Rare disease diagnosis: A review of web search, social media and large-scale data-mining approaches

Figures & data

Figure 1. Performance of different web based search tools on a query collection consisting of 56 queries from Ref. 6.

Table 1. Overview of the different diagnostic tools mentioned in the article

	Use	Purpose	Web page
FindZebra	Free text search with automatic symptom extraction and inference using Bayesian networks. Faceted search.	Diagnostic tool for rare diseases (for professionals)	www.findzebra.com
London Dysmorphology Database	Database on rare dysmorphic syndromes	Browsable database (for professionals)	www.lmdatabases.com/
OMIM	Free text search	Search for articles on ge-Netic diseases (for professionals)	www.omim.org
Phenomizer	Patient features and symptoms are input using a HPO (Human Phenotype Ontology) questionnaire. The system uses custom inference specialized for ontologies	Diagnostic tool for ge-netic diseases (for professionals)	compbio.charite.de/phenomizer
POSSUM	A dysmorphology database of multiple malformations, metabolic, teratogenic, chromosomal and skeletal syndromes and their images	Used for diagnosis and learning (for professionals)	www.possum.net.au
PubMed	Free text search	General medical information search (for professionals)	www.ncbi.nlm.nih.gov/pubmed
SimulConsult	Uses Bayesian inference to compile differential diagnosis	Diagnostic tool specialized for neurology and genetics (for professionals)	www.simulconsult.com/
Watson	Custom inferential system based on various statistical methods (see text for details)	General diagnostic tool (for professionals)	Not publicly accessible
WebMD	Search is performed using either free text search or by using a knowledge based symptom questionnaire system.	General diagnostic tool (for non-professionals)	www.webmd.com

Figure 2. Performance of FindZebra on each of the major disease categories in the Doctor's dilemma dataset.