ADHD-200 competition
Competition, leading to several papers. Some press, as our team WON the competition, by ignoring the fMRI data, and just using the non-fMRI features.
See http://www.talyarkoni.org/blog/2011/10/12/brain-based-prediction-of-adhd-now-with-100-fewer-brains/ and https://www.amii.ca/fmri-analysis/. Since then, we have significantly improved the results. related publications
|
Prognostic prediction of treatment response and clinical outcome in Drug-free Schizophrenia
The goal of this ongoing project is to develop models to predict short term clinical outcomes in largely antipsychotic-naive schiozophrenia cohort, using multimodal clinical and neuroimaging data.
|
EMPaSchiz: Multi-parcellation ensmble model to diagnose Schizophrenia
We developed an ensemble model that stacks predictions from several ‘single-source’ models, each based on features of regional activity and functional connectivity, over a range of different a priori parcellation schemes -- that outperform current methods in Schizophrenia diagnosis.
related publications
|
Schizophrenia diagnosis using functional network features
LEARNING STABLE NETWORK-BASED PATTERNS OF SCHIZOPHRENIA - We used machine learning techniques and whole-brain functional connectivity features, extracted from brain fMRI data of patients diagnosed with schizophrenia and healthy control subjects, to discriminate between the two group. We achived 74% classification accuracy in the fBIRN multi-site fMRI dataset. Also, using whole-brain functional connectivity and sparse regression (Elastic Net), we accurately predicted severity of several symptoms, such as inattentiveness and bizarre behaviour in patients.
related publications
|
|
Learning neural markers of schizophrenia using recurrent neural networks
Here we propose a new method based on recurrent-convolutional neural networks to automatically learn useful representations from segments of 4-D fMRI recordings. Our goal is to exploit both spatial and temporal information in the functional MRI movie (at the whole-brain voxel level) for identifying patients with schizophrenia.
Related publications
|
Dealing with batch effects in fMRI data for diagnosis of mental disorders
A common approach to increase the size of fMRI datasets is to combine data coming from different sources. Datafrom different sources have different probability distributions, which complicates the job of machine learning algorithms. We are exploring different approaches to decrease the impact of this problem, known as batch effects.
related publications
|
Using graphical models on cognitive data to identify schizophrenia and depression
Here, we used graphical LASSO to build graphs for healthy controls, people with major depression, and people with schizophrenia, then used a maximum likelihood approach to classify subjects on a hold-set. Finally, we created a partial correlation graphs to find correlations among features.
reLated publications
|
Validation of biomaker identification in OMICS research
Graphical models for studying brain-gene relationships
Goal: Develop probabilistic graphical models to map brain MRI features to the genetic SNP dependencies using Schizophrenia datasets (NIMHANS imaging genetics data and more).
|
Canadian Biomarker Integration Network in Depression (CAN-BIND) escitalopram treatment response
This project aims to develop models to predict response to escitalopram using MR data.
|
Predicting patient response to desvenlafaxine using clinical trial data
Goal: Develop a model to predict DVS response, using data from multiple stage III/IV clinical trials (Pfizer).
|
IDChatbot - An artificially intelligent chatbot that provides useful information for children suffering from intellectual disabilities
We propose a model for a chatbot application, Wallace, that asks parents/caregivers questions about their child and uses that information to suggest websites and other resources designed to provide relevant information. We implement an expert system that decides the best questions to ask given the answers provided, to build a patient profile and suggest relevant websites and resources.
|