Previous Page  11 / 26 Next Page
Information
Show Menu
Previous Page 11 / 26 Next Page
Page Background

Fully Convolutional Network for Liver

Segmentation and Lesions Detection

This Friday,

MICCAI 2016

will host workshops that will be

held after the publication of our last

MICCAI Daily

. Among

them,

DLMIA 2016

: the 2nd Workshop on

Deep Learning

in Medical Image Analysis

. We have asked

Avi Ben Cohen

,

one of the poster presenters at the morning session, to tell

This project presents an automatic system for

liver segmentation and

lesions detection in CT examinations using fully convolutional neural

network architecture

. Currently this task is manually handled by

radiologists and can be very time consuming. In addition, the variability in

texture, size and different contrast enhancement behavior of liver lesions

and parenchyma make it in many cases a difficult task. One of the

difficulties comes from

detecting small liver lesions

: the model explores

different scales of the fully convolutional architecture for that purpose and

although in some cases this was helpful, in other cases small lesion can be

missed, meaning that more improvements and work needs to be done.

This project is a part of an ongoing project for creating a package of tools

which is necessary for

radiologists

, when they go over a liver CT

examination. These tools include the

liver segmentation

,

lesion detection

and segmentation

,

follow-up tools

for treatment evaluation and

classification tools

for different lesion types.

A package of tools necessary for radiologists

going over a liver CT examination

Presentation

11

us about the work he is presenting with his colleagues (

Idit Diamant

,

Eyal

Klang, Michal Amitai

and

Hayit Greenspan

). The project is a part of a joint

effort of the

Medical Image Processing lab in Tel Aviv University

(headed

by Prof. Hayit Greenspan) and the

Sheba Medical Center

.

MICCAI Daily: Thursday