TY - GEN
T1 - Evaluating the Segmentation Performance of Gross Volume Tumor in Cervical Cancer Using MRI Images
AU - Zaki, Nazar
AU - Krishnan, Anusuya
AU - Damseh, Rafat
AU - Qin, Wenjian
AU - Zaitouny, Ayham
AU - Ghebrehiwet, Isaias
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - Cervical Cancer (CC) is the most prevalent gynecologic malignancy worldwide. Tumor segmentation in CC is a crucial step for radiotherapy treatment and planning. The clinical practice involves laborious slice-by-slice segmentation of the primary tumor using simultaneous assessments from several image modalities, and ignores spatial ambiguity in tumor delineation. This work evaluates the performance of a state-of-the-art Landing AI for automated 3D medical image segmentation applied to Gross Tumor Volume (GTV) in CC from Magnetic Resonance Imaging (MRI) scans. Our work provides a novel in-house labeled dataset with a systematic assessment of the segmented lesions after network training on various voxel spacing MRI images. The segmentation performance was assessed using the dice coefficient. We demonstrated that training on MRI images to optimize Landing AI achieves an improved dice score of 0.92, outperforming other MRI models.
AB - Cervical Cancer (CC) is the most prevalent gynecologic malignancy worldwide. Tumor segmentation in CC is a crucial step for radiotherapy treatment and planning. The clinical practice involves laborious slice-by-slice segmentation of the primary tumor using simultaneous assessments from several image modalities, and ignores spatial ambiguity in tumor delineation. This work evaluates the performance of a state-of-the-art Landing AI for automated 3D medical image segmentation applied to Gross Tumor Volume (GTV) in CC from Magnetic Resonance Imaging (MRI) scans. Our work provides a novel in-house labeled dataset with a systematic assessment of the segmented lesions after network training on various voxel spacing MRI images. The segmentation performance was assessed using the dice coefficient. We demonstrated that training on MRI images to optimize Landing AI achieves an improved dice score of 0.92, outperforming other MRI models.
KW - Cervical cancer
KW - Deep learning
KW - Landing AI
KW - Medical image segmentation
UR - http://www.scopus.com/inward/record.url?scp=85207487553&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85207487553&partnerID=8YFLogxK
U2 - 10.1109/CIBCB58642.2024.10702146
DO - 10.1109/CIBCB58642.2024.10702146
M3 - Conference contribution
AN - SCOPUS:85207487553
T3 - 21st IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology, CIBCB 2024
BT - 21st IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology, CIBCB 2024
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 21st IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology, CIBCB 2024
Y2 - 27 August 2024 through 29 August 2024
ER -
