View in EDS

Lesion-Harvester: Iteratively Mining Unlabeled Lesions and Hard-Negative Examples at Scale.

Saved in:
Bibliographic Details
Title: Lesion-Harvester: Iteratively Mining Unlabeled Lesions and Hard-Negative Examples at Scale.
Authors: Cai, Jinzheng1 (AUTHOR) caijinzheng883@paii-labs.com, Harrison, Adam P.1 (AUTHOR) adampharrison070@paii-labs.com, Zheng, Youjing2 (AUTHOR) zhengyoujing@vt.edu, Yan, Ke1 (AUTHOR) yanke383@paii-labs.com, Huo, Yuankai3 (AUTHOR) yuankai.huo@vanderbilt.edu, Xiao, Jing4 (AUTHOR) xiaojing661@pingan.com.cn, Yang, Lin5 (AUTHOR) lin.yang@bme.ufl.edu, Lu, Le1 (AUTHOR) le.lu@paii-labs.com
Source: IEEE Transactions on Medical Imaging. Jan2021, Vol. 40 Issue 1, p59-70. 12p.
Subjects: Mines & mineral resources, Machine learning, Computer-assisted image analysis (Medicine), Detectors, Medical imaging systems, Conveyor belts
Abstract: The acquisition of large-scale medical image data, necessary for training machine learning algorithms, is hampered by associated expert-driven annotation costs. Mining hospital archives can address this problem, but labels often incomplete or noisy, e.g., 50% of the lesions in DeepLesion are left unlabeled. Thus, effective label harvesting methods are critical. This is the goal of our work, where we introduce Lesion-Harvester—a powerful system to harvest missing annotations from lesion datasets at high precision. Accepting the need for some degree of expert labor, we use a small fully-labeled image subset to intelligently mine annotations from the remainder. To do this, we chain together a highly sensitive lesion proposal generator (LPG) and a very selective lesion proposal classifier (LPC). Using a new hard negative suppression loss, the resulting harvested and hard-negative proposals are then employed to iteratively finetune our LPG. While our framework is generic, we optimize our performance by proposing a new 3D contextual LPG and by using a global-local multi-view LPC. Experiments on DeepLesion demonstrate that Lesion-Harvester can discover an additional 9,805 lesions at a precision of 90%. We publicly release the harvested lesions, along with a new test set of completely annotated DeepLesion volumes. We also present a pseudo 3D IoU evaluation metric that corresponds much better to the real 3D IoU than current DeepLesion evaluation metrics. To quantify the downstream benefits of Lesion-Harvester we show that augmenting the DeepLesion annotations with our harvested lesions allows state-of-the-art detectors to boost their average precision by 7 to 10%. [ABSTRACT FROM AUTHOR]
Copyright of IEEE Transactions on Medical Imaging is the property of IEEE and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
FullText Text:
  Availability: 0
Header DbId: egs
DbLabel: Engineering Source
An: 148071999
AccessLevel: 6
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Lesion-Harvester: Iteratively Mining Unlabeled Lesions and Hard-Negative Examples at Scale.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Cai%2C+Jinzheng%22">Cai, Jinzheng</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> caijinzheng883@paii-labs.com</i><br /><searchLink fieldCode="AR" term="%22Harrison%2C+Adam+P%2E%22">Harrison, Adam P.</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> adampharrison070@paii-labs.com</i><br /><searchLink fieldCode="AR" term="%22Zheng%2C+Youjing%22">Zheng, Youjing</searchLink><relatesTo>2</relatesTo> (AUTHOR)<i> zhengyoujing@vt.edu</i><br /><searchLink fieldCode="AR" term="%22Yan%2C+Ke%22">Yan, Ke</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> yanke383@paii-labs.com</i><br /><searchLink fieldCode="AR" term="%22Huo%2C+Yuankai%22">Huo, Yuankai</searchLink><relatesTo>3</relatesTo> (AUTHOR)<i> yuankai.huo@vanderbilt.edu</i><br /><searchLink fieldCode="AR" term="%22Xiao%2C+Jing%22">Xiao, Jing</searchLink><relatesTo>4</relatesTo> (AUTHOR)<i> xiaojing661@pingan.com.cn</i><br /><searchLink fieldCode="AR" term="%22Yang%2C+Lin%22">Yang, Lin</searchLink><relatesTo>5</relatesTo> (AUTHOR)<i> lin.yang@bme.ufl.edu</i><br /><searchLink fieldCode="AR" term="%22Lu%2C+Le%22">Lu, Le</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> le.lu@paii-labs.com</i>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%22IEEE+Transactions+on+Medical+Imaging%22">IEEE Transactions on Medical Imaging</searchLink>. Jan2021, Vol. 40 Issue 1, p59-70. 12p.
– Name: Subject
  Label: Subjects
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Mines+%26+mineral+resources%22">Mines & mineral resources</searchLink><br /><searchLink fieldCode="DE" term="%22Machine+learning%22">Machine learning</searchLink><br /><searchLink fieldCode="DE" term="%22Computer-assisted+image+analysis+%28Medicine%29%22">Computer-assisted image analysis (Medicine)</searchLink><br /><searchLink fieldCode="DE" term="%22Detectors%22">Detectors</searchLink><br /><searchLink fieldCode="DE" term="%22Medical+imaging+systems%22">Medical imaging systems</searchLink><br /><searchLink fieldCode="DE" term="%22Conveyor+belts%22">Conveyor belts</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: The acquisition of large-scale medical image data, necessary for training machine learning algorithms, is hampered by associated expert-driven annotation costs. Mining hospital archives can address this problem, but labels often incomplete or noisy, e.g., 50% of the lesions in DeepLesion are left unlabeled. Thus, effective label harvesting methods are critical. This is the goal of our work, where we introduce Lesion-Harvester—a powerful system to harvest missing annotations from lesion datasets at high precision. Accepting the need for some degree of expert labor, we use a small fully-labeled image subset to intelligently mine annotations from the remainder. To do this, we chain together a highly sensitive lesion proposal generator (LPG) and a very selective lesion proposal classifier (LPC). Using a new hard negative suppression loss, the resulting harvested and hard-negative proposals are then employed to iteratively finetune our LPG. While our framework is generic, we optimize our performance by proposing a new 3D contextual LPG and by using a global-local multi-view LPC. Experiments on DeepLesion demonstrate that Lesion-Harvester can discover an additional 9,805 lesions at a precision of 90%. We publicly release the harvested lesions, along with a new test set of completely annotated DeepLesion volumes. We also present a pseudo 3D IoU evaluation metric that corresponds much better to the real 3D IoU than current DeepLesion evaluation metrics. To quantify the downstream benefits of Lesion-Harvester we show that augmenting the DeepLesion annotations with our harvested lesions allows state-of-the-art detectors to boost their average precision by 7 to 10%. [ABSTRACT FROM AUTHOR]
– Name: AbstractSuppliedCopyright
  Label:
  Group: Ab
  Data: <i>Copyright of IEEE Transactions on Medical Imaging is the property of IEEE and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.)
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=148071999
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1109/TMI.2020.3022034
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 12
        StartPage: 59
    Subjects:
      – SubjectFull: Mines & mineral resources
        Type: general
      – SubjectFull: Machine learning
        Type: general
      – SubjectFull: Computer-assisted image analysis (Medicine)
        Type: general
      – SubjectFull: Detectors
        Type: general
      – SubjectFull: Medical imaging systems
        Type: general
      – SubjectFull: Conveyor belts
        Type: general
    Titles:
      – TitleFull: Lesion-Harvester: Iteratively Mining Unlabeled Lesions and Hard-Negative Examples at Scale.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Cai, Jinzheng
      – PersonEntity:
          Name:
            NameFull: Harrison, Adam P.
      – PersonEntity:
          Name:
            NameFull: Zheng, Youjing
      – PersonEntity:
          Name:
            NameFull: Yan, Ke
      – PersonEntity:
          Name:
            NameFull: Huo, Yuankai
      – PersonEntity:
          Name:
            NameFull: Xiao, Jing
      – PersonEntity:
          Name:
            NameFull: Yang, Lin
      – PersonEntity:
          Name:
            NameFull: Lu, Le
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Text: Jan2021
              Type: published
              Y: 2021
          Identifiers:
            – Type: issn-print
              Value: 02780062
          Numbering:
            – Type: volume
              Value: 40
            – Type: issue
              Value: 1
          Titles:
            – TitleFull: IEEE Transactions on Medical Imaging
              Type: main
ResultId 1