Next Generation Neonatal Health Informatics with Artemis Carolyn McGregor a, , Christina Catley a , Andrew James b , and James Padbury c a University of Ontario Institute of Technology, Oshawa, ON, Canada b The Hospital for Sick Children, University of Toronto, Toronto, ON, Canada c Women & Infant's Hospital of Rhode Island, The Warren Alpert Medical School of Brown University, Providence, RI, USA carolyn.mcgregor@uoit.ca
Motivation: Earlier Onset Detection 15/11/06 16/11/06 Absolute times Diagnosis Baby 1 Diagnosis Baby 2 Diagnosis Baby 3 • Hourly spot readings from medical devices recorded on paper or electronic charts. • Babies weight between 2.5-3kg at discharge. Paper notes can weight 6kg 2
Motivation
Motivation: Earlier Onset Detection 15/11/06 16/11/06 Absolute times Diagnosis Baby 1 Diagnosis Baby 2 Diagnosis Baby 3 4
Motivation • Behaviour of physiological data streams that describe respiratory and cardiac function . . . – Pneumothorax (McIntosh et al, 2000) – Nosocomial infection (Griffin and Moorman, 2001) – Periventricular leucomalacia (Shankaran et al, 2006) – Intraventricular haemorrhage (Fabres et al, 2006; Tuzcu et al, 2009)
Knowledge Translation Challenge • Retrospective Analysis • Not implemented in clinical practice • Not scalable • Either or combination of: – Patient centric – Condition centric – Stream centric
Objectives • The provision of this knowledge requires a multidimensional approach: – multiple conditions – multiple streams of data – for which multiple behaviours can exist • In addition, integrate of – real-time synchronous medical device data – asynchronous clinical data
Care ¡provider ¡views Apnea Apnea Patients Diagnoses Apnoea Sepsis Patients Diagnoses Neonatologist Sepsis Patients Diagnoses Sepsis Hypoglycemia HR RR IRW SpO 2 Hypoglycemia Streams/Behaviours HR RR IRW SpO 2 Hypoglycemia WIHRI Streams/Behaviours HR RR IRW SpO 2 SickKids Streams/Behaviours UOIT Locations
Artemis SpO2 BP ECG HR MP50 Babylog8000 Online Analysis Result Data USER INTERFACE FA Presentation Medical AR CapsuleTech Aquisition EP QRS RR PT HR Source Op Server Data Alert Sink SpO2 Source Op Hub Op Sepsis BP BPA BP Source Op Configuration WTA WT CIS Source Op Server Clinical Information CIS Adapter InfoSphere Streams Runtime System Cognos Deployment Server Knowledge Extraction Data Integration Mgr SPADE IDE (Re)deployment Stream Persistency Ontology Driven Data Miner HIR Data Mover Stream Patient Rule Modifier Knowledge Extraction 9
SickKids (a) SpO2 BP ECG HR MP50 Babylog8000 Online Analysis Result Data USER INTERFACE FA Presentation Medical AR CapsuleTech Aquisition EP QRS RR PT HR Source Op Server Data Alert Sink SpO2 Source Op Hub Op Sepsis BP BPA BP Source Op Configuration WTA WT CIS Source Op Server Clinical Information CIS Adapter InfoSphere Streams Runtime System Cognos Deployment Server Knowledge Extraction Data Integration Mgr SPADE IDE (Re)deployment Stream Persistency Ontology Driven Data Miner HIR Data Mover Stream Patient Rule Modifier Stream Persistency Knowledge Extraction 10
SickKids (a) • Clinical research into new earlier onset detection of LONS. • 174 patients, representing 4.1 patient years of data. • Currently supporting eight concurrent patients and collecting approximately 1250 readings a second
Artemis Cloud WIHRI Artemis Cloud Define Monitor Web Web Service Service InfoSphere Streams Runtime Data Integration Manager FA ECG ECG AR Physiological HR HR EP QRS RR PT Stream HR Source Op SpO2 SpO2 Web Alert Sink SpO2 Source Op BP BP Op Service Sepsis BP BPA BP Source Op Patient Clinical WT WTA CIS Source Op Web Data Mover Service Deployment Server Knowledge Extraction TA Stream Ontology Driven Rules Temporal Rule Modifier Data Miner Patient TAs Clinical Rule Analyse Hospital Web Web Service Service McGregor, C., 2011, “A Cloud Computing Framework for Real-time Rural and Remote Service of Critical Care”, IEEE Computer Based Medical Systems, Bristol, UK, 6 pages CDROM
WIHRI • Clinical research into neonatal instability • Enrolled 203 patients, representing 10.6 patient years of data • Spot readings every minute
SickKids (b) SpO2 BP ECG HR MP50 Babylog8000 Online Analysis Result Data USER INTERFACE FA Presentation Medical AR CapsuleTech Aquisition EP QRS RR PT HR Source Op Server Data Alert Sink SpO2 Source Op Hub Op Sepsis BP BPA BP Source Op Configuration WTA WT CIS Source Op Server Clinical Information CIS Adapter InfoSphere Streams Runtime System Cognos Deployment Server Knowledge Extraction Data Integration Mgr SPADE IDE (Re)deployment Stream Persistency Ontology Driven Data Miner HIR Data Mover Stream Patient Rule Modifier Knowledge Extraction 14
SickKids (b) • Clinical research into new earlier onset detection of LONS. • Nearly two years of 30 second spot reading data • Obtained from 1151 patients
Conclusion • Artemis provides clinical decision support in a flexible and transparent manner and instantiate clinical knowledge into the information processing pathway. • This is in direct contrast to many CDSSs based on complex mathematical processing, such as artificial neural networks, which from the clinicians’ viewpoint operate as black boxes
Future Work • Artemis has quickly become ubiquitous as it is used to support the clinical research • Clinical result publication is pending • We are currently installing Artemis in two NICUs in China to support cross cultural clinical research • We expanding the clinical research studies
Artemis Clinical Partners Shenzhen Maternity & Child Health Hospital 深圳市 妇 幼保健院 Nepean Hospita Westmead Hosp
Funding Acknowledgements Canada TJ Watson Research Center, NY
Next Generation Neonatal Health Informatics with Artemis Carolyn McGregor a, , Christina Catley a , Andrew James b , and James Padbury c a University of Ontario Institute of Technology, Oshawa, ON, Canada b The Hospital for Sick Children, University of Toronto, Toronto, ON, Canada c Women & Infant's Hospital of Rhode Island, The Warren Alpert Medical School of Brown University, Providence, RI, USA carolyn.mcgregor@uoit.ca
Recommend
More recommend
