Artificial Intelligence and Its Effect on Dermatologists’ Accuracy in Dermoscopic Melanoma Image Classification: Web-Based Survey Study (Preprint)

BACKGROUND Early detection of melanoma can be lifesaving but this remains a challenge. Recent diagnostic studies have revealed the superiority of artificial intelligence (AI) in classifying dermoscopic images of melanoma and nevi, concluding that these algorithms should assist a dermatologist’s diagnoses. OBJECTIVE The aim of this study was to investigate whether AI support improves the accuracy and overall diagnostic performance of dermatologists in the dichotomous image–based discrimination between melanoma and nevus. METHODS Twelve board-certified dermatologists were presented disjoint sets of 100 unique dermoscopic images of melanomas and nevi (total of 1200 unique images), and they had to classify the images based on personal experience alone (part I) and with the support of a trained convolutional neural network (CNN, part II). Additionally, dermatologists were asked to rate their confidence in their final decision for each image. RESULTS While the mean specificity of the dermatologists based on personal experience alone remained almost unchanged (70.6% vs 72.4%; P=.54) with AI support, the mean sensitivity and mean accuracy increased significantly (59.4% vs 74.6%; P=.003 and 65.0% vs 73.6%; P=.002, respectively) with AI support. Out of the 10% (10/94; 95% CI 8.4%-11.8%) of cases where dermatologists were correct and AI was incorrect, dermatologists on average changed to the incorrect answer for 39% (4/10; 95% CI 23.2%-55.6%) of cases. When dermatologists were incorrect and AI was correct (25/94, 27%; 95% CI 24.0%-30.1%), dermatologists changed their answers to the correct answer for 46% (11/25; 95% CI 33.1%-58.4%) of cases. Additionally, the dermatologists’ average confidence in their decisions increased when the CNN confirmed their decision and decreased when the CNN disagreed, even when the dermatologists were correct. Reported values are based on the mean of all participants. Whenever absolute values are shown, the denominator and numerator are approximations as every dermatologist ended up rating a varying number of images due to a quality control step. CONCLUSIONS The findings of our study show that AI support can improve the overall accuracy of the dermatologists in the dichotomous image–based discrimination between melanoma and nevus. This supports the argument for AI-based tools to aid clinicians in skin lesion classification and provides a rationale for studies of such classifiers in real-life settings, wherein clinicians can integrate additional information such as patient age and medical history into their decisions.

Download Full-text

Artificial Intelligence and Its Effect on Dermatologists’ Accuracy in Dermoscopic Melanoma Image Classification: Web-Based Survey Study

Journal of Medical Internet Research ◽

10.2196/18091 ◽

2020 ◽

Vol 22 (9) ◽

pp. e18091 ◽

Cited By ~ 1

Author(s):

Roman C Maron ◽

Jochen S Utikal ◽

Achim Hekler ◽

Axel Hauschild ◽

Elke Sattler ◽

...

Keyword(s):

Artificial Intelligence ◽

Personal Experience ◽

Real Life ◽

Survey Study ◽

Final Decision ◽

Web Based ◽

Additional Information ◽

Control Step ◽

Average Confidence ◽

The Mean

Background Early detection of melanoma can be lifesaving but this remains a challenge. Recent diagnostic studies have revealed the superiority of artificial intelligence (AI) in classifying dermoscopic images of melanoma and nevi, concluding that these algorithms should assist a dermatologist’s diagnoses. Objective The aim of this study was to investigate whether AI support improves the accuracy and overall diagnostic performance of dermatologists in the dichotomous image–based discrimination between melanoma and nevus. Methods Twelve board-certified dermatologists were presented disjoint sets of 100 unique dermoscopic images of melanomas and nevi (total of 1200 unique images), and they had to classify the images based on personal experience alone (part I) and with the support of a trained convolutional neural network (CNN, part II). Additionally, dermatologists were asked to rate their confidence in their final decision for each image. Results While the mean specificity of the dermatologists based on personal experience alone remained almost unchanged (70.6% vs 72.4%; P=.54) with AI support, the mean sensitivity and mean accuracy increased significantly (59.4% vs 74.6%; P=.003 and 65.0% vs 73.6%; P=.002, respectively) with AI support. Out of the 10% (10/94; 95% CI 8.4%-11.8%) of cases where dermatologists were correct and AI was incorrect, dermatologists on average changed to the incorrect answer for 39% (4/10; 95% CI 23.2%-55.6%) of cases. When dermatologists were incorrect and AI was correct (25/94, 27%; 95% CI 24.0%-30.1%), dermatologists changed their answers to the correct answer for 46% (11/25; 95% CI 33.1%-58.4%) of cases. Additionally, the dermatologists’ average confidence in their decisions increased when the CNN confirmed their decision and decreased when the CNN disagreed, even when the dermatologists were correct. Reported values are based on the mean of all participants. Whenever absolute values are shown, the denominator and numerator are approximations as every dermatologist ended up rating a varying number of images due to a quality control step. Conclusions The findings of our study show that AI support can improve the overall accuracy of the dermatologists in the dichotomous image–based discrimination between melanoma and nevus. This supports the argument for AI-based tools to aid clinicians in skin lesion classification and provides a rationale for studies of such classifiers in real-life settings, wherein clinicians can integrate additional information such as patient age and medical history into their decisions.

Download Full-text

POS0733 REAL-LIFE PHYSICAL ACTIVITY IN SLE PATIENTS: ASSOCIATIONS WITH FATIGUE AND QUALITY OF SLEEP

Annals of the Rheumatic Diseases ◽

10.1136/annrheumdis-2021-eular.1775 ◽

2021 ◽

Vol 80 (Suppl 1) ◽

pp. 617.1-617

Author(s):

H. Wohland ◽

N. Leuchten ◽

M. Aringer

Keyword(s):

Physical Activity ◽

Sleep Quality ◽

Disease Activity ◽

Lupus Erythematosus ◽

Real Life ◽

Survey Study ◽

Cross Sectional Survey ◽

Cross Sectional ◽

Fatigue Score ◽

The Mean

Background:Fatigue is among the top complaints of patients with systemic lupus erythematosus (SLE), but only in part associated with SLE disease activity. Physical activity can help to reduce fatigue and should therefore be recommended to SLE patients. Vice versa, fatigue may arguably lead to reduced physical activity.Objectives:To investigate the extent of physical activity and the perception of fatigue and sleep quality in patients with SLE.Methods:Starting in February 2019, SLE patients were invited to participate in a cross-sectional survey study of fatigue and physical exercise during their routine outpatient clinic visits. Participants filled out a ten-page paper questionnaire focused on physical activity. To evaluate fatigue, we primarily used a 10 cm visual analogue scale (0-100 mm, with 100 meaning most fatigued), but also the FACIT fatigue score (range 0-52). Sleep quality was estimated using grades from 1 (excellent) to 6 (extremely poor).Results:93 SLE patients took part in the study. All patients fulfilled the European League Against Rheumatism/ American College of Rheumatology (EULAR/ACR) 2019 classification criteria for SLE. 91% of the patients were female. Their mean (SD) age was 45.5 (14.3) years and their mean disease duration 12.1 (9.4) years. The mean BMI was 25.2 (5.6). Of all patients, 7.5% had a diagnosis of (secondary) fibromyalgia. The mean fatigue VAS was 32 (27) mm and the mean FACIT fatigue score 35.7 (10.3). As expected, fatigue by VAS and FACIT was correlated (Spearman r=-0.61, p<0.0001). The mean SLEDAI was 1 (1) with a range of 0 to 6. Median glucocorticoid doses were 2 mg prednisolone equivalent, with a range from 0 to 10 mg.Out of 66 patients in payed jobs, 64 (97%) reported details on their working space. One person (2%) worked in a predominanty standing position, 37 (58%) worked in essentially sedentary jobs and 26 (40%) were in positions where they were mildly physically active in part. The mean fatigue VAS was 31 (24) mm for patients with partly active jobs and 27 (30) mm for those in sedentary jobs. Sleep was graded 2.9 (0.9) by those with active and 3.1 (1.3) by those with sedentary jobs.Half of the patients (51%) reported more than one physical recreational activity. 44 (47%) were walking and for five persons (5%) this was the only form of activity. Cycling was reported by 19 patients (20%), 18 of whom also practiced other activities. For transport, 52 (56%) in part chose active modes, such as walking and cycling. Patients who reported any of the above activities showed a mean fatigue VAS of 28 (25) mm, compared to 36 (28) mm in the patient group without a reported activity. Sleep quality was very similar: 3.1 (1.2) and 3.2 (1.1) for more active and more passive patients, respectively.65 (70%) patients regularly practiced sports. Of these, 39 (60%) practiced one kind of sport, 15 (23%) two, 7 (11%) three, and 2 (3%) each four and five kinds of sports. Fatigue VAS of patients practicing sports was 27 (25) mm versus 43 (28) in those who did not (p=0.0075). Sleep quality was 2.9 (1.1) in the sports cohort and 3.5 (1.1) in the no-sports cohort (p=0.0244).Conclusion:A majority of SLE patients in remission or low to moderate disease activity regularly practiced sports, and those doing so reported lesser fatigue and better sleep quality. The absolute values on the fatigue VAS were in a moderate range that made fatigue as the main cause of not performing sports rather unlikely for most patients.Disclosure of Interests:Helena Wohland: None declared, Nicolai Leuchten Speakers bureau: AbbVie, Janssen, Novartis, Roche, UCB, Consultant of: AbbVie, Janssen, Novartis, Roche, Martin Aringer Speakers bureau: AbbVie, Astra Zeneca, BMS, Boehringer Ingelheim, Chugai, Gilead, GSK, HEXAL, Lilly, MSD, Novartis, Pfizer, Roche, Sanofi, UCB, Consultant of: AbbVie, Astra Zeneca, BMS, Boehringer Ingelheim, GSK, Lilly, MSD, Roche, Sanofi, UCB

Download Full-text

The Impact of COVID-19 on Surgical Education: A Monocentric Survey of Residents Training in Surgical Specialties

10.21203/rs.3.rs-602831/v1 ◽

2021 ◽

Author(s):

Nicolas Tassé ◽

Etienne Auger-Dufour

Keyword(s):

Resident Education ◽

Clinical Teaching ◽

Survey Study ◽

Quebec City ◽

Surgical Resident ◽

Web Based ◽

Surgical Residents ◽

The Mean ◽

The Impact ◽

Descriptive Statistical Analysis

Abstract BACKGROUND: This study aims to identify the effects of the COVID-19 on surgical resident education at University Laval during first wave of the pandemic of spring 2020.METHODS: We conducted a web-based survey study to all residents training within one of the ten surgical specialties at University Laval, Quebec City. The survey focused on clinical teaching hours, appreciation of activities and novelties experienced and the impacts of virtual teaching. Descriptive statistical analysis was performed to summarize the data.RESULTS: There were 48 surgical residents who responded to our survey. There were participants from ten specialties. During the pandemic the mean number of weekly teaching hours dropped from 4.31 to 3.69 hours. The most appreciated activity was teaching sessions lead by a staff surgeon. More than 80% of respondents reported having partaken in other activities at some time during an online class while over 70% expressed retaining less when material was taught online rather than in person.CONCLUSION: Our survey provides insight for surgical programs to improve resident teaching and illustrates the necessity to optimize teaching schedules rapidly in times of pandemic. Even though the appreciation of virtual learning seems unsatisfactory by certain residents, trainees still require and appreciate teaching by their mentors.

Download Full-text

Preparedness and attitude toward personal protective equipment among house officers during COVID-19 pandemic in Egypt

Journal of Public Health ◽

10.1093/pubmed/fdab078 ◽

2021 ◽

Author(s):

Asmaa Ahmed Sayed ◽

Marwa Mostafa Ahmed ◽

Inas Talaat Elsayed ◽

Soliman Saeed ◽

Alsallout Inas ◽

...

Keyword(s):

Personal Protective Equipment ◽

Odds Ratio ◽

Hospital Staff ◽

Survey Study ◽

P Value ◽

Protective Equipment ◽

Web Based ◽

The World ◽

The Mean ◽

Low Levels

Abstract Background Coronavirus disease 2019 (COVID-19) struck the world by surprise by the rising numbers that required prompt governmental and hospital staff reaction to the ongoing crisis. A robust preparedness and personal protective equipment (PPE) were yet to be regarded as our best plan. Methods A survey study was conducted on 254 Egyptian house officers using an anonymous web-based questionnaire that was filled using Google Forms after obtaining online informed consent. Results The mean age of the participants was 25 years. Only 28.74% of the house officers were categorized as having a good preparedness, while 85.83% of them have a good PPE attitude. The preparedness and willingness were significantly associated with the overall worry related to the pandemic (P value = 0.012). Fear of contracting COVID-19 infection negatively affected their preparedness by 60% (odds ratio (OR) 0.40, 95% confidence interval (CI), 0.17–0.93, P value = 0.034). The House officers with family members at-risk for severe COVID-19 were less likely to be prepared and willing by 70% (OR 0.30, 95% CI 0.15–0.60, P value = 0.001). The house officers with good preparedness and willingness to deal with COVID-19 seemed to have a good PPE attitude (OR 11.48, 95% CI 2.43-54.34, P value = 0.002). Conclusion A significant number of house officers expressed low levels of preparedness, while most of them have a good PPE attitude.

Download Full-text

One-year real-life results on effect of intravitreal aflibercept in patients with diabetic macular oedema switched from ranibizumab

European Journal of Ophthalmology ◽

10.1177/1120672120927275 ◽

2020 ◽

pp. 112067212092727

Author(s):

Marko Lukic ◽

Gwyn Williams ◽

Zaid Shalchi ◽

Praveen J Patel ◽

Philip G Hykin ◽

...

Keyword(s):

Visual Acuity ◽

Diabetic Retinopathy ◽

Early Treatment ◽

Macular Oedema ◽

Real Life ◽

Diabetic Macular Oedema ◽

Foveal Thickness ◽

Macular Volume ◽

Central Foveal Thickness ◽

The Mean

Purpose To assess visual and optical coherence tomography–derived anatomical outcomes of treatment with intravitreal aflibercept (Eylea®) for diabetic macular oedema in patients switched from intravitreal ranibizumab (Lucentis®). Design Retrospective, cohort study. Participants Ninety eyes (of 67 patients) receiving intravitreal anti–vascular endothelial growth factor therapy were included. Methods This is a retrospective, real-life, cohort study. Each patient had visual acuity measurements and optical coherence tomography scans performed at baseline and 12 months after the first injection of aflibercept was given. Main Outcome Measures We measured visual acuities in Early Treatment Diabetic Retinopathy Study letters, central foveal thickness and macular volume at baseline and at 12 months after the first aflibercept injection was given. Results Ninety switched eyes were included in this study. The mean (standard deviation) visual acuity was 63 (15.78) Early Treatment Diabetic Retinopathy Study letters. At baseline, the mean (standard deviation) central foveal thickness was 417.7 (158.4) μm and the mean macular volume was 9.96 (2.44) mm3. Mean change in visual acuity was +4 Early Treatment Diabetic Retinopathy Study letters (p = 0.0053). The mean change in macular volume was −1.53 mm 3 in SW group (p = 0.21), while the change in central foveal thickness was −136.8 μm (p = 0.69). Conclusion There was a significant improvement in visual acuity and in anatomical outcomes in the switched group at 12 months after commencing treatment with aflibercept for diabetic macular oedema.

Download Full-text

Aflibercept in real-life for the treatment of age-related macular degeneration using a treat and extend protocol: The Armada study

European Journal of Ophthalmology ◽

10.1177/11206721211005703 ◽

2021 ◽

pp. 112067212110057

Author(s):

Pierre Gascon ◽

Prithvi Ramtohul ◽

Charles Delaporte ◽

Sébastien Kerever ◽

Danièle Denis ◽

...

Keyword(s):

Macular Degeneration ◽

Real Life ◽

Age Related Macular Degeneration ◽

Cumulative Number ◽

Treat And Extend ◽

Treatment Interval ◽

Age Related ◽

Treatment Naïve ◽

The Mean

Purpose: To report the visual and anatomic outcomes in treatment-naïve neovascular age-related macular degeneration (nAMD) patients treated with aflibercept under a standardized Treat and Extend (T&E) protocol for up to 3 years of follow-up in “real-life” practice. Methods: This retrospective, observational, multicenter study included patients with treatment-naïve nAMD and at least 12 months of follow-up. T&E regimen adjustment was initiated after loading phase. At each visit best-corrected visual acuity (BCVA) and optical coherence tomography parameters were performed. Results: One hundred and thirty-six eyes of 115patients had at least 1 year of follow-up with 114 and 82 eyes completing at least 2 and 3 years of follow-up, respectively (mean follow-up duration: 2.7 ± 1.3 years). Mean age was 78.6 ± 8.6 years old and 52% were women. Mean BCVA increased from 60.6 ± 18.7 letters at diagnosis to 66.9 ± 16.2 letters at 1 year (+6.3 letters, p = 0.003) and remained stable throughout the follow-up period (63.1 ± 20.3 letters (+2.5, p = 0.1) and 64.0 ± 20.1 letters (+3.4, p = 0.27) at 2 and 3 years, respectively). The mean central retinal thickness decreased significantly from 358.2 ± 87.9 µm at baseline to 302 ± 71.7 µm at 12 months and maintained stable after 36 months of follow-up (297.1 ± 76 µm, p < 0.0001). Mean number of injections was 6.6 ± 2.2, 4.8 ± 1.9, and 5.6 ± 1.7 at 1, 2, and 3 years, respectively. Mean cumulative number of 16.4 ± 5.6 injections after 3 years. Mean treatment interval was 6.8 ± 2.5 weeks at 1 year. Eight-week and 12-week treatment interval were achieved in 59.5% and 19.1%, 65.8%, and 36.8% and 69.5% and 41.5% at 1, 2, and 3 years, respectively. Conclusions: Our study demonstrated that intravitreal injections of aflibercept initiated under a standardized T&E for patients with treatment-naïve nAMD allow for significant visual improvement at 12 months, which was maintained over a 3-year follow-up period.

Download Full-text

A Web-Based Rendering Application for Communicating Dental Conditions

Healthcare ◽

10.3390/healthcare9080960 ◽

2021 ◽

Vol 9 (8) ◽

pp. 960

Author(s):

Hudson D. Spangler ◽

Miguel A. Simancas-Pallares ◽

Jeannie Ginnis ◽

Andrea G. Ferreira Zandoná ◽

Jeff Roach ◽

...

Keyword(s):

Stainless Steel ◽

Latent Class ◽

Real Life ◽

Tooth Surface ◽

Caries Experience ◽

Visual Aids ◽

Web Based ◽

Surface Level ◽

Ceramic Crowns ◽

Dental Conditions

The importance of visual aids in communicating clinical examination findings or proposed treatments in dentistry cannot be overstated. Similarly, communicating dental research results with tooth surface-level precision is impractical without visual representations. Here, we present the development, deployment, and two real-life applications of a web-based data visualization informatics pipeline that converts tooth surface-level information to colorized, three-dimensional renderings. The core of the informatics pipeline focuses on texture (UV) mapping of a pre-existing model of the human primary dentition. The 88 individually segmented tooth surfaces receive independent inputs that are represented in colors and textures according to customizable user specifications. The web implementation SculptorHD, deployed on the Google Cloud Platform, can accommodate manually entered or spreadsheet-formatted tooth surface data and allows the customization of color palettes and thresholds, as well as surface textures (e.g., condition-free, caries lesions, stainless steel, or ceramic crowns). Its current implementation enabled the visualization and interpretation of clinical early childhood caries (ECC) subtypes using latent class analysis-derived caries experience summary data. As a demonstration of its potential clinical utility, the tool was also used to simulate the restorative treatment presentation of a severe ECC case, including the use of stainless steel and ceramic crowns. We expect that this publicly available web-based tool can aid clinicians and investigators deliver precise, visual presentations of dental conditions and proposed treatments. The creation of rapidly adjustable lifelike dental models, integrated to existing electronic health records and responsive to new clinical findings or planned for future work, is likely to boost two-way communication between clinicians and their patients.

Download Full-text

Improved Visual Localization via Graph Filtering

Journal of Imaging ◽

10.3390/jimaging7020020 ◽

2021 ◽

Vol 7 (2) ◽

pp. 20

Author(s):

Carlos Lassance ◽

Yasir Latif ◽

Ravi Garg ◽

Vincent Gripon ◽

Ian Reid

Keyword(s):

Image Retrieval ◽

Large Scale ◽

Visual Localization ◽

Localization Accuracy ◽

Additional Information ◽

Feature Representations ◽

The Mean ◽

Classical Image ◽

Vision Based Localization ◽

Improve Reliability

Vision-based localization is the problem of inferring the pose of the camera given a single image. One commonly used approach relies on image retrieval where the query input is compared against a database of localized support examples and its pose is inferred with the help of the retrieved items. This assumes that images taken from the same places consist of the same landmarks and thus would have similar feature representations. These representations can learn to be robust to different variations in capture conditions like time of the day or weather. In this work, we introduce a framework which aims at enhancing the performance of such retrieval-based localization methods. It consists in taking into account additional information available, such as GPS coordinates or temporal proximity in the acquisition of the images. More precisely, our method consists in constructing a graph based on this additional information that is later used to improve reliability of the retrieval process by filtering the feature representations of support and/or query images. We show that the proposed method is able to significantly improve the localization accuracy on two large scale datasets, as well as the mean average precision in classical image retrieval scenarios.

Download Full-text

Application of Artificial Intelligence (AI) for Sustainable Highway and Road System

Symmetry ◽

10.3390/sym13010060 ◽

2020 ◽

Vol 13 (1) ◽

pp. 60

Author(s):

Md Arifuzzaman ◽

Muhammad Aniq Gul ◽

Kaffayatullah Khan ◽

S. M. Zakir Hossain

Keyword(s):

Experimental Data ◽

High Performance ◽

Adhesion Force ◽

Model Development ◽

Real Life ◽

Bayesian Optimization ◽

Support Vector ◽

Adhesive Properties ◽

Bayesian Optimization Algorithm ◽

The Mean

There are several environmental factors such as temperature differential, moisture, oxidation, etc. that affect the extended life of the modified asphalt influencing its desired adhesive properties. Knowledge of the properties of asphalt adhesives can help to provide a more resilient and durable asphalt surface. In this study, a hybrid of Bayesian optimization algorithm and support vector regression approach is recommended to predict the adhesion force of asphalt. The effects of three important variables viz., conditions (fresh, wet and aged), binder types (base, 4% SB, 5% SB, 4% SBS and 5% SBS), and Carbon Nano Tube doses (0.5%, 1.0% and 1.5%) on adhesive force are taken into consideration. Real-life experimental data (405 specimens) are considered for model development. Using atomic force microscopy, the adhesive strength of nanoscales of test specimens is determined according to functional groups on the asphalt. It is found that the model predictions overlap with the experimental data with a high R2 of 90.5% and relative deviation are scattered around zero line. Besides, the mean, median and standard deviations of experimental and the predicted values are very close. In addition, the mean absolute Error, root mean square error and fractional bias values were found to be low, indicating the high performance of the developed model.

Download Full-text