BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260525T145718EDT-6380VTAxXU@132.216.98.100
DTSTAMP:20260525T185718Z
DESCRIPTION:Abstract\n\nAlthough Deep Learning (DL) models have been shown 
 to perform very well on various medical imaging tasks\, inference in the p
 resence of pathology presents several challenges to common models. These c
 hallenges impede the integration of DL models into real clinical workflows
 . Deployment of these models into real clinical contexts requires: (1) tha
 t the confidence in DL model predictions be accurately expressed in the fo
 rm of uncertainties and (2) that they exhibit robustness and fairness acro
 ss different sub-populations. Quantifying the reliability of DL model pred
 ictions in the form of uncertainties could enable clinical review of the m
 ost uncertain regions\, thereby building trust and paving the way toward c
 linical translation. Similarly\, by embedding uncertainty estimates across
  cascaded inference tasks\, prevalent in medical image analysis\, performa
 nce on the downstream inference tasks should also be improved. In this the
 sis\, we develop an uncertainty quantification score for the task of Brain
  Tumour Segmentation. We evaluate the score's usefulness during the two ch
 allenges\, BraTS 2019 and BraTS 2020. Overall\, our findings confirm the i
 mportance and complementary value that uncertainty estimates provide to se
 gmentation algorithms\, highlighting the need for uncertainty quantificati
 on in medical image analyses. We further show the importance of uncertaint
 y estimates in medical image analysis by propagating uncertainty generated
  by upstream tasks into the downstream task of interest. Our results on th
 ree different clinically relevant tasks indicate that uncertainty propagat
 ion helps improve the performance of the downstream task of interest. Addi
 tionally\, we combine the aspect of uncertainty estimates with fairness ac
 ross demographic subgroups into the picture. With extensive experiments on
  multiple tasks\, we show that popular ML methods for achieving fairness a
 cross different subgroups\, such as data-balancing and distributionally ro
 bust optimization\, succeed in terms of the model performances for some of
  the tasks. However\, this can come at the cost of poor uncertainty estima
 tes associated with the model predictions. This tradeoff must be mitigated
  if fairness models are to be adopted in medical image analysis. In the la
 st part of the thesis\, we look at Active Learning (AL) for reduced manual
  labeling of a dataset. Specifically\, we present an information-theoretic
  AL framework that guides the optimal selection of images for labeling. Re
 sults indicate that the proposed framework outperforms several existing me
 thods\, and by careful design choices\, it can be integrated into existing
  DL classifiers with minimal computational overhead.\n
DTSTART:20230705T180000Z
DTEND:20230705T200000Z
LOCATION:Room 603\, McConnell Engineering Building\, CA\, QC\, Montreal\, H
 3A 0E9\, 3480 rue University
SUMMARY:PhD defence of Raghav Mehta – Integrating Bayesian Deep Learning Un
 certainties in Medical Image Analysis
URL:/ece/channels/event/phd-defence-raghav-mehta-integ
 rating-bayesian-deep-learning-uncertainties-medical-image-analysis-348930
END:VEVENT
END:VCALENDAR