1. Sven Van Poucke, MD, PhD
RIP “one-size-fits-all”:
welcome to the new era of personalized medicine
Wisdom 2020 Boston
2. 2
Maria, aged 74 years admitted to the hospital (ED)
Reason for encounter:
Signs and Symptoms: weight loss 25 pounds last 2
months, flu-like symptoms, yellowish discoloration of
the white area of the eye (sclera) and the skin, pale
stool, dark urine….
Jaundice may indicate a serious problem with the function of:
your red blood cells, liver, gallbladder, or pancreas.
Most likely causes (differential diagnosis):
-hepatitis
-alcohol
-auto-immune disorders
-rare genetic metabolic defects
-medicines, including acetaminophen toxicity, penicillins, oral
contraceptives, estrogenic or anabolic steroids etc.
-gallstones
-inflammation of the gallbladder
-gall bladder cancer
-pancreatic cancer
3. EHR EHR-LOINC codes
Clinical Biologist
Initial laboratory evaluation
Surgeon/Anesthesiologist
Internal Medicine
Curative or palliative treatment
ED Physician
Perform history &
physical examination
ICU Specialist
Critical care
Researcher
Where can we improve...
Dicom/(AI)
Radiologist
Ultrasonography or
computed tomography
EHR, Decision Support Multivariate Time Series Filling the gap between data
generated and data used
Data&Medicine Diagnostics Treatment
Uncertain observations
+changing over time
3
AI/ML/Process
5. 5
A Generalized Linear Model (GLM) (RapidMiner, Inc
AutoModel extension), was developed (with standard
parameter settings) including automatic feature
selection and generation providing a predictive model,
weights related to the most predictive attributes, the
performance of the model, a simulator, the
performance of the model, the predictions, prediction
chart and production model, weights by correlation and
correlations.
Management
16. 16
The models have yet to be tested
prospectively and their benefit on
clinical or other outcomes
demonstrated.
Healthcare organizations currently
lack the expert human resources
and the information technology or
IT capabilities to implement,
evaluate, incorporate, and
continuously monitor and
re-calibrate ML into clinical practice
Many of the studies developed
and validated models within the
same patient cohort; therefore,
generalizability to other
populations has not been
evaluated.
There remains significant trust
issues around the collection and
safe use of digital data that
need to be addressed.