eduzhai > Applied Sciences > Engineering >

Feature based Sequential Classifier with Attention Mechanism

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 9 pages

Abstract: Cervical cancer is one of the deadliest cancers affecting women globally.Cervical intraepithelial neoplasia (CIN) assessment using histopathologicalexamination of cervical biopsy slides is subject to interobserver variability.Automated processing of digitized histopathology slides has the potential formore accurate classification for CIN grades from normal to increasing grades ofpre-malignancy: CIN1, CIN2 and CIN3. Cervix disease is generally understood toprogress from the bottom (basement membrane) to the top of the epithelium. Tomodel this relationship of disease severity to spatial distribution ofabnormalities, we propose a network pipeline, DeepCIN, to analyzehigh-resolution epithelium images (manually extracted from whole-slide images)hierarchically by focusing on localized vertical regions and fusing this localinformation for determining Normal CIN classification. The pipeline containstwo classifier networks: 1) a cross-sectional, vertical segment-level sequencegenerator (two-stage encoder model) is trained using weak supervision togenerate feature sequences from the vertical segments to preserve thebottom-to-top feature relationships in the epithelium image data; 2) anattention-based fusion network image-level classifier predicting the final CINgrade by merging vertical segment sequences. The model produces the CINclassification results and also determines the vertical segment contributionsto CIN grade prediction. Experiments show that DeepCIN achievespathologist-level CIN classification accuracy.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×