BEGIN:VCALENDAR
VERSION:2.0
PRODID:talks.ox.ac.uk
BEGIN:VEVENT
SUMMARY:Towards Safe & Robust AI for Image-guided Diagnosis and Interventi
 on - Dr Mobarakol Islam (University College London)
DTSTART;VALUE=DATE-TIME:20230526T143000
DTEND;VALUE=DATE-TIME:20230526T160000
UID:https://talks.ox.ac.uk/talks/id/af798453-a86e-49a9-b264-f84884d16d89/
DESCRIPTION:Dr Mobarakol Islam is a senior research fellow at WEISS\, Univ
 ersity College London. Before that\, he was a postdoctoral research associ
 ate at the Department of Computing\, Imperial College London\, under the s
 upervision of Dr. Ben Glocker in BioMedIA Lab. He holds a PhD degree from 
 the Dept. of ISEP at National University of Singapore (NUS) and afterward 
 worked as a research fellow in the same Lab. Before that\, he was a lead s
 oftware engineer at Samsung R&D Institute. His research focuses on enhanci
 ng deep neural network robustness\, fairness\, and reliability using uncer
 tainty\, calibration and causality to improve image-guided disease diagnos
 is and intervention. Overall\, his research covers the medical imaging and
  video sources of MRI\, CT\, X-ray\, Ultrasound\, Endoscope\, and Microsco
 pe and non-imaging data sources of DNA\, Genomic\, Radiomic\, and clinical
  information. He is also involved in teaching undergraduate and postgradua
 te students and supervising PhD students with collaborative projects betwe
 en UCL\, ICL and NUS. He has received several awards including Turing Post
 doctoral Enrichment Award\, AUAPAF Conference Scholarship\, ISEP PhD Schol
 arship\, ICRA/MICCAI travel Awards\, and KUETEF Best Paper Award. He is se
 rving as an area-chair at MICCAI 2023\, organizing of the MICCAI DART work
 shop and reviewer of the several top conferences and journals in Healthcar
 e AI such as TPAMI\, MedIA\, IEEE TMI\, MICCAI\, ICRA\, IROS\, IJCARS\, IE
 EE RA-L\, and Neurocomputing.\n\nAbstract:\nAlthough AI has enormous poten
 tial to accelerate healthcare\, there are very few examples of AI-based me
 dical systems to translate into clinical practice due to the concerns of A
 I algorithmic trust\, safety\, and transparency. The key limitations of cu
 rrent AI-enabled systems\, including recent foundation models\, are reliab
 ility\, technical robustness\, fairness\, and transparency. In particular\
 , AI models are (i) poorly robust: the performance drops significantly on 
 data variation\; (ii) unreliable: overly confident in prediction and unabl
 e to provide feedback when a prediction is wrong or confusing\; (iii) fair
 ness and bias: underdiagnosis towards a certain population\; (iv) catastro
 phic forgetting: disruption of previously learned tasks with the training 
 of novel tasks in a constantly changing environment. In this talk\, I will
  discuss some of my works toward safe and reliable AI in the applications 
 of image-guided diagnosis and intervention. More specifically\, the novel 
 methods on uncertainty and confidence calibration\, perturbation\, computa
 tional stress testing\, feature-level regularization\, curriculum Fourier 
 domain adaptation\, and synthetic continual learning with vision-language 
 modeling.\nSpeakers:\nDr Mobarakol Islam (University College London)
LOCATION:Wolfson College (Levett Room)\, Linton Road OX2 6UD
TZID:Europe/London
URL:https://talks.ox.ac.uk/talks/id/af798453-a86e-49a9-b264-f84884d16d89/
BEGIN:VALARM
ACTION:display
DESCRIPTION:Talk:Towards Safe & Robust AI for Image-guided Diagnosis and I
 ntervention - Dr Mobarakol Islam (University College London)
TRIGGER:-PT1H
END:VALARM
END:VEVENT
END:VCALENDAR