BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.2.3.1//EN
TZID:Asia/Kolkata
X-WR-TIMEZONE:Asia/Kolkata
BEGIN:VEVENT
UID:102@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20250212T110000
DTEND;TZID=Asia/Kolkata:20250212T120000
DTSTAMP:20250128T022300Z
URL:https://cds.iisc.ac.in/events/mtech-research-thesis-defense-hybrid-cds
 -12-february-2025-lesion-synthesis-using-physics-based-noise-models-for-lo
 w-data-medical-imaging-regime-applications/
SUMMARY:Mtech Research Thesis Defense: HYBRID: CDS: 12\, February 2025 "Les
 ion Synthesis using Physics-Based Noise Models for Low-Data Medical Imagin
 g Regime applications.
DESCRIPTION:DEPARTMENT OF COMPUTATIONAL AND DATA SCIENCES\nMtech Research T
 hesis Defense\n\n\n\nSpeaker : Mr. Ramanujam Narayanan\nS.R. Number : 06-1
 8-01-10-22-22-1-21772\nTitle : Lesion Synthesis using Physics-Based Noise 
 Models for Low-Data Medical Imaging Regime applications.\nThesis examiner 
 : Dr. Pradipta Maji\nResearch Supervisor: Dr. Vaanathi Sundaresan\nDate &a
 mp\; Time : February 12\, 2025 (Wednesday) at 11:00 AM\n\nVenue : The Thes
 is Défense will be held on HYBRID Mode\n\n# 102 CDS Seminar Hall /MICROSO
 FT TEAMS\n\nPlease click on the following link to join the Thesis Defense:
 \n\nMS Teams link\n\n\n\nABSTRACT\n\nLesion segmentation and their progres
 sion prediction in medical imaging relies critically on the availability o
 f manually annotated\, heterogeneous large pathological datasets. Acquirin
 g such diverse large datasets is also challenging because they require coo
 rdination and ethical clearances from multiple sites\, and the manual anno
 tation of these images is both time-consuming and expensive. On the other 
 hand\, this data diversity is difficult to achieve in low-data regimes\, h
 indering the robust training of the models.\n\nOur study presents a lesion
  simulation method involving structural localized perturbation of healthy 
 tissue using noise models based on the physics of modalities. Later\, we l
 ocalize these perturbations within masks defined by composites of ellipsoi
 dal polygons (thus forming random shapes) and blended them with the input 
 image with varying contrast. The lesion simulation step does not require t
 raining and can generate any number of lesions with texture\, size\, and s
 cale variations\, injecting sufficient variability in the training data po
 ol in low-data regimes. We evaluate the performance of our simulated lesio
 ns for a downstream lesion segmentation task and show superior performance
  than its fully supervised counterpart. We also performed extensive ablati
 on studies and experimentally determined the optimal simulation data and m
 inimal training data required for training the segmentation model. Further
 \, we also showed detailed variation charts depicting the possible simulat
 ions the method can generate across different datasets. We evaluate the pe
 rformance on publicly available pathological brain MRI\, liver CT\, retina
 l fundus imaging and breast Ultrasound datasets with diverse lesions. Usin
 g only 75% of labelled real-world data\, the proposed method significantly
  improves the segmentation performance compared to the fully supervised tr
 aining\, with a 16% mean increase in the Dice score (DSC) and a 33% mean d
 ecrease in the 95th percentile of the normalised Hausdorff distance (HD95 
 (norm)).\nWe also discuss potential use case of the method for prediction 
 of post-treatment DWI and penumbra using pre-treatment NCCT and perfusion 
 maps.\n\n\n\nALL ARE WELCOME
CATEGORIES:Events,Thesis Defense
END:VEVENT
BEGIN:VTIMEZONE
TZID:Asia/Kolkata
X-LIC-LOCATION:Asia/Kolkata
BEGIN:STANDARD
DTSTART:20240213T110000
TZOFFSETFROM:+0530
TZOFFSETTO:+0530
TZNAME:IST
END:STANDARD
END:VTIMEZONE
END:VCALENDAR