MU Net: Ovarian Follicle Segmentation Using Modified U-Net Architecture
Debasmita Saha1, Ardhendu Mandal2, Rinku Ghosh3

1Debasmita Saha*, Department of Computer Science, University of Gour Banga, Malda, Pin- 732103, West Bengal, India.
2Dr. Ardhendu Mandal, Department of Computer Science and Application, University of North Bengal, Siliguri, West Bengal, Pin-734013, India. 
3Rinku Ghosh, Department of Computer Science, University of Gour Banga, Malda, Pin- 732103, West Bengal, India.
Manuscript received on February 18, 2022. | Revised Manuscript received on February 24, 2022. | Manuscript published on April 30, 2022. | PP: 30-35 | Volume-11 Issue-4, April 2022. | Retrieval Number: 100.1/ijeat.D34190411422 | DOI: 10.35940/ijeat.D3419.0411422
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (

Abstract: Ovaries play a pivotal role in production by generating eggs through oogenesis in the female reproductive system. This is one crucial aspect of reproduction as eggs are fertilized by the sperm which eventually leads to fertilization and eventually ending in embryo formation. Ovaries are often susceptible to diseases like infertility, polycystic ovarian syndrome (PCOS), ovarian cancer etc. Screening of ovarian follicles via ultrasound images can be of great help in the diagnosis of these abnormal situations. However, screening in most scenarios is still carried out manually by doctors and sonographers leading it to be a monotonous, time consuming and laborious job as well. Thus automatic detection of follicles can reduce the burden of doctors. In our work, we propose MU-net, a novel 2D segmentation network, combination of both MobileNetV2 and U-Net for segmentation of the follicles from ovarian ultrasound images. The test is conducted on the USOVA3D Training Set 1. Although low contrast issues are common setback for ultrasound images, our model has achieved a descent accuracy rate of 98.4%. 
Keywords: CNN, Deep Learning, Follicle, Ovary, Segmentation. 
Scope of the Article: Deep Learning