aquitaine providing dermatological counselling to general practitioners (GPs). Objective The main objective was to determine the concordance between the diagnosis of skin cancer made by dermatologists and the pathologists’ diagnosis. Methods GPs in 3 pilot departments of Nouvelle-Aquitaine (Lot-Et-Garonne, Deux-Sèvres, Creuse) sent their dermatology requests using their smartphone, via an app called PAACO/Globule; dermatologists at the University Hospital of Bordeaux answered within 48-72 hours. Consecutive cases of skin cancer suspected by the referent dermatologists during the intervention were included, if the result of biopsy interpreted by a certified pathologist was available at the time of the study. Results Among the 1727 requests, 163 (9%) concerned a possible diagnosis of skin cancer and were eligible. For 61 cases, the histopathological findings were not available. Eventually, 93 patients with a total of 102 skin lesions were included. Median age was 75 years (range 26-97 years), with 53% women. The skin lesions had progressed for 8 months on average (range 0.5-36 months). The median response time was 1 day (range 0-61 days); 65 days (range 1-667 days) elapsed on average between the SAF opinion and the histological sample. Histopathology diagnosed 83 malignant lesions (57 basal cell carcinomas, 69%; 18 squamous cell carcinomas, 22%; 6 melanomas, 7%; 1 cutaneous lymphoma, 1%; 1 secondary location of a primary cancer, 1%), 1 precancerous lesion, and 18 benign lesions. The concordance between the opinion of the referent dermatologist and the final pathological finding was 83% for nonmelanocytic lesions and 67% for melanocytic lesions. Conclusions This study showed the reliability of SAF teledermatology in the diagnosis of skin cancer, comparable to literature data in the absence of dermatoscopy. The median delay of about two months between request and histology was an improvement compared to the delay of usual appointments in the intervention area. The lack of data for 61 patients showed that SAF telemedicine requires better coordination and follow-up, especially for the management of skin cancer. With this reservation in mind, teledermatology offers an alternative answer for the triage of patients with skin cancer residing in areas with low medical density. Conflicts of Interest None declared.

Early Implementation and Evaluation of a Teledermatology Virtual Clinic Within an Academic Medical Center

Background Teledermatology can increase patient access; however, its optimal implementation remains unknown. Objective This study aimed to describe and evaluate the implementation of a pilot virtual clinic teledermatology service at Duke University. Methods Leaders at Duke Dermatology and Duke Primary Care identified a teledermatology virtual clinic to meet patients’ access needs. Implementation was planned over the exploration, preparation, implementation, and sustainment phases. We evaluated the implementation success of teledermatology using the Reach, Effectiveness, Adoption, Implementation, and Maintenance framework and prioritized outcome collection through a stakeholder survey. We used the electronic health record and patient surveys to capture implementation outcomes. Results Our process consisted of primary care providers (PCPs) who sent clinical and dermatoscopic images of patient lesions or rashes via e-communication to a teledermatology virtual clinic, with a subsequent virtual clinic scheduling of a video visit with the virtual clinic providers (residents or advanced practice providers, supervised by Duke Dermatology attending physicians) within 2-5 days. The teledermatology team reviews the patient images on the day of the video visit and gives their diagnosis and management plan with either no follow-up, teledermatology nurse follow-up, or in-person follow-up evaluation. Implementation at 4 pilot clinics, involving 19 referring PCPs and 5 attending dermatologists, began on September 9, 2021. As of October 31, 2021, a total of 68 e-communications were placed (50 lesions and 18 rashes) and 64 virtual clinic video visits were completed. There were 3 patient refusals and 1 conversion to a telephonic visit. Participating primary care clinics differed in the number of patients referred with completed visits (range 2-32) and the percentage of providers using e-communications (range 13%-53%). Patients were seen soon after e-communication placement; compared to in-person wait times of >3 months, the teledermatology virtual clinic video visits occurred on average 2.75 days after e-communication. In total, 20% of virtual clinic video visits were seen as in-person visit follow-up, which suggests that the majority of patients were deemed treatable at the virtual clinic. All patients who returned the patient survey (N=10, 100%) agreed that their clinical goals were met during the virtual clinic video visits. Conclusions Our virtual clinic model for teledermatology implementation resulted in timely access for patients, while minimizing loss to follow-up, and has promising patient satisfaction outcomes. However, participating primary care clinics differ in their volume of referrals to the virtual clinic. As the teledermatology virtual clinics scale to other clinic sites, a systematic assessment of barriers and facilitators to its implementation may explain these interclinic differences. Acknowledgments We are grateful to the Private Diagnostic Clinic and Duke Institute for Health Innovation for their support. Conflicts of Interest None declared.

A 9-Year Teledermoscopy Service: Retrospective Service Review

Background A teledermoscopy service was established in January 2010, where patients attended nurse-led clinics for imaging of lesions of concern and remote diagnosis by a dermatologist. Objective The study aimed to review the number of visits, patient characteristics, the efficiency of the service, and the diagnoses made. Methods We evaluated the waiting time and diagnosis of skin lesions for all patient visits from January 1, 2010, to May 31, 2019. The relationships between patient characteristics and the diagnosis of melanoma were specifically analyzed. Results The teledermoscopy clinic was attended by 6479 patients for 11,005 skin lesions on 8805 occasions. Statistically significant risk factors for the diagnosis of melanoma/melanoma in situ were male sex, European ethnicity, and Fitzpatrick skin type 2. Attendance was maximal during 2015 and 2016. The seasonal variation in visits 2011-2018 revealed a consistent peak at the end of summer and a dip at the end of winter. In the year 2010, 306 patients attended; 76% (233/306) of these were discharged to primary care and 24% (73/306) were referred to hospital for specialist assessment. For patients diagnosed by the dermatologist with suspected melanoma from January 1, 2010, to May 31, 2019, the median waiting time for an imaging appointment was 44.5 days (average 57.9 days, range 8-218 days). The most common lesions diagnosed were benign naevus (2933/11,005, 27%), benign keratosis (2576/11,005, 23%), and keratinocytic cancer (1707/11,005, 15%); melanoma was suspected in 5% (507/11,005) of referred lesions (Multimedia Appendix 1). The positive predictive value of melanoma/melanoma in situ was 61.1% (320 true positives and 203 false positives). The number needed to treat (ie, the ratio of the total number of excisions to the number with a histological diagnosis of melanoma/melanoma in situ) was 2.02. Conclusions Diagnoses were comparable to the experience of other teledermoscopy services. Teledermoscopy using a nurse-led imaging clinic can provide efficient and convenient access to dermatology by streamlining referrals to secondary care and prioritizing patients with skin cancer for treatment. Conflicts of Interest None declared.

A Validated Questionnaire to Evaluate Primary Care Pediatrician Satisfaction With the Use of Teledermatology

Background Teledermatology (TD) is a branch of telemedicine focused on the evaluation of cutaneous lesions by dermatologists remotely, in order to avoid unnecessary in-person consults that could be otherwise resolved by this method, and to shorten the time required for prompt evaluation of cutaneous diseases. Objective This study aimed to create and validate a questionnaire to evaluate satisfaction with the use of TD among primary care pediatricians (PCPs) and to test the questionnaire in our health area before performing an intervention for the optimization of TD. Methods We first created a questionnaire based on previous publications. Then, an expert consultation was made before drafting the final version of the questionnaire. We tested it twice among pediatricians of different health areas, with a 1-month gap between both evaluations. Internal consistency, reproducibility, and validity of the questionnaire were evaluated. Finally, the validated questionnaire was tested among the PCPs of our health area, to analyze their responses. Results We registered 38 questionnaire responses. In all, 30 (78.9%) PCPs actively used TD several times within a month or a year; none of them used TD daily. Technical and health care quality of TD was mostly considered as good or very good. TD was regarded as similar or even better than face-to-face evaluation for most PCPs, whereas 7.9% (3/38) of PCPs thought TD was worse than conventional consults. Most PCPs considered TD as an effective, self-learning, and trustable tool, and 10.5% (4/38) of them identified that pictures captured by mobile phones were a barrier for its use, as it affects patient privacy. Technical problems, absence of exclusive devices for image taking, and delayed answers are some other barriers for TD that need to be overcome. Nonetheless, all PCPs were satisfied with TD, and all of them reported they would continue or start to use this tool. Conclusions TD has demonstrated to be an efficient tool, as it reduces waiting time and costs for dermatology evaluation, and it increases satisfaction among professionals. With our proposed questionnaire, we validated that quality, usability, efficacy, and satisfaction related to TD in our health area had a positive consideration among PCPs in general, but there still are barriers to overcome. Conflict of Interest None declared.

Privacy Protection With Facial Deidentification Machine Learning Methods: Can Current Methods Be Applied to Dermatology?

Background In the era of increasing tools for automatic image analysis in dermatology, new machine learning models require high-quality image data sets. Facial image data are needed for developing models to evaluate attributes such as redness (acne and rosacea models), texture (wrinkles and aging models), pigmentation (melasma, seborrheic keratoses, aging, and postinflammatory hyperpigmentation), and skin lesions. Deidentifying facial images is critical for protecting patient anonymity. Traditionally, journals have required facial feature concealment typically covering the eyes, but these guidelines are largely insufficient to meet ethical and legal guidelines of the Health Insurance Portability and Accountability Act for patient privacy. Currently, facial feature deidentification is a challenging task given lack of expert consensus and lack of testing infrastructure for adequate automatic and manual facial image detection. Objective This study aimed to review the current literature on automatic facial deidentification algorithms and to assess their utility in dermatology use cases, defined by preservation of skin attributes (redness, texture, pigmentation, and lesions) and data utility. Methods We conducted a systematic search using a combination of headings and keywords to encompass the concepts of facial deidentification and privacy preservation. The MEDLINE (via PubMed), Embase (via Elsevier), and Web of Science (via Clarivate) databases were queried from inception to May 1, 2021. Studies with the incorrect design and outcomes were excluded during the screening and review process. Results A total of 18 studies, largely focusing on general adversarial network (GANs), were included in the final review reporting various methodologies of facial deidentification algorithms for still and video images. GAN-based studies were included owing to the algorithm’s capacity to generate high-quality, realistic images. Study methods were rated individually for their utility for use cases in dermatology, pertaining to skin color or pigmentation and texture preservation, data utility, and human detection, by 3 human reviewers. We found that most studies notable in the literature address facial feature and expression preservation while sacrificing skin color, texture, pigmentation, which are critical features in dermatology-related data utility. Conclusions Overall, facial deidentification algorithms have made notable advances such as disentanglement and face swapping techniques, while producing realistic faces for protecting privacy. However, they are sparse and currently not suitable for complete preservation of skin texture, color, and pigmentation quality in facial photographs. Using the current advances in artificial intelligence for facial deidentification summarized herein, a novel approach is needed to ensure greater patient anonymity, while increasing data access for automated image analysis in dermatology. Conflicts of Interest None declared.

Attitudes Toward Artificial Intelligence Among Dermatologists in Morocco: A National Survey

Background Artificial intelligence (AI) is a hot topic, and the use of AI in our day-to-day lives has increased exponentially. AI is becoming increasingly important in dermatology, with studies reporting accuracy matching or exceeding that of dermatologists in the diagnosis of skin lesions from clinical and dermoscopic images. However, little is known about the attitudes of dermatologists in Morocco toward AI. Objective The purpose of this cross-sectional study was to evaluate the attitudes of dermatologists in Morocco toward AI. Methods An online survey was distributed through Google Forms (Google LLC) to dermatologists in Morocco and was open from January to June 2021. Statistical analysis of the data collected was performed using Jamovi software. Any association for which the P value was <.05 was considered statistically significant. Results In total, 183 surveys were completed and analyzed. Overall, 79.8% of respondents were female, and the median age was 35 years (IQR 25-74 years). A total of 30.6% stated that they were not aware of AI, and 34.4% had a basic knowledge of AI technologies. Only 7.7% of the respondents strongly agreed that the human dermatologist will be replaced by AI in the foreseeable future. Of the entire group, 61.8% agreed or strongly agreed that AI will improve dermatology, and 70% thought that AI should be part of medical training. In addition, only 32.2% reported having read publications about AI. Female dermatologists showed more fear pertaining to the use of AI within dermatology (P=.01); this group also suggested that AI has a very strong potential in the detection of skin diseases using dermoscopic images (P=.03). Conclusions Our results demonstrate an overall optimistic attitude toward AI among dermatologists in Morocco. The majority of respondents believed that it will improve diagnostic capabilities. Conflict of Interest None declared.

Trends in Teledermatology Utilization in the United States

Background Teledermatology is an effective health care delivery model that has seen tremendous growth over the last decade. This growth can be attributed to a variety of factors, including but not limited to an increased access to dermatologic care for those with socioeconomic or geographic barriers, a reduction in health care costs for both the patient and the physician, and the delivery of high-quality dermatologic care. However, the associated barriers include practice reimbursements, interstate licensing, and liability. Despite these apparent barriers, the emergence of COVID-19 afforded teledermatology a surge of demand and loosened regulations, allowing dermatologists to see higher volumes of teledermatology patients. In this paper, we analyzed the American Academy of Dermatology’s DataDerm registry teledermatology utilization and patient demographic trends throughout the COVID-19 pandemic. Objective The aim of this paper was to characterize national-level teledermatology demographic data in the setting of the COVID-19 pandemic. Methods National-level data were curated for all practices enrolled in the American Academy of Dermatology’s DataDerm registry from April 1, 2020, through June 30, 2021. Encounter utilization rates were collected for visit type (ie, teledermatology versus in person), sex, race, age, insurance provider, and location (ie, in state versus out of state). The aggregate total data, as opposed to individual encounter data, were collected. Results The proportion of women who utilized services via teledermatology (65,023/98,642, 65.9%) was greater than that of those who utilized in-person services (29,40,122/50,48,450, 58.2%). Non-White patients made up a higher percentage of teledermatology utilizers (8920/62,324, 15%) when compared with in-person utilizers (3,94,580/35,08,150, 11.7%). Younger patients (aged <40) contributed more to teledermatology service utilization (62,695/75,319, 83.2%) when compared with in-person services (13,29,218/33,01,175, 40.3%). Medicare was a larger payor contributor for in-person services (8232/1,53,279, 25.2%) than for teledermatology services (10,89,777/43,30,882, 5.4%). Utilization by out-of-state patients was proportionally higher for teledermatology services (19,422/1,33,416, 14.6%) compared with in-person services (5,80,358/1,38,31,400, 4.2%). Conclusions Teledermatology services may reach and benefit certain populations (female, younger patients, those with non-White racial backgrounds, and out-of-state patients) more so than others. These baseline demographics may also serve to highlight populations for potential future teledermatology outreach efforts. Conflict of Interest None declared.

The Clinical Utility of a Handheld Elastic Scattering Spectroscopy Tool and Machine Learning in the Diagnosis and Referral Management of Skin Cancer by Primary Care Physicians

Background Elastic scattering spectroscopy (ESS) is a noninvasive optical biopsy technique that can distinguish between normal and abnormal tissue in vivo. The handheld device measures ESS spectra of skin lesions and classifies lesions with an output of “Investigate Further” or “Monitor.” The algorithm was trained and validated with over 11,000 spectral scans from over 3500 skin lesions. The device performance was also evaluated in an associated clinical study. Objective The aim of this paper was to establish whether the use of a handheld ESS tool can improve the detection of skin malignancies by evaluating clinical performance while emulating a real-world telemedicine clinical care setting. Methods The associated clinical study examined an independent test set of 332 lesions in a prospective multicenter study that compared algorithm performance to biopsy results for diagnosing malignant lesions. A total of 50 cases were randomly selected from the study data base (25 malignant and 25 benign lesions). Device performance on these lesions had a 96% sensitivity. High-resolution digital images and the patient’s clinical information including prior skin cancer history, risk factors, and physical examination results were available for evaluation. A total of 57 primary care physicians participated in this study in 2 phases, the first phase with their standard-of-care diagnostic and the second phase regarding their evaluation with the device output. The physicians were educated on the ESS device before evaluating the cases in a random order. Case evaluation included the physician reporting their diagnosis, management decision, and confidence level without the device output in the first phase and with the device output in the second phase. The results were evaluated for sensitivity and specificity with confidence intervals. Results The diagnostic sensitivity of the readers without and with the use of the handheld ESS device increased significantly from 67% to 88% (P<.001). There was no significant difference in specificity at 40% and 53% (P=.05). The management sensitivity of the readers increased significantly with and without the use of the device, which, respectively, was 94% (91%-96%) and 81% (77%-85%) (P<.001), suggesting that the use of the device may reduce false negatives by 68%. Specificity was comparable for management decisions (P=.36) at 31% compared to 36% without the device. Conclusions The use of the handheld ESS device significantly improved diagnostic and management sensitivity over standard-of-care, with comparable specificity. While telemedicine has shown promise in many fields, studies have shown that in-person skin evaluation is superior to telemedicine evaluations; however, integration with this type of tool has the potential to improve early detection.

The Safety and Effectiveness of Elastic Scattering Spectroscopy and Machine Learning in the Evaluation of Skin Lesions for Cancer

Background Elastic scattering spectroscopy (ESS) is an optical biopsy technique that can distinguish between a normal and abnormal tissue in vivo without the need to remove it. The handheld device measures ESS spectra of skin lesions and classifies lesions as either malignant or benign with an output of “Investigate Further” or “Monitor,” respectively, with positive results accompanied by a spectral score output from 1 to 10, indicating how similar the lesion is to the malignant lesions the device was trained on. The algorithm was trained and validated with over 11,000 spectral scans from over 3500 skin lesions. Objective The purpose of this study was to evaluate the safety and effectiveness of the handheld ESS device in detecting the most common types of skin cancer. Methods A prospective, single-arm, investigator-blinded, multicenter study conducted at 4 investigational sites in the United States was performed. Patients who presented with skin lesions suggestive of melanoma, basal cell carcinoma, squamous cell carcinoma, and other highly atypical lesions were evaluated with the handheld ESS device. A validation performance analysis was performed with 553 lesions from 350 subjects with algorithm version 2.0. An independent test set of 281 lesions was selected and used to evaluate device performance in the detection of melanoma, basal cell carcinoma (BCC), and squamous cell carcinoma (SCC). Statistical analyses included overall effectiveness analyses for sensitivity and specificity as well as subgroup analyses for lesion diagnoses. Results The overall sensitivity of the device was 92.3% (95% CI: 87.1 to 95.5%). The sensitivity for subgroups of lesions was 95% (95% CI 75.1% to 99.9%) for melanomas, 94.4% (95% CI 86.3% to 98.4%) for BCCs, and 92.5% (95% CI 83.4% to 97.5%) for SCCs. The overall device specificity was 36.6% (95% CI 29.3% to 44.6%). There was no statistically significant difference between the dermatologist performance and the ESS device (P=.2520). The specificity of the device was highest for benign melanocytic nevi (62.5%) and seborrheic keratoses (78.2%). The overall positive predictive value (PPV) was 59.8%, and the negative predictive value (NPV) was 81.9% with the study’s malignancy prevalence rate of 51%. For a prevalence rate of 5%, the PPV was estimated to be 7.1%, and the NPV was estimated to be 98.9%. For a prevalence rate of 7%, the PPV was estimated to be 9.8%, and the NPV was estimated to be 98.4%. For a prevalence rate of 15%, the PPV was estimated to be 20.3%, and the NPV was 96.4%. Conclusions The handheld ESS device has a high sensitivity for the detection of melanoma, BCC, and SCC. Coupled with clinical exam findings, this device can aid physicians in detecting a variety of skin malignancies. The device output can aid teledermatology evaluations by helping frontline providers determine which lesions to share for teledermatologist evaluation as well as potentially benefitting teledermatologists’ virtual evaluation, especially in instances of suboptimal photo quality. Acknowledgments This study was sponsored by Dermasensor Inc. Conflicts of Interest None declared.

Use of Artificial Intelligence as a Predictor of the Response to Treatment in Alopecia Areata

Background Artificial intelligence (AI) has emerged in dermatology with some studies focusing on skin disorders such as skin cancer, atopic dermatitis, psoriasis, and onychomycosis. Alopecia areata (AA) is a dermatological disease whose prevalence is 0.7%-3% in the United States, and is characterized by oval areas of nonscarring hair loss of the scalp or body without evident clinical variables to predict its response to the treatment. Nonetheless, some studies suggest a predictive value of trichoscopic features in the evaluation of treatment responses. Assuming that black dots, broken hairs, exclamation marks, and tapered hairs are markers of negative predictive value of the treatment response, while yellow dots are markers of no response to treatment according to recent studies, the absence of these trichoscopic features could indicate favorable disease evolution without treatment or even predict its response. Nonetheless, no studies have reportedly evaluated the role of AI in AA on the basis of trichoscopic features. Objective This study aimed to develop an AI algorithm to predict, using trichoscopic images, those patients diagnosed with AA with a better disease evolution. Methods In total, 80 trichoscopic images were included and classified in those with or without features of negative prognosis. Using a data augmentation technique, they were multiplied to 179 images to train an AI algorithm, as previously carried out with dermoscopic images of skin tumors with a favorable response. Subsequently, 82 new images of AA were presented to the algorithm, and the algorithm classified these patients as responders and non-responders; this process was reviewed by an expert trichologist observer and presented a concordance higher than 90% with the algorithm identifying structures described previously. Evolution of the cases was followed up to truly determine their response to treatment and, therefore, to assess the predictive value of the algorithm. Results In total, 32 of 40 (80%) images of patients predicted as nonresponders scarcely showed response to the treatment, while 34 of 42 (81%) images of those predicted as responders showed a favorable response to the treatment. Conclusions The development of an AI algorithm or tool could be useful to predict AA evolution and its response to treatment. However, further research is needed, including larger sample images or trained algorithms, by using images previously classified in accordance with the disease evolution and not with trichoscopic features. Conflicts of Interest None declared.