We build technologies that support Clinical Decision Support Systems (CDSS) based on patient record at the point of care. We use AI technology and up-to-date knowledgebase to provide suggestions on differential diagnosis, medication and treatment. We also help streamline clinical processes through mechanisms such as generating conditional order sets, patient-focused reports/forms/templates and contextually relevant reference information.
We build our technologies with interoperability and information exchange in mind. As such, our knowledgebase and implementations are fully based on semantic ontologies and clinical information models such as HL7 CDA and FHIR resources.
Natural Language Processing (NLP): We use natural language processing pipeline to identify concepts and extract relationships between concepts in medical journals and articles to be imported into our knowledgebase. This can also be used to extract structured information from unstructured patient record such as discharge summaries for example.
Concepts are identified using concept unique identifier (CUI) from Unified Medical Language System (UMLS) meta-thesaurus. Relationships are extracted as subject-predicate-object triples. E.g.
(C0036421: Systemic Scleroderma)-[causes]->
(C0014866: Esophageal Stenosis).
Knowledge Graph: A graph is the best way to model knowledge in medical domain and relationships between medical concepts. We use a graphical db to build this knowledge store and run graph compute algorithms to help answer interesting questions and build models for decision support.
We can further augment the knowlege graph by linking it to external resources such as ontology based semantic network or publicly available linked open data for example or even genomic data.
Ontology and Semantic Interoperability: Adopting Unified Medical Language System (UMLS) meta-thesaurus to represent concepts in knowledge graph and result sets, we ensure interoperability of our inference engines and helps us integrate our decision support tools seamlessly with existing processes and workflows.
Inference Engine: A probabilistic inference engine based on a growing knowledge base is used to provide decision support as opposed to a deterministic rule based inference engine. As a result, we have a CDSS that learns and grows with the underlying knowledgebase and patient data.
We use AI and machine learning techniques to automatically learn inference rules and optimize/train our engine for particular speciality/domain - e.g. oncology etc.
Ravi specializes in healthcare informatics and ontologies. He has over 10 years of experience in software design and development. He has worked in various industries including fin-tech and online marketing.