BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:A Distance Function based Cascaded Neural Network for accurate Pol
 yps Segmentation and Classification -  Yuanhong Jiang\, Shanghai Jiao Tong
  University
DTSTART:20221201T140000Z
DTEND:20221201T150000Z
UID:TALK176813@talks.cam.ac.uk
CONTACT:Pietro Lio
DESCRIPTION:In clinical practice\, it is often difficult to locate and mea
 sure the size of polyps accurately for the follow-up surgical operation de
 cision. In this paper\, based on the position constraint between the prima
 ry organ and polyps boundary\, we propose a U-Net based cascaded neural ne
 twork for the joint segmentation of the organ of interest and polyps. The 
 constraint on their position relation is further imposed by adding a narro
 w-band distance function and complimentary dice function to the loss funct
 ion. Through a series of comparisons and ablation study\, the proposed met
 hod with the cascaded network architecture and the additional loss functio
 ns was validated on an in-house dataset for gallbladder polyps segmentatio
 n and classification. It has been demonstrated that the proposed method ac
 hieved a significant improvement over conventional U-Net\, U-Net++ etc.. E
 ventually\, the pathological type classification based on the segmented po
 lys shows 30% higher accuracy compared to those conventional ResNet based 
 results.                                       \n*Biography*\nYuanhong Jia
 ng is a Ph.D student from Institute of Natural Sciences\, Shanghai Jiao To
 ng University. His research interest lies broadly in the area of medical i
 mage processing\, deep learning and graph neural networks. His recent work
 s include medical image segmentation and MR image reconstruction with deep
  learning approaches\, and he is researching on image processing task with
  graph neural networks.
LOCATION:Lecture Theatre 2
END:VEVENT
END:VCALENDAR