Improved brain metastases segmentation using generative adversarial network and conditional random field optimization mask R-CNN

Yiren Wang; Zhongjian Wen; Lei Su; Hairui Deng; Jiali Gong; Hongli Xiang; Yongcheng He; Huaiwen Zhang; Ping Zhou; Haowen Pang

doi:10.1002/mp.17176

Improved brain metastases segmentation using generative adversarial network and conditional random field optimization mask R-CNN

Med Phys. 2024 May 22. doi: 10.1002/mp.17176. Online ahead of print.

Authors

Yiren Wang^{1

2}, Zhongjian Wen^{1

2}, Lei Su³, Hairui Deng^{1

2}, Jiali Gong¹, Hongli Xiang¹, Yongcheng He⁴, Huaiwen Zhang^{5

6}, Ping Zhou^{2

7

8}, Haowen Pang⁹

Affiliations

¹ School of Nursing, Southwest Medical University, Luzhou, Sichuan, China.
² Wound Healing Basic Research and Clinical Application Key Laboratory of Luzhou, School of Nursing, Southwest Medical University, Luzhou, Sichuan, China.
³ School of Medical Information and Engineering, Southwest Medical University, Luzhou, Sichuan, China.
⁴ Department of Pharmacy, College of Veterinary Medicine, Sichuan Agricultural University, Chengdu, Sichuan, China.
⁵ Department of Radiotherapy, Jiangxi Cancer Hospital, The Second Affiliated Hospital of Nanchang Medical College, Jiangxi Clinical Research Center for Cancer, Nanchang, Jiangxi, China.
⁶ Department of Oncology, The Third People's Hospital of Jingdezhen, Jingdezhen, Jiangxi, China.
⁷ Department of Nursing, The Affiliated Hospital of Southwest Medical University, Luzhou, Sichuan, China.
⁸ Department of Radiology, The Affiliated Hospital of Southwest Medical University, Luzhou, Sichuan, China.
⁹ Department of Oncology, The Affiliated Hospital of Southwest Medical University, Luzhou, Sichuan, China.

PMID: 38775791
DOI: 10.1002/mp.17176

Abstract

Background: In radiotherapy, the delineation of the gross tumor volume (GTV) in brain metastases using computed tomography (CT) simulation localization is very important. However, despite the criticality of this process, a pronounced gap exists in the availability of tools tailored for the automatic segmentation of the GTV based on CT simulation localization images.

Purpose: This study aims to fill this gap by devising an effective tool specifically for the automatic segmentation of the GTV using CT simulation localization images.

Methods: A dual-network generative adversarial network (GAN) architecture was developed, wherein the generator focused on refining CT images for more precise delineation, and the discriminator differentiated between real and augmented images. This architecture was coupled with the Mask R-CNN model to achieve meticulous GTV segmentation. An end-to-end training process facilitated the integration between the GAN and Mask R-CNN functionalities. Furthermore, a conditional random field (CRF) was incorporated to refine the initial masks generated by the Mask R-CNN model to ensure optimal segmentation accuracy. The performance was assessed using key metrics, namely, the Dice coefficient (DSC), intersection over union (IoU), accuracy, specificity, and sensitivity.

Results: The GAN+Mask R-CNN+CRF integration method in this study performs well in GTV segmentation. In particular, the model has an overall average DSC of 0.819 ± 0.102 and an IoU of 0.712 ± 0.111 in the internal validation. The overall average DSC in the external validation data is 0.726 ± 0.128 and the IoU is 0.640 ± 0.136. It demonstrates favorable generalization ability.

Conclusion: The integration of the GAN, Mask R-CNN, and CRF optimization provides a pioneering tool for the sophisticated segmentation of the GTV in brain metastases using CT simulation localization images. The method proposed in this study can provide a robust automatic segmentation approach for brain metastases in the absence of MRI.

Keywords: artificial intelligence; automatic segmentation; brain metastases; convolutional neural network; deep learning; gross tumor volume.

Abstract

Grants and funding