Functional Brain Imaging Synthesis Based on Image Decomposition and Kernel Modeling: Application to Neurodegenerative Diseases

Abstract

The rise of neuroimaging in research and clinical practice, together with the development of new machine learning techniques has strongly encouraged the Computer Aided Diagnosis (CAD) of different diseases and disorders. However, these algorithms are often tested in proprietary datasets to which the access is limited and, therefore, a direct comparison between CAD procedures is not possible. Furthermore, the sample size is often small for developing accurate machine learning methods. Multi-center initiatives are currently a very useful, although limited, tool in the recruitment of large populations and standardization of CAD evaluation. Conversely, we propose a brain image synthesis procedure intended to generate a new image set that share characteristics with an original one. Our system focuses on nuclear imaging modalities such as PET or SPECT brain images. We analyze the dataset by applying PCA to the original dataset, and then model the distribution of samples in the projected eigenbrain space using a Probability Density Function (PDF) estimator. Once the model has been built, we can generate new coordinates on the eigenbrain space belonging to the same class, which can be then projected back to the image space. The system has been evaluated on different functional neuroimaging datasets assessing the: resemblance of the synthetic images with the original ones, the differences between them, their generalization ability and the independence of the synthetic dataset with respect to the original. The synthetic images maintain the differences between groups found at the original dataset, with no significant differences when comparing them to real-world samples. Furthermore, they featured a similar performance and generalization capability to that of the original dataset. These results prove that these images are suitable for standardizing the evaluation of CAD pipelines, and providing data augmentation in machine learning systems -e.g. in deep learning-, or even to train future professionals at medical school.

Keywords: Alzheimer's Disease (AD); Neuroimaging; Parkinson's Disease (PD); Synthesis; data augmentation; density estimation; evaluation; validation.

Figure 1

Schema of the proposed synthesis procedure.

Figure 2

SPM analysis of the ADNI dataset (AD vs. NOR), Family-Wise Error (FWE) corrected, with p = 0.05, for the original and the synthetic images.

Figure 3

SPM analysis of the PPMI dataset (PD vs. NOR), Family-Wise Error (FWE) corrected, with p = 0.05, for the original and the synthetic images.

Figure 4

Evolution of the accuracy when varying the number of components L.

Figure 5

Evolution of the accuracy when varying the number of synthetic images N.

Figure 6

Outline of the experimental setup for E1.3 to test the generalization ability of the synthetic images.

Figure 7

Outline of the experimental setup for E2 to test the resubstitution and prediction accuracies.

Figure 8

Illustration of the first four eigenbrains (components 0 to 3) of the PET ADNI dataset.

Figure 9

Comparison between the MVN and KDE PDF estimation methods for the component 2 and 10 in the PET ADNI dataset (AD and NOR groups for simplicity), setting the histogram as reference.

Figure 10

Partial PDF modeling of component 2 (AD and NOR classes) using the MVN estimator with different Ls. PDFs scaled for comparison.

Figure 11

Examples of some original and synthetic subjects from the ADNI-PET dataset.

Figure 12

Examples of some original and synthetic subjects from the PPMI dataset.