This paper proposes a novel method to learn a 3D non-rigid deformation for automatic image registration between Positron Emission Tomography (PET) and Computed Tomography (CT) scans obtained from the same patient. There are two modules in the proposed scheme including (1) low-resolution displacement vector field (LR-DVF) estimator, which uses a 3D deep convolutional network (ConvNet) to directly estimate the voxel-wise displacement (a 3D vector field) between PET/CT images, and (2) 3D spatial transformer and re-sampler, which warps the PET images to match the anatomical structures in the CT images using the estimated 3D vector field. The parameters of the ConvNet are learned from a number of PET/CT image pairs via an unsupervised learning method. The Normalized Cross Correlation (NCC) between PET/CT images is used as the similarity metric to guide an end-to-end learning process with a constraint (regular term) to preserve the smoothness of the 3D deformations. A dataset with 170 PET/CT scans is used in experiments based on 10-fold cross-validation, where a total of 22,338 3D patches are sampled from the dataset. In each fold, 3D patches from 153 patients (90%) are used for training the parameters, while the remaining whole-body voxels from 17 patients (10%) are used for testing the performance of the image registration. The experimental results demonstrate that the image registration accuracy (the mean value of NCCs) is increased from 0.402 (the initial situation) to 0.567 on PET/CT scans using the proposed scheme. We also compare the performance of our scheme with previous work (DIRNet) and the advantage of our scheme is confirmed via the promising results.

Learning 3D non-rigid deformation based on an unsupervised deep learning for PET/CT image registration

PSR-Nets: Deep neural networks with prior shift regularization for PET/CT based automatic, accurate, and calibrated whole-body lymphoma segmentation

Function Integrated Diagnostic Assistance Based on MCA Models

A better understanding of the standardized uptake value (SUV) ranges of fludeoxyglucose positron emission tomography (FDG-PET) is crucial for radiologists. We have developed a statistical image analysis method for FDG-PET imaging of the torso, based on comparisons with normal data. The purpose of this study was to verify the accuracy of the normal model and usefulness of the statistical image analysis method by using typical cancer cases in the liver, lungs, and abdomen. Our study and the data collection (49 normal and 34 abnormal cases, in terms of PET/CT findings) were approved by the institutional review board. Our scheme consisted of the following steps: (1) normal model construction, (2) anatomical standardization of patient images, and (3) Z-score calculation to show the results of the statistical image analysis. To validate the Z-score index, we sampled 3603 and 1270 voxels in normal organs and abnormal regions, respectively, from the liver, lungs, and the abdomen. We then obtained the SUV and Z-score for each region. A receiver operating characteristics (ROC) analysis-based method was performed to evaluate the discrimination performances of the SUV and Z-score. The discrimination performances of the SUV and Z-score for the objective regions of interest (ROIs) were evaluated by the areas under the ROC curves (AUCs). As a result of the ROC analysis and statistical tests, all AUCs were found to be larger than 0.98. When the ROIs in the objective regions were combined, the mean AUCs of the Z-score and SUV were 0.99 and 0.98, respectively, the difference being statistically significant (
 $$p < 0.001$$
 ). The results suggested the possibility of applying a quantitative image reading method for torso FDG-PET imaging. Furthermore, a combination of the SUV and Z-score may provide increased accuracy of the determination methods, such as computer-aided detection and diagnosis.

Normal model construction for statistical image analysis of torso FDG-PET images based on anatomical standardization by CT images from FDG-PET/CT devices

Kenshiro Takeda

Papers

Normal model construction for statistical image analysis of torso FDG-PET images based on anatomical standardization by CT images from FDG-PET/CT devices