Click here to


Are you sure ?

Yes, do it No, cancel

Synthetic MRI-Aided Prostate Segmentation in CT Image

Y Lei*, S Tian , T Wang , Y Liu , X Dong , X Jiang , P Patel , A Jani , W Curran , T Liu , X Yang , Emory Univ, Atlanta, GA


(Thursday, 7/18/2019) 7:30 AM - 9:30 AM

Room: 221AB

Purpose: Accurately defining prostate volume on CT for treatment planning is challenging due to its poor soft-tissue contrast. MRI has been used to aid prostate delineation, but its final accuracy is limited by that of the MRI-CT registration. Therefore, this study aims to develop an accurate prostate segmentation strategy on CT images with CT-based synthetic MRI (sMRT) to deal with this prostate radiotherapy treatment planning challenge.

Methods: We propose a new prostate segmentation strategy which apply a MRI-based pre-trained deep attentional network to CT-based sMRI to accurately segment the prostate on CT. Our method consists of three major steps. First, a cycle generative adversarial network (cycle-GAN) was used to estimate sMRI from CT. Second, a pre-trained deep attentional fully convolution network (DAFCN), which had been previously trained on MRIs with prostate contours delineated on them, was fine-tuned by sMRI and the corresponding prostate contours deformed from MRIs. Attention models were introduced to retrieve the most relevant features to classify prostate and non-prostate regions. Deep supervision was also incorporated into this DAFCN to enhance the features’ discriminative characteristic. The final prostate contour was obtained by feeding new CT into the well-trained cycle-GAN and DAFCN. We performed leave-one-out cross-validation to evaluate the proposed algorithm. Our segmented prostate contours were compared with the contours manually delineated by physicians on MRI to quantitatively evaluate segmentation accuracy.

Results: The segmentation technique was validated with a clinical study of 20 patients’ CT and MR images. The mean Dice similarity coefficient, sensitivity and specificity indexes between our segmented and MRI-defined prostate volume were 90.1±5.2%, 87.6±4.9%, and 94.3±4.0%.

Conclusion: We have proposed a novel prostate CT segmentation strategy using CT-based sMRI and validated its accuracy against MRI-define prostate. This technique could provide accurate prostate volume for treatment planning without requiring MRI acquisition, greatly facilitating the routine clinical workflow.

Funding Support, Disclosures, and Conflict of Interest: NIH R01 CA215718


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email