Dermatology
Dermatologists are physicians who treat adult and pediatric patients with disorders of the skin, hair, nails, and adjacent mucous membranes. They diagnose everything from skin cancer, tumors, inflammatory diseases of the skin, and infectious diseases. They also perform skin biopsies and dermatological surgical procedures.
Subspecialties within the dermatology field include the following:
-
Dermatopathology
-
Pediatric dermatology
-
Procedural dermatology
1. Data Sources
1.1 Academic Journals
-
Journal of the American Academy of Dermatology
-
British Journal of Dermatology
-
Journal of Investigative Dermatology
-
Journal of the European Academy of Dermatology and Venereology
-
JAMA Dermatology
-
American Journal of Clinical Dermatology
-
Experimental Dermatology
-
Acta Dermato-Venereologica
-
Journal of Dermatological Science
-
The Journal of Dermatology
-
Contact Dermatitis
-
Clinics in Dermatology
-
Dermatologic Clinics
-
International Journal of Dermatology
-
Dermatologic Surgery
-
Dermatology and Therapy
-
Journal of Cosmetic Dermatology
-
Journal of Dermatological Treatment
-
Journal of the German Society of Dermatology: JDDG
-
Pediatric Dermatology
1.2 Textbooks
-
Dermatology: An Illustrated Colour Text
-
Dermatology A Pictorial Review
-
Ferri's Fast Facts in Dermatology: A Practical Guide to Skin Diseases and Disorders
-
Dermatology: Illustrated Study Guide and Comprehensive Board Review
2. NLP for Dermatology
2.1 Language Modeling
DermaBERT is a new transformer-based pre-trained language model deveoped for the specific field of Dermatology. A sample architecture to train a domain-centric language model from scratch is shown below.
2.2 Downstream Tasks
2.2.1 Question Answering
OxBioNLP's question answering component consists of six sub-components:
-
Extractive QA
-
Abstractive QA
-
Multi-choice QA
-
Explainable QA
-
Knowledge-Base QA
-
Medical Question Generation
2.2.2 Automated Fact Checking
OxBioNLP's automated fact checking component consists of four sub-components:
-
Claim Detection
-
Evidence Retrieval
-
Verdict Prediction
-
Justification Generation
2.2.3 Automation of Systematic Reviews
OxBioNLP's component for the automation of systematic reviews consists of four sub-components:
-
Abstractive Summarization
-
Extractive Summarization
-
Bias Detection
-
Federated Search
3. OxBioNLP for Dermatology
3.1 System Architecture
OxBioNLP's system architecture is consisting of multiple components organized into multiple layers.
Data sources are consisting of top-tier 20 dermatology journals and 5 textbooks which is given to the language modeling layer.
​
The output of T-PTLMs layer is a domain-centric transformer-based pre-trained language model; DermaBERT.
​
DermaBERT is used for the following downstream tasks: medical question answering, automated fact checking and automation of systematic reviews.
​
​
3.2 Use Cases
3.2.1 Medical Examinations Question Answering
3.2.2 Automation of Systematic Reviews
OxBioNLP's unsupervised information extraction component extracts fine-grained hierarchical entity mentions and categories. This makes it possible for research scientists to get access to the information using federated search.
