The Medical Knowledge Discovery Project carried out preliminary research over a period of 3 years to identify the methods and structure by which a statistical analysis and profile can be created of web based health and medical information and make it usefull in the daily lifes of patients, physicians, clinicians, researchers, students, and healthcare professionals. The preliminary research also set out to define the structure of web-based Medical Knowledge Applications that assists the user in accessing, identifying, and interpreting World Wide Web health and medical information to suite his particular needs.
The following preliminary study papers are available in PDF format.
Mapping Medical Ontologies created a Medical Knowledge Base of 700 diseases and disorders in 5 classes from the Disease Ontology. Emphasis is given to 42 cardiovascular, 206 skin diseases (68 melanoma and 138 skin cancer disorders), 65 diabetes mellitus (endocrine and metabolic), 220 hematology and oncology (including 37 cancer and 23 leukemia disorders), and 154 musculoskeletal disorders. The study mapped 27 prescription medications from 19 categories to the Medical Knowledge Base.
Web Medical Information research created a “disease information profile” database and statistical analysis for 700 diseases in 5 major disease classes (cardiovascular, diseases of the skin, endocrine & metabolic, hematology & oncology, musculoskeletal). For each of these 700 diseases 6 search engine statistics were recorded to create a disease information profile (web page references, average references per disease class, images, videos, sponsored links, rare diseases). The study identified an average of 1 Million web pages per disease, 99% have an average of 15,000 images, and 44% have an average of 55 videos.
Medical Context Descriptor research identified the methodology to define and map terms that are commonly used to describe the context of web-based disease, medication, and gene information and knowledge. For the research, the matrix database of 500 combinations of medical context descriptors for 10 diseases was created and measured using Google search.
Identifying, qualifying, and ranking medical web site research analyzed 3,500 web page references to 7 cardiovascular, 7 diabetes, and 6 skin cancer disease searches. The analysis identified 9 categories of medical web sites: general health & medical, medical publications, research articles, medical libraries, medical associations, healthcare associations, medical schools and academic, government, and healthcare & pharmaceutical industry. Within each category, the web sites can be ranked by measuring their Medical Information Density per disease class and entity. The 90 web sites with the highest Medical Information Density cover 2,100 or approximately 2/3 of the 3,500 researched web pages.
Web based clinical application research analyzed 11 free text medical symptom searches that described 2 cancer symptoms, 4 cardiovascular disease symptoms, 2 diabetes symptoms, and 2 musculoskeletal disease symptoms. For each symptom search 60 web pages were analyzed for disease references. On the average, each search identified 32 references to 7 diseases with a primary disease identified by 13 references or 32% of the references.
Ontology Views research investigated the methods and structures that would enable user groups to define different representations and navigation of the Disease Ontology to meet their needs while taking advantage of the disease relationships and definitions of the ontology. The study compared the mappings of two diseases, abdominal aortic aneurysm and basal cell carcinoma between the Disease Ontology and the Merck Medical Library.
