BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260622T215832EDT-4131WTOPJF@132.216.98.100
DTSTAMP:20260623T015832Z
DESCRIPTION:Speaker: Raghav Mehta\n\nSupervised  by Prof. Tal Arbel\n\nThes
 is defence\n\nAbstract: Although Deep Learning (DL) models have been shown
  to perform very well on various medical imaging tasks\, inference in the 
 presence of pathology presents several challenges to common models. These 
 challenges impede the integration of DL models into real clinical workflow
 s. Deployment of these models into real clinical contexts requires: (1) th
 at the confidence in DL model predictions be accurately expressed in the f
 orm of uncertainties and (2) that they exhibit robustness and fairness acr
 oss different sub-populations. Quantifying the reliability of DL model pre
 dictions in the form of uncertainties could enable clinical review of the 
 most uncertain regions\, thereby building trust and paving the way toward 
 clinical translation. Similarly\, by embedding uncertainty estimates acros
 s cascaded inference tasks\, prevalent in medical image analysis\, perform
 ance on the downstream inference tasks should also be improved. In this th
 esis\, we develop an uncertainty quantification score for the task of Brai
 n Tumour Segmentation. We evaluate the score's usefulness during the two c
 hallenges\, BraTS 2019 and BraTS 2020.  Overall\, our findings confirm the
  importance and complementary value that uncertainty estimates provide to 
 segmentation algorithms\, highlighting the need for uncertainty quantifica
 tion in medical image analyses. We further show the importance of uncertai
 nty estimates in medical image analysis by propagating uncertainty generat
 ed by upstream tasks into the downstream task of interest. Our results on 
 three different clinically relevant tasks indicate that uncertainty propag
 ation helps improve the performance of the downstream task of interest. Ad
 ditionally\, we combine the aspect of uncertainty estimates with fairness 
 across demographic subgroups into the picture. With extensive experiments 
 on multiple tasks\, we show that popular ML methods for achieving fairness
  across different subgroups\, such as data-balancing and distributionally 
 robust optimization\, succeed in terms of the model performances for some 
 of the tasks. However\, this can come at the cost of poor uncertainty esti
 mates associated with the model predictions. This tradeoff must be mitigat
 ed if fairness models are to be adopted in medical image analysis. In the 
 last part of the thesis\, we look at Active Learning (AL) for reduced manu
 al labeling of a dataset. Specifically\, we present an information-theoret
 ic AL framework that guides the optimal selection of images for labeling. 
 Results indicate that the proposed framework outperforms several existing 
 methods\, and by careful design choices\, it can be integrated into existi
 ng DL classifiers with minimal computational overhead.\n\n \n
DTSTART:20230705T180000Z
DTEND:20230705T190000Z
LOCATION:MC 603\, McConnell Engineering Building\, CA\, QC\, Montreal\, H3A
  0E9\, 3480 rue University
SUMMARY:Integrating Bayesian Deep Learning Uncertainties in Medical Image A
 nalysis
URL:/cim/channels/event/integrating-bayesian-deep-lear
 ning-uncertainties-medical-image-analysis-351716
END:VEVENT
END:VCALENDAR