Maitrei Kohli & Ashkan Pakzad - CMIC/WEISS Joint Seminar Series

Speaker: Maitrei Kohli

Title: From developmental behavioural genetics to predicting disease progression: transfer & ensemble learning approaches 

Abstract: 

This talk is split into two parts, first I’ll give a brief overview of my PhD research which spanned two different disciplines, contributing both to cognitive psychology and machine learning. The next part of the talk will focus on my current research; using ensemble methods to improve clinical predictive utility in HD. Ensembles have proven to be extremely effective and versatile in a broad spectrum of problem-domains and real-world applications. However, there has been no coordinated effort of using an ensemble ML method for making patient-specific and data-driven predictions of disease state in HD. So, we propose a computational framework for predictive-classification of disease state in Huntington’s disease (HD). The framework comprises of (a) an array of standard machine learning (ML) models; and (b) a stacked ensemble model. I’ll present the ensemble-based framework and some of our key findings.

Speaker: Ashkan Pakzad

Title: AirQuant: measuring airway morphology in lung disease

Abstract: 

Airway structure can be dramatically affected by lung disease. Indeed, specific patterns of structural airway damage can allow diagnosis of discrete diseases. But how do we quantify patterns of airway damage? AirQuant is an opensource tool that systematically parcellates the airways into their corresponding lung lobes and segments and extracts their morphological properties. It therefore allows us to quantify properties of airway structure and present these in clinically intuitive representations for easy translation. In this presentation, we consider AirQuant’s inner workings and demonstrate results associating AirQuant metrics with patient mortality in fibrosing lung disease.

Chair: Joe Jacob