Breast cancer ultrasound image segmentation using improved 3DUnet++
Breast cancer is the most common cancer and the main cause of cancer-related deaths in women around the world. Early detection reduces the number of deaths. Automated breast ultrasound (ABUS) is a new and promising screening method for examining the entire breast. Volumetric ABUS examination is time...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-06-01
|
| Series: | WFUMB Ultrasound Open |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2949668324000363 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1823856656046358528 |
|---|---|
| author | Saba Hesaraki Abdul Sajid Mohammed Mehrshad Eisaei Ramin Mousa |
| author_facet | Saba Hesaraki Abdul Sajid Mohammed Mehrshad Eisaei Ramin Mousa |
| author_sort | Saba Hesaraki |
| collection | DOAJ |
| description | Breast cancer is the most common cancer and the main cause of cancer-related deaths in women around the world. Early detection reduces the number of deaths. Automated breast ultrasound (ABUS) is a new and promising screening method for examining the entire breast. Volumetric ABUS examination is time-consuming, and lesions may be missed during the examination. Therefore, computer-aided cancer diagnosis in ABUS volume is highly expected to help the physician for breast cancer screening. In this research, we presented 3D structures based on UNet, ResUNet, and UNet++ for the automatic detection of cancer in ABUS volume to speed up examination while providing high detection sensitivity with low false positives (FPs). The three investigated approaches were evaluated on equal datasets in terms of training and testing as well as with proportional hyperparameters. Among the proposed approaches in classification and segmentation problems, the UNet++ approach was able to achieve more acceptable results. The UNet++ approach on the dataset of the Tumor Segmentation, Classification, and Detection Challenge on Automated 3D Breast Ultrasound 2023 (Named TSCD-ABUS2023) was able to achieve Accuracy = 0.9911 and AUROC = 0.9761 in classification and Dice = 0.4930 in segmentation. |
| format | Article |
| id | doaj-art-462177acf2244fa6a65bec26a19a191c |
| institution | Kabale University |
| issn | 2949-6683 |
| language | English |
| publishDate | 2025-06-01 |
| publisher | Elsevier |
| record_format | Article |
| series | WFUMB Ultrasound Open |
| spelling | doaj-art-462177acf2244fa6a65bec26a19a191c2025-02-12T05:33:11ZengElsevierWFUMB Ultrasound Open2949-66832025-06-0131100068Breast cancer ultrasound image segmentation using improved 3DUnet++Saba Hesaraki0Abdul Sajid Mohammed1Mehrshad Eisaei2Ramin Mousa3Islamic Azad University Science and Research Branch, Iran; Corresponding author.University of the Cumberlands, USAUniversity of Mazandaran, IranUniversity of Zanjan, IranBreast cancer is the most common cancer and the main cause of cancer-related deaths in women around the world. Early detection reduces the number of deaths. Automated breast ultrasound (ABUS) is a new and promising screening method for examining the entire breast. Volumetric ABUS examination is time-consuming, and lesions may be missed during the examination. Therefore, computer-aided cancer diagnosis in ABUS volume is highly expected to help the physician for breast cancer screening. In this research, we presented 3D structures based on UNet, ResUNet, and UNet++ for the automatic detection of cancer in ABUS volume to speed up examination while providing high detection sensitivity with low false positives (FPs). The three investigated approaches were evaluated on equal datasets in terms of training and testing as well as with proportional hyperparameters. Among the proposed approaches in classification and segmentation problems, the UNet++ approach was able to achieve more acceptable results. The UNet++ approach on the dataset of the Tumor Segmentation, Classification, and Detection Challenge on Automated 3D Breast Ultrasound 2023 (Named TSCD-ABUS2023) was able to achieve Accuracy = 0.9911 and AUROC = 0.9761 in classification and Dice = 0.4930 in segmentation.http://www.sciencedirect.com/science/article/pii/S2949668324000363UNet++ResUNetUNetUltrasound imageBreast tumor segmentation |
| spellingShingle | Saba Hesaraki Abdul Sajid Mohammed Mehrshad Eisaei Ramin Mousa Breast cancer ultrasound image segmentation using improved 3DUnet++ WFUMB Ultrasound Open UNet++ ResUNet UNet Ultrasound image Breast tumor segmentation |
| title | Breast cancer ultrasound image segmentation using improved 3DUnet++ |
| title_full | Breast cancer ultrasound image segmentation using improved 3DUnet++ |
| title_fullStr | Breast cancer ultrasound image segmentation using improved 3DUnet++ |
| title_full_unstemmed | Breast cancer ultrasound image segmentation using improved 3DUnet++ |
| title_short | Breast cancer ultrasound image segmentation using improved 3DUnet++ |
| title_sort | breast cancer ultrasound image segmentation using improved 3dunet |
| topic | UNet++ ResUNet UNet Ultrasound image Breast tumor segmentation |
| url | http://www.sciencedirect.com/science/article/pii/S2949668324000363 |
| work_keys_str_mv | AT sabahesaraki breastcancerultrasoundimagesegmentationusingimproved3dunet AT abdulsajidmohammed breastcancerultrasoundimagesegmentationusingimproved3dunet AT mehrshadeisaei breastcancerultrasoundimagesegmentationusingimproved3dunet AT raminmousa breastcancerultrasoundimagesegmentationusingimproved3dunet |