Fluorodeoxyglucose PET in Neurology and Psychiatry

Low resolution of positron emission tomography (PET) limits its diagnostic performance. Deep learning has been successfully applied to achieve super-resolution PET. However, commonly used supervised learning methods in this context require many pairs of low- and high-resolution (LR and HR) PET images. Although unsupervised learning utilizes unpaired images, the results are not as good as that obtained with supervised deep learning. In this paper, we propose a quasi-supervised learning method, which is a new type of weakly-supervised learning methods, to recover HR PET images from LR counterparts by leveraging similarity between unpaired LR and HR image patches. Specifically, LR image patches are taken from a patient as inputs, while the most similar HR patches from other patients are found as labels. The similarity between the matched HR and LR patches serves as a prior for network construction. Our proposed method can be implemented by designing a new network or modifying an existing network. As an example in this study, we have modified the cycle-consistent generative adversarial network (CycleGAN) for super-resolution PET. Our numerical and experimental results qualitatively and quantitatively show the merits of our method relative to the state-of-the-art methods. The code is publicly available at https://github.com/PigYang-ops/CycleGAN-QSDL.

As a deep learning method, graph convolution network (GCN) has the advantage of dealing with non-Euclidean domain problems and is constantly applied in the research of computer-aided diagnosis of Alzheimer's disease (AD). In graph-based methods for AD diagnosis, nodes can represent potential subjects with a set of vectors, and edges combine the interaction and similarity between subjects. However, for the three-dimensional neuroimage like structural Magnetic Resonance Imaging (sMRI) or Positron Emission Tomography (PET), due to the non-sequential of ROI (Region of Interest) features (compared with four-dimensional neuroimage), which makes the graph-based analysis approach more difficult. In this study, we obtained individual features by constructing brain network via health group indirectly, and then constructed a population-based GCN framework by expressing the subject population as adjacency matrix in graph to achieve the diagnosis of AD. The nodes in graph are associated with individual features, and edges are weighted by combining the phenotypic information, we further discuss the influence of phenotypic information on the classification performance of GCN. Compared with acquiring the ROI features of the brain regions as the input features of GCN, our proposed method remarkably improved the prediction accuracy based on both sMRI and PET images by about 5 to 10 percentage. Through our testing and experimental analysis on the public ADNI dataset, our method achieved improved performance for AD diagnosis and mild cognitive impairment conversion prediction tasks. Our proposed method also provides technical support for AD diagnosis using GCN method based on three-dimensional brain images.

The lack of standardization of intensity normalization methods and its unknown effect on the quantification output is recognized as a major drawback for the harmonization of brain FDG-PET quantification protocols. The aim of this work is the ground truth-based evaluation of different intensity normalization methods on brain FDG-PET quantification output.

Realistic FDG-PET images were generated using Monte Carlo simulation from activity and attenuation maps directly derived from 25 healthy subjects (adding theoretical relative hypometabolisms on 6 regions of interest and for 5 hypometabolism levels). Single-subject statistical parametric mapping (SPM) was applied to compare each simulated FDG-PET image with a healthy database after intensity normalization based on reference regions methods such as the brain stem (RRBS), cerebellum (RRC) and the temporal lobe contralateral to the lesion (RRTL), and data-driven methods, such as proportional scaling (PS), histogram-based method (HN) and iterative versions of both methods (iPS and iHN). The performance of these methods was evaluated in terms of the recovery of the introduced theoretical hypometabolic pattern and the appearance of unspecific hypometabolic and hypermetabolic findings.

Detected hypometabolic patterns had significantly lower volumes than the introduced hypometabolisms for all intensity normalization methods particularly for slighter reductions in metabolism . Among the intensity normalization methods, RRC and HN provided the largest recovered hypometabolic volumes, while the RRBS showed the smallest recovery. In general, data-driven methods overcame reference regions and among them, the iterative methods overcame the non-iterative ones. Unspecific hypermetabolic volumes were similar for all methods, with the exception of PS, where it became a major limitation (up to 250 cm³) for extended and intense hypometabolism. On the other hand, unspecific hypometabolism was similar far all methods, and usually solved with appropriate clustering.

Our findings showed that the inappropriate use of intensity normalization methods can provide remarkable bias in the detected hypometabolism and it represents a serious concern in terms of false positives. Based on our findings, we recommend the use of histogram-based intensity normalization methods. Reference region methods performance was equivalent to data-driven methods only when the selected reference region is large and stable.

Biomarkers can be used in medicine for diagnosis, prognosis, or treatment guidance. In bipolar disorder, biomarkers have been studied not only to identify specific features associated with the diagnosis but also to detect illness activity related to mood episodes. Although the clinical use of biological markers is not possible yet, biomarker research in bipolar disorder has helped advance understanding of the pathophysiology and the neurobiological processes underlying this complex illness. With the increasing shift in psychiatry toward personalized medicine, biomarkers may aid in developing improved tools for risk assessment and treatment selection. In this chapter, we review candidate biomarkers studied in bipolar disorder, exploring evidence for the peripheral molecules (including BDNF, inflammatory markers, oxidative stress markers, and hypothalamic-pituitary-adrenal axis markers), “omics,” and neuroimaging markers, both structural and functional.