HOW WILL BIG DATA AFFECT RADIOLOGY (RESEARCH / ANALYTICS)? Ronald Arenson, MD
DEFINITION OF BIG DATA Big data is a broad term for data sets so large or complex that traditional data processing applications are inadequate. Challenges include analysis, capture, data curation, search, sharing, storage, transfer, visualization, and information privacy. The work instead requires "massively parallel software running on tens, hundreds, or even thousands of servers" Jacobs, A. (6 July 2009). "The Pathologies of Big Data". ACMQueue.
BIG DATA IN HEALTH CARE Genomics and proteomics Phenotype information Electronic Health Record Includes pathology, cytology, and lab Medical images Scientific literature Important components are changes over time
IMPORTANT CONCEPTS ABOUT BIG DATA All but literature are specific for patients Important that data is anonymized yet reversible Data security and confidentiality a top priority Data must be verified and dated with the ID of responsible owner Must be maintained once verified Data must be carefully defined, formatted, and precise EHR data notoriously free text and not structured
AUTHORIZED USERS Users must be trained on proper use of data Can not let anyone access data without careful training and monitoring Needs audit trail of accesses, modifications, and uses Processes for copyrights and patents are important Must manage who publishes where
VALUE OF BIG DATA With big data it is possible to correlate health information such as similar diagnoses, findings, genetics, clinical presentations, response to therapies, outcomes, prognoses, etc. Must be careful not to assume casual relationships Does not establish cause and effect
QUALITY OF DATA EHR full of inaccurate information, free text, conjecture, assumptions, not all proven diagnoses Health care vocabulary not precise with many synonyms with varying overlapping definitions Patients are not identified the same way across health systems EHRs also not uniformly used across health care systems or doctors Difficult to assume population statistics
PHYSICIAN USE OF EHR S No doubt that EHRs have improved data collection, communication and billing In general, EHRs require physicians to do more work on the computer Distraction from patient interaction Patient email curse
IMPLICATIONS OF BIG DATA FOR IMAGING Image Analysis is getting more sophisticated Using big data, researchers are exploring new techniques similar to Neural Networks Lessons from oil and space exploration and military applications Not blinded by knowledge Now called Deep Learning Research applications Image analysis and correlation with other data True outcomes evaluations for imaging Molecular imaging and diagnostics
DEEP LEARNING Deep learning (deep machine learning, or deep structured learning, or hierarchical learning, or sometimes DL) is a branch of machine learning based on a set of algorithms that attempt to model high-level abstractions in data by using model architectures, with complex structures or otherwise, composed of multiple non-linear transformations. From Wikipedia
CLINICAL APPLICATIONS Better decision support for order entry Improved quantitative imaging More Computer Assisted Diagnoses (CAD) Example: CureMetrix for breast cancer detection Computer assisted radiology (CAR) Form of structured reporting to help radiologist create report that is most useful to referring MD Precision in diagnoses Need real-time correlation with other data
BREAKING NEWS Helping Watson see : IBM plans to acquire Merge for $1B
SUMMARY Big Data is transforming health care Implications for Imaging Research Image analysis, outcomes and molecular imaging Clinical Decision support, quantitative imaging, CAD and CAR Problems include poor quality data, difficulty identifying patients, massive data, security and management