Efficient Multiple Organ Localization in CT Image using 3D Region Proposal Network
X. Xu, F. Zhou, B. Liu, D. Fu, X. Bai
IEEE Trans Med Imaging., vol. 38, issue 1, January 2019
Multiple organ localization is a fundamental but frequently required procedure in many medical image analysis tasks, such as organ segmentation, image registration, and lesion detection. Accurate and efficient localization of the target organs could largely improve the performance of the subsequent algorithms. However, organ localization is also a challenging problem due to the large variation of organ appearance and the complex background across different patients.
In this work, we propose an efficient method for multiple organ localization in CT image using 3D region proposal network. This ConvNet-based method is fully implemented in 3D manner, thus it can take full advantages of the spatial context information in CT image to perform accurate and efficient organ localization with only one prediction. As our experimental results shown, the proposed method could provide higher localization accuracy than the current state-of-the-art methods with approximate 4 to 18 times faster processing speed (around 0.3 seconds per CT scan). The full implementation of the proposed method has been made publicly available on https://github.com/superxuang/caffe_3d_faster_rcnn.
Another major contribution of this work is that we have established a public dataset dedicated for body organ localization based on the MICCAI Liver Tumor Segmentation (LiTS) image dataset. 3D bounding box of 12 body organs in 201 abdominal CT scans are included in this dataset. The dataset has been shared on http://dx.doi.org/10.21227/df8g-pq27.