Hackathon Research for Team# 30

1. Telemedicine is effective in patients with three of the most common skin conditions diagnosed in dermatology. It has not only proven to be an effective diagnostic tool, but also has been shown to be effective for monitoring the treatment progression of these diseases. 
   A large study from Brazil published in Frontiers assessed the concordance between inflammatory dermatoses  diagnoses made by in-person dermatologists and teledermatologists. The most common  inflammatory dermatoses diagnosed by teledermatologist was Dermatophytosis. Additional frequent diagnosis were acne, chloasma and Atopic Dermatitis.

A smaller study compared diagnostic agreement between telemedicine on social networks (Twitter and MedPics) and standard teledermatology services. The common diagnosis were purpura (8.3%), eczema (6.7%), mycosis (6.7%), and viral infections (6.7%). In addition, they concluded that diagnostic agreement using social network images may serve as a reliable telemedicine tool.

      2. Telemedicine may be described as a modern technology supporting health care at a distance. Dermatology, as a visually-dependent specialty, is particularly suited for this kind of the health care model.  

Poor image quality is a significant issue in teledermatology. A study from September 2022,  introduced ImageQX, a first of its kind explainable image quality assessor which leverages domain expertise to improve the quality and efficiency of dermatological care in a virtual setting. The authors found that around 20% of the images collected through the mobile application were labeled as poor quality by dermatologists.

A report from JAMA Dermatology reviewed 2915 patient-submitted images. Fifty five precepts of the images were useful for medical decision-making and 62.2% were of sufficient quality. The study results suggest that images are most likely to be useful when they are in focus and reviewed by experienced attending physicians for wound surveillance.

The recommended standards for teledermatology are detailed in the ‘QUALITY STANDARDS FOR TELEDERMATOLOGY’, published by 

In conclusion, by reviewing the literature in the field of teledermatology, we can conclude that good image quality is necessary for the diagnosis of skin conditions in dermatology.   

      3. Several models for diagnosis of skin pathologies exist and potentially can be used in telemedicine. The majority is still under research and yet to be in everyday clinical use. 

Study evaluated an artificial intelligence (AI)–based tool that assists with diagnoses of dermatologic conditions. Artificial intelligence assistance was associated with improved diagnoses by primary care physicians and nurse practitioners. The benefit could be quantified as an improved diagnosis of 1 in every 8 to 10 cases.

Another study from 2022 evaluated the utility of (AI) in telemedicine triage and diagnosis of malignant lesions. 100 images were presented to AI software and to three dermatologists and then compared the diagnosis to the biopsy results. The AI correctly identified 63% of the cases, a similar success rate as the dermatologists.

Study used  convolutional neural networks (CNN) and artificial neural networks (ANN)  for melanoma detection based on dermoscopic images. The accuracy for CNN+ANN model was high (92.34%). 

Another study evaluated the ability of AI to detect skin lesions in unprocessed clinical photographs. The algorithm reviewed 673 patients with different skin pathologies (185 malignant, 305 benign, and 183 normal conditions) and was able to localize and diagnose skin cancer without manual preselection of suspicious lesions by dermatologists.

In this study, CNN had a similar sensitivity and higher specificity in detecting melanoma in dermoscopic images.

The use of models in dermatology and teledermatology is not limited only to malignant conditions. A group from France developed an algorithm for acne grading from smartphone photographs. 

In conclusion, we could not find models that interpret pathological findings from dermosope image. The existing models have high accuracy in the diagnosis of malignant and non-malignant skin conditions by dermosope images, but AI algorithms analyzing pathologic data only work on laboratory slides