TRUE or FALSE: artificial intelligence (AI) exists
FALSE: AI does not exist, and may never exist
What is machine learning (ML)?
“a logical contradiction, and thus nonsense”
What is deep learning?
“learning”
Arrange the following from oldest to newest (in terms of when these concepts came into existence): artificial intelligence, deep learning, machine learning
artificial intelligence > machine learning > deep learning
How many people are challenged by mental illness in Canada? Provide a number and a percentage.
- 20-25%
- 6.7 million +
How many people are challenged by mental illness in Alberta? Provide a number and a percentage.
- 25%
- 1 million +
TRUE or FALSE: there are more people with heart disease and diabetes in Canada than there are people with mental illness.
FALSE: there are MORE people with MENTAL ILLNESS than there are people with heart disease and diabetes in Canada
What are the limits in mental health services?
- symptom-biased diagnosis
- low-efficiency high-cost treatment (trial and error)
- lack of accessible long-term support
- complexity of mental illness and limited information (personal, populational, longitudinal level)
How does machine learning work? Provide an example.
- models are inputted (e.g. atrophied and normal brains)
- put a line in the space between the distinct models (i.e. line = modifier)
what is the difference between supervised and unsupervised learning?
- supervised: human teaches the machine to label clusters (e.g. apples and strawberries)
- unsupervised: no human labelling; machine defines distinct groups as clusters (e.g. cluster 1 and cluster 2)
Why use machine learning? (i.e. what are the benefits?)
- predictivity (individual-level prediction)
- generalizability (cross-validation, avoid over-fitting)
- finding the best features
TRUE or FALSE: in machine learning, significance = individual prediction
FALSE: significance does NOT equal individual prediction
TRUE or FALSE: machine learning can predict treatment response to ECT with hippocampal subfield volumes
TRUE
Machine learning can be used to identify life-time suicide attempters. What percentage of predicted non-attempters are actual non-attempters?
69%
Machine learning can be used to identify life-time suicide attempters. What percentage of predicted non-attempters are actual attempters?
31%
Machine learning can be used to identify life-time suicide attempters. What percentage of predicted attempters are actual non-attempters?
28%
Machine learning can be used to identify life-time suicide attempters. What percentage of predicted attempters are actual attempters?
72%
TRUE or FALSE: machine learning is used in administrative health data
TRUE
In collaboration with Alberta Health, machine learning has been used to predict what mental health data? The preliminary analysis achieved what percentage of accuracy of individual prediction of OUD?
opioid use disorder (OUD); 86%
What percentage of predicted non-OUD individuals are actual non-OUD individuals?
79%
What percentage of predicted non-OUD individuals are actual OUD individuals?
7%
What percentage of predicted OUD individuals are actual non-OUD individuals?
21%
What percentage of predicted OUD individuals are actual OUD individuals?
93%
What are some ongoing machine learning projects?
- health administration data and medical records
- clinical trials
- genetic imaging
- aging
-
CNS Structures6
-
Neuronal Excitability19
-
CNS Development54
-
Locus Coeruleus72
-
Respiratory System54
-
Locomotion81
-
Interneuron Network Connnectivity84
-
Claustrum51
-
Pain System 179
-
Pain System 238
-
Stroke75
-
Artificial Intelligence25
-
MRI60
-
Motor disorders - Spasticity78
-
Plasticity and Learning39
-
Spinal Cord Injury37
-
Addiction0
-
Schizophrenia27
-
Depression39
-
Neural Prostheses 120
-
Neural Prostheses 235
