BROWSE

ITEM VIEW & DOWNLOAD

Large-scale Medical Image Processing

DC Field Value Language
dc.contributor.advisor Chun, Se Young -
dc.contributor.author Kim, Jisoo -
dc.date.accessioned 2020-04-01T08:56:04Z -
dc.date.available 2020-04-01T08:56:04Z -
dc.date.issued 2020-02 -
dc.identifier.other 200000288768 -
dc.identifier.uri http://unist.dcollection.net/common/orgView/200000288768 en_US
dc.identifier.uri https://scholarworks.unist.ac.kr/handle/201301/31775 -
dc.description Department of Electrical Engineering -
dc.description.abstract Deep learning based approaches for vision motivated many researchers in Medical Image processing fields due to the powerful performances. Compare to the natural image data, the medical image data set commonly consumes huge memory with complex data structures. In addition, to demonstrate the large scale images for clinical purpose such as CT scans or pathological image data, it is commonly known to be difficult so that direct application of conventional deep models with typical GPU usage should be considered. For example, in the pathological data which is the image of microscope of human cells to classify tumor cells or not, the size of image slide is far larger than natural high resolution images while the field of view (FOV) that we are interested region is tiny. On the other hand, to handle the large scale of CT data which is using X-ray beams to visualize in-vivo hardness structures, due to the memory limitation of GPU device, the patch-wise method is suppressed to yield high performance and is disturbed to compute faster. Thus, in this paper, we investigate the how data balancing method effectively enhance the deep approach method when there is only unbalanced dataset. Furthermore, we propose the efficient memory utilization of multi-gpu method for deep learning with large scale CT images. -
dc.description.statementofresponsibility close -
dc.language ENG -
dc.publisher Graduate School of UNIST -
dc.title Large-scale Medical Image Processing -
dc.type Master's thesis -
dc.administration.regnum 200000288768 -
Appears in Collections:
EE_Theses_Master

find_unist can give you direct access to the published full text of this article. (UNISTARs only)

Show simple item record

qrcode

  • mendeley

    citeulike

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

MENU