NICU. Has the future arrived? Anirudha Das MD Neonatologist Asst - - PowerPoint PPT Presentation

Artificial intelligence in the NICU…. Has the future arrived?

Anirudha Das MD Neonatologist Asst Professor of Pediatrics and Neonatology Cleveland Clinic Foundation Cleveland, Ohio, United States

Research interest in Artificial Intelligence?

Definition

Decision Support System (DSS): computer-based

application that collects, organizes and analyzes data to facilitate/assist healthcare professionals make clinical decisions.

Knowledge Based Data driven aims to capture the knowledge of human experts to support decision-making Processing large amounts of data to abstract patterns utilized for predictions,

• r emulate human behavior in similar

conditions in the past

(Malak et al., 2019) (Montani and Striani, 2019)

Decision support systems

Knowledge based:

• Top down approach
• Lack of knowledge in many areas
• r lack of management

consensus

• Transparency and explainability

is present Data driven:

• Bottoms up
• Recent surge in capability of

collecting large amounts of data

• Lack of transparency and

explainability

(Montani and Striani, 2019)

Decision Support Systems

Clinical Decision Support System (CDSS): computer-

based application that collects, organizes and analyzes data to facilitate/assist healthcare professionals make clinical decisions.

Prognosis Diagnosis Resource utilization Treatment Outcomes Cost estimation

(Malak et al., 2019)

Parameters: demographics/characteristics

• Maternal factors: Age, race, diagnosis, complications, GBS, previous

still birth, magnesium therapy, antenatal steroids, antepartum hemorrhage etc.

• Infant: Birth weight, gestation, sex, physiological scores (SNAPPE)

(CRIB II), weight gain

• Labs: Glucose, sodium, potassium, chloride, blood parameters,

cultures

• Management: Number of ventilated/oxygenated days, fluids, lipids,

dextrose, calorie intake including carbohydrates, proteins and fats

(Shillan et al., 2019)

Parameters: Vitals

TCO2 RR, BP and Temp Pulse oximeter Heart Rate aEEG NIRS

Data

• About 20,000 terabytes of data is being produced in the United states

each year

• Healthcare professionals have access to this data because they are

managing the patient

• There are limitations of organizing, processing and analyzing this data
• AI can help with this data!
• When models are built on the basis of quantitative data, objective

measures can be provided and cut-offs for patient management established

(SCCM, 2018)

Types of Artificial Intelligence

• Support vector machines
• Artificial neural networks
• Decision tree
• K-nearest newborns
• Gaussian mixture model
• Regression
• Rule based
• Fuzzy logic
• Multi agent-based approaches
• Biosensors

Seizures Plus ROP Jaundice Sepsis/NEC HIE NICU Diagnosis

Seizures Plus ROP Jaundice Sepsis HIE NICU Diagnosis

(Temko et al., 2011)

Seizures Plus ROP Jaundice Sepsis/NEC HIE NICU Diagnosis

(Peter Campbell et al., 2016)

Seizures Plus ROP Jaundice Sepsis HIE NICU Diagnosis

(Lake, Fairchild and Moorman, 2014)

Seizures Plus ROP Jaundice Sepsis HIE NICU Diagnosis

(Joshi et al., 2019)

Seizures Plus ROP Jaundice Sepsis/NEC HIE NICU Diagnosis

(Irles et al., 2018)

Mortality Discharge NICU Prognosis

Mortality Discharge NICU Prognosis

(Frize et al., 2005)

Mortality Discharge NICU Prognosis

(Podda et al., 2018)

Mortality Discharge NICU Prognosis

(Chaves and Nascimento, 2014)

(Morton, 2017)

Mechanical Ventilation Apnea monitoring Brain maturity Pain NICU Monitoring

(Morton, 2017)

Mechanical Ventilation Apnea monitoring Brain maturity Pain NICU Monitoring

Our project:

Data captor application Data captor server Interface Engine Data converted to HL-7 format EMR Repository Main Campus Fairview Hillcrest Beachwood

Beachwood