1 Linguamatics 2013 Linguamatics NLP for Electronic Health Records
2 Linguamatics 2013 Click Agenda to to edit edit Master Master title style title style Linguamatics and I2E Overview Customer case studies Application examples I2E Demonstration
Click Linguamatics to to edit edit Master Master Overview title style title style Formed in 2003 Based on technology developed at SRI, which also produced Apple Siri Leader in text mining/nlp in Life Sciences Pharmaceuticals: 16 of top 20 at every stage of drug discovery Government institutions: Workin with the FDA and divisions of the NIH Leading Healthcare Institutions in a variety of applications 40-50% growth year on year 3 Linguamatics 2013
4 Linguamatics 2013 Click EHRs: to to edit A edit Vital Master Master Opportunity title style title style to Improve Patient Care Not for profit Approvers Funders Payers Prescribers ICT The challenge is to unlock the value of the huge investment being made in EHRs Education Electronic Health Record Medical devices Government Providers Research Patient Dispensers In order to arrive at the depth of understanding Biotech they need from analytics, healthcare organizations will need to integrate unstructured data Pharma IDC Health Industry Insights
@Linguamatics 2013 Click Most to to people edit edit Master Master familiar title style title with style searching key terms Breast Cancer All these documents contain the keywords breast cancer. Read ALL the document to find the relevant bit to you
Efficient review, without reading every document @Linguamatics 2013 Click Text to Mining to edit edit Master Master Approach title style title style All human genes, and all ways of referring to each gene Any relationship between genes and diseases All Classes of Breast Cancer NLP uses the grammar within the sentence to extract precise results Use of terminologies, clustered results based on preferred terms
Click Linguamatics, to to edit edit Master Master Transforming title style title styletext into Patient Insights Turn this Into this To drive these Data warehouse Analytics Accurate results small focussed results set Complete results comprehensive and systematic 8 Linguamatics 2013 Company Confidential
9 Linguamatics 2013 Company Confidential Click Introduction to to edit edit Master Master to title Natural style title Language style Processing Groups words into meaningful units Morphology allows search for different forms of words sentences noun phrases verb groups morphology - match entities match actions different forms We find that p42mapk phosphorylates c-myb on serine and threonine. Purified recombinant p42 MAPK was found to phosphorylate Wee1.
11 Linguamatics 2013 Company Confidential Click Linguamatics to to edit edit Master Master I2E title style title style Decision Support Structured Results (Actionable Information) Agile NLP Querying Actionable Information Ad hoc queries Smart queries Multi queries Batch queries NLP Class, concept Reg exp Disambiguation Indexing Domain Knowledge Documents Flexible, Highly Scalable Rich Indexes Domain knowledge ontologies, thesauri, coding standards Internal/External Sources
@Linguamatics 2013 Click I2E is to to used edit edit Master in Master a title variety style title of style applications Vocabulary Development Gene Profiling Target Identification & Prioritization Safety/Tox Sentiment Analysis in Social Media Mining FDA Drug In the Labels beginning... Competitive IntelligenceProtein- Protein Interactions Workflow Integration Conference Abstract Mining Key Opinion Leader Identification Biomarker Discovery Mining Electronic Medical Records Clinical Trial Analysis Mutations and Gene Expression In-licensing Opportunities Patent Analysis Extracting Numerical and Experimental Data Systems Biology Drug Repositioning
13 Linguamatics 2013 Company Confidential Customer Case Studies
14 Linguamatics 2013 Confidential Click Linguamatics to to edit edit Master Master Healthcare title style title style Use Cases Pathology, radiology, initial assessment, discharge, check up Structured data Patient characteristics FDA Drug labels Patient characteristics Electronic Health Record Enterprise Data Warehouse Potential adverse drug reactions Patient characteristics Scientific literature Clinical case histories and/or genomic interpretation Patient characteristics Care gap models Patient lists Matching Clinical trials Clinical Trials. gov
15 Linguamatics 2013 Huntsman Cancer Institute: Click to to edit edit Master Master title style Turning Pathology Reports title style into Structured Data Challenge Solution Benefit Valuable patient data and insights about malignant tumours trapped in pathology reports, prohibit use in medical research and clinical trial selection Linguamatics I2E used to automatically identify and extract tumour facts from text, using advanced NLP capabilities and medical ontologies Pathology data now systematically extracted from pathology reports and loaded into data warehouse for improved medical research
16 Linguamatics 2013 Payer/Provider: Click to to edit edit Master Master title style Predicting Pneumonia title from style Radiology Reports Challenge Solution Benefit Diagnosis of pneumonia is a complex procedure requiring assessment of detailed radiologist s reports In collaboration with Linguamatics and I2E, this Payer/Provider has constructed a model that predicts pneumonia s presence or absence at 93% accuracy More guided diagnosis is allowing radiologists to focus on the uncertain cases and enables faster treatment and reduced risk of side effects
Georgetown University: Click to to edit edit Master Master title style Real-time Clinical Decision title style Support Using ipads Challenge Solution Benefit 17 Published literature provides valuable insights into disease comorbidity and treatments from past case histories. These are complex questions that can not be easily answered, causing delays in treatment decisions Linguamatics 2013 Georgetown University and Linguamatics have developed an application that enables rapid identification of case histories from PubMed during hospital rounds This rapid access to relevant data can save days and allows faster decisions to be made, leading to improved patient outcomes
18 Copyright Linguamatics 2013 Company Confidential Cardiovascular Disease: Nuclear Stress Test Assessment Click to to edit edit Master Master title style title style Challenge Solution Benefit Nuclear stress tests are a standard part of cardiovascular pathways but may be overlooked resulting in higher patient risk and lost re-imbursement Linguamatics are jointly developing a solution that automatically identifies patients that should be tested based on risk factors identified in their notes Cardiovascular patients are properly monitored, risks identified early and hospitals are reimbursed
20 Copyright Linguamatics 2013 Company Confidential Application Examples
Click Multiple to to edit edit Ways Master Master title of Interacting style title style Knowledge extraction Patient profile Specific parameters Tumour size, change in size, etc Full extraction Aspiration score (cytology), Peripheral blood smear, Reticulin fibrosis, Diagnosis, Core biopsy, Immunochemistry strain, Cytogenetic, Stain, Symptoms, Chemotherapy, Patient history queries Join to other systems EHR to adverse events EHR to clinical trials 21 Linguamatics 2013
22 Copyright Linguamatics 2013 Company Confidential Click Profiling to to edit edit Patients Master Master title from style title Doctors style Notes Collect pieces of information from the records for the patient across discharge summaries, consultations, radiology reports etc. Present as a summary with links to the evidence within the documents themselves for fast access to relevant information
Click Making to to edit edit Precise Master Master title Distinctions style title style for Mining and Coding Distinguish family history of disease/history of disease Distinguish history of disease/no history/family history Copyright Linguamatics 2013 Company Confidential 23
24 Copyright Linguamatics 2013 Company Confidential Click to edit Master title style Click Change to edit from Master Previous title styleobservation Multiple linguistic constructs Developed vocabulary list of terms to consider
25 Copyright Linguamatics 2013 Company Confidential Click to edit Master title style Click Extract to edit Dimensions Master title style in Context Extracted multiple dimensions and different units in same statement
26 Copyright Linguamatics 2013 Company Confidential Click to to edit edit Master Master title style title style Match Patients to Relevant Clinical Trials Find conditions of patients Link patients with the condition to appropriate trials Extract inclusion and exclusion conditions Patient Record Information Clinical Trial Information
27 Copyright Linguamatics 2013 Company Confidential Click Identify to to edit Potential edit Master Master title Adverse style title Events style Mine EHRs to identify condition and treatment combinations Identify patients with conditions that are known side effects of the drugs they are taking, based on contraindications in FDA label data e.g. rhinitis and dyspnea
28 Copyright Linguamatics 2013 Company Confidential Click Identifying to to edit edit Master Potential Master title Safety style title Issues style Compound Potential safety issues In this organ At this dosage
Click Translational to to edit edit Master Master Medicine title style title style Connect the patient condition to known information interpretation of sequencing results from research literature genes and known mutations find known adverse events, issues with co-medication 29 Copyright Linguamatics 2013 Company Confidential
30 Copyright Linguamatics 2013 Company Confidential Click to to edit edit Master Master title style title style Identify Alternate Treatments, Off Label Uses Discover indirect associations across disease, genes and drug interactions
31 Copyright Linguamatics 2013 Company Confidential I2E Demonstration