Click here to

  1. Home
  2. Meeting Program
  3. All Sessions
  4. Abstracts
Session: [Return to Session]

Structure Boundary Preserving U-Net for Prostate Ultrasound Images Segmentation

H Bi1,2*, J Sun1, L Gao1, X Ni 1, (1)Changzhou No.2 People's Hospital ,Changzhou, 32, CN, (2)Changzhou University ,Changzhou, 32, CN

Presentations

PO-GePV-I-99 (Sunday, 7/10/2022)   [Eastern Time (GMT-4)]

ePoster Forums

Purpose: The prostate cancer diagnosis are performed under ultrasound-guided puncture for pathological cell extraction. However, the accurate prostate location still is a challenge from two aspects: (1) the prostate boundary in ultrasound images always ambiguous; (2) the delineation of radiologist always occupies multiple pixels lead to many disturbing points around the actual contour. To achieve precise prostate contour, we proposed a structure boundary preserving U-Net (SBP-UNet) in this paper.

Methods: The SBP-UNet incorporates prostate shape prior into traditional UNet. The shape prior is build by the key points selection module which is an active shape model (ASM) based method. Then, the module plug into the tradition U-net structure network to achieve the prostate segmentation.

Results: The experimental are conducted on two dataset: PH2 + ISBI 2016 challenge and our private prostate ultrasound dataset. The results on PH2 + ISBI 2016 challenge achieve dice coefficient score of 95.94% and jaccard coefficient of 88.58%. The results of prostate contour based on our method achieve higher pixel accuracy of 97.05%, mean intersection over union of 93.65%, dice coefficient score of 92.54%, and jaccard coefficient of 93.16%.

Conclusion: The experimental results show that the proposed SBP-Unet has good performance on PH2 + ISBI 2016 challenge and prostate ultrasound image segmentation that outperforms the state-of-the-art methods.

Funding Support, Disclosures, and Conflict of Interest: This work was supported by the National Natural Science Foundation No.62171125, the National Postdoctoral Programs No.2020M671277, the Science and Technology Project of Changzhou City No.CE20215045, the Key Laboratory of Computer Network and Information Integration (Southeast University) of Ministry of Education No.K93-9-2021-08, the Key Medical Physics Laboratory of Changzhou No. CM20193005.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email

Share:

Additional Links