"Artificial Intelligence-Empowered Breast Cancer Radiogenomics for Precision Medicine"

Join Qian Liu, a PhD student at the University of Manitoba for their talk on Thursday, February 17th about “Artificial Intelligence-Empowered Breast Cancer Radiogenomics for Precision Medicine”.


Breast cancer (BC) is heterogeneous both gnomically and phenotypically. In precise medicine, it is of great significant to develop computational frameworks for identifying prognostic biomarkers which can capture both genomic and phenomic heterogeneity of BC. Radiogenomics is a field where medical images and genomic measurements are integrated and mined to solve challenging clinical problems. There are some limitations in previous radiogenomic studies such as data incompleteness, feature subjectivity and low interpretability. Majority of radiogenomic studies miss one or two of medical imaging data, genomic data, and clinical outcome data, which results in the data incomplete issue. The feature subjectivity comes from the extraction of imaging features with significant human involvement. Although there are some automatic feature extraction methods, such as deep learning (DL), they have limited interpretability. Therefore, there is a pressing need to address the limitations so that fully automatic and transparent radiogenomic prognostic biomarkers could be identified for BC.

In this talk, I will introduce two computational frameworks to address the radogenomic analysis issues. The first framework mainly focuses on exploring deep learning approaches for automatic medical imaging feature extraction and their association to single type of genomic features and clinical outcomes. The second framework discusses the application of tensor factorization technique to integrate and extract multiple types of omics features, the new deep learning strategy to extract imaging features, and a leverage strategy to take advantage of unpaired imaging, genomic, and clinical outcome data. I will also discuss how to apply statistic mediation analysis for providing further interpretation of identified radiogenomic biomarkers. This work provides a new perspective for conducting a comprehensive radiogenomic study when only limited resources are given. Comparing with traditional radiogenomic biomarkers, the biomarkers identified by the proposed frameworks have significantly higher prognostic value, and their interpretability is guaranteed by different levels of build-in and follow-up analyses.

DATE: Thursday, February 17th, 2022

WHERE: Zoom (see below for more information)

WHEN: 2:30pm, meet & greet at 2:00pm


The zoom call will open at 2:00pm to allow you to join before the seminar begins. Qian Liu will be there. You are welcome to join the meeting and talk to them.

Zoom Link:

MEETING ID: 666 9874 2330

PASSCODE: 103307

If you run into any troubles please email in the Statistics Office.

See poster HERE
Feb 15, 2022