Lee, Jeyeon

AbstractBackgroundFluorodeoxyglucose positron emission tomography (FDG‐PET) is an established modality for differential diagnosis of dementia. Deriving phenotypic signatures from FDG‐PET via machine learning is challenging due to the high dimensional nature of brain images relative to the generally small number of labeled examples available for training, the class imbalance among those labels, and the cooccurrence of multiple pathologies. In this study, we developed a multi‐class, multi‐label framework to address these challenges.MethodA database of clinically acquired PET/CT images from 3,000 unique patients was used to develop a latent space model using matrix decomposition. This model was then applied to images from a separate cohort of Mayo Clinic Alzheimer’s Disease Research Center participants (n=1,745) labeled as cognitively unimpaired (CU) (n=1,436) or with the following potentially co‐occurring phenotypes: Alzheimer’s disease (AD) (n=165), Lewy body dementia (DLB) (n=92), behavioral variant frontotemporal dementia (bvFTD) (n=43), semantic (svPPA) (n=10) and logopenic (lvPPA) (n=13) variant PPA, and posterior cortical atrophy (PCA) (n=17). A k‐nearest neighbors classifier that is robust to these imbalanced and overlapping labels was then trained on these examples. The resulting classifier was evaluated by area under receiver‐operator characteristic curve (ROC‐AUC) via leave one out cross validation, using clinical diagnosis as the gold standard.ResultROC curves and AUC scores for each phenotype are illustrated in Fig. 1a. Because the classifier is based on a k‐nearest neighbors connectivity matrix, it has a convenient graphical representation, where images are nodes and edges are drawn between an image and its set of nearest neighbors in latent space. A self‐organizing force directed graph constructed in this way is illustrated in Fig 1b, highlighting the strong separation of CU and degenerative images, as well as the segregation of each phenotype within the neurodegenerative region of the graph.ConclusionIn this study, we developed a machine learning framework for classification of neurodegenerative disease based on k‐nearest neighbor analysis in a low dimensional latent space projection of FDG‐PET images. By leveraging low‐dimensional representations and k‐nearest neighbors analysis, this framework is robust in multi‐class, multi‐label tasks with strong class imbalance and provides a highly interpretable graphical representation.

AbstractBackgroundIn vivo tau‐positron emission tomography (PET) is an attractive biomarker for Alzheimer’s disease (AD) diagnosis and treatment. However, tau‐PET is less widely available than other modalities. In this study, we tested cross‐modality synthesis of tau‐PET brain images from fluorodeoxyglucose F‐18 (FDG)‐PET using a deep convolutional neural network (CNN).MethodParticipants (n=1,192) who had brain FDG‐PET with 18F‐FDG and tau‐PET with Flortaucipir (F‐18‐AV‐1451) were included for training and testing. This cohort spanned normal aging (ages 26‐98), pre‐clinical, and clinical AD and related disorders including the FTD and DLB spectrum. External validation was done using ADNI (n=288). The PET scans were co‐registered to the corresponding MRI and subsequently warped to Mayo Clinic Adult Lifespan Template (MCALT) space. Tau‐PET images were SUVR‐normalized to the cerebellar crus, and FDG to the pons. A 3D dense‐U‐net model was utilized as an architecture. Cross‐validation experiments were conducted using 5‐fold validations (60% training set, 20% validation set, and 20% test set) with mean squared error as the loss function.ResultOur dense‐U‐net model successfully synthesized tau‐PET from metabolic images with good correlation and low prediction error for regional SUVRs (Figure 1A‐C). The model showed a robust prediction ability, performing accurately in an independent, external ADNI cohort (Figure 1D‐F). The model‐imputed tau‐PET significantly improved performance in classifying tau positivity (mean AUROC(±SD)=0.78±0.04 and 0.85±0.03 for FDG‐PET and synthesized tau‐PET, respectively) and diagnostic groups (cognitively unimpaired with abnormal amyloid‐PET vs. cognitively impaired with abnormal amyloid‐PET) compared to the original input FDG data (mean AUROC(±SD)=0.89±0.04, 0.85±0.05 0.91±0.04 for actual tau‐PET, FDG‐PET and synthesized tau‐PET, respectively) (Figure 2), suggesting enhanced clinical utility for metabolic images. The ADNI cohort also showed similar results (for tau positivity: AUROC=0.66 and 0.78 for FDG‐PET and synthesized tau‐PET, respectively; for CU A+ vs. CI A+: AUROC=0.86, 0.62, and 0.73 for actual tau‐PET, FDG‐PET, and synthesized tau‐PET; Figure 3).ConclusionWe showed that using a CNN model to predict tau‐PET from FDG‐PET is feasible. The synthesized tau‐PET can augment the value of FDG‐PET, facilitating the multi‐modal diagnosis of AD.

AbstractBackgroundDysexecutive Alzheimer's disease (dAD) is a relatively early‐onset variant of AD primarily degrading core executive functions in the absence of predominant behavioral symptoms. Clinical observations suggest substantial variability in clinical, cognitive and neuroimaging profiles within this syndrome. In this presentation, we will discuss our ongoing work disentangling this heterogeneity using unsupervised machine learning techniques.MethodWe collected clinical, neuropsychological and multimodal imaging (MRI, FDG‐PET, amyloid‐PET, tau‐PET) data on 52 dAD patients assessed in our behavioral neurology clinic at Mayo Clinic Rochester. We first performed a spectral clustering analysis based on FDG‐PET to delineate latent patterns of network degeneration (referred as “eigenbrains”) and assessed the relationships between these eigenbrains and clinical and cognitive symptomatology. We then performed a hierarchical clustering on these eigenbrains to derive data‐driven subtypes of dAD. We compared clinical and cognitive data between subtypes, and then compared the imaging profiles to 52 amyloid‐negative age‐ and sex‐matched controls.ResultSix eigenbrains explained approximatively 48% of the variance in FDG‐PET patterns and primarily reflected heteromodal to primary motor/sensory, left‐right hemispheric asymmetry, and anterior‐to‐posterior gradients of macro‐scale cortical organization (Fig 1). These eigenbrains differentially related to reported age at symptom onset, degree of clinical impairment, and performance on a wide range of cognitive domains (executive functions, episodic memory, and visuospatial). Hierarchical clustering revealed four dAD subtypes (diffuse, biparietal, left‐hemisphere, right‐hemisphere). These subtypes differed in reported age at symptom onset and cognitive profile, where the heteromodal‐diffuse subtype exhibited an overall worse clinical picture and the biparietal had a milder profile compared to other subtypes, which was not explained by disease duration. Additionally, spatial patterns of tau distribution and neurodegeneration overlapped with patterns of FDG‐PET hypometabolism in each subtype, whereas patterns of amyloid deposition were similar across subtypes (Figs 2 & 3).ConclusionAlmost half of the variance in macro‐scale patterns of hypometabolism in this dAD cohort was accounted for by six eigenbrains. These eigenbrains captured the inter‐individual variability in age at symptom onset and cognitive impairment. Four dAD subtypes derived from these eigenbrains revealed meaningful differences in clinical, cognitive, and imaging profiles. Recognizing this heterogeneity has significant clinical implications for diagnosis, prognosis, and symptom monitoring.

AbstractBackgroundFluorodeoxyglucose positron emission tomography (FDG‐PET) is an established modality for differential diagnosis of dementia. Deriving phenotypic signatures from FDG‐PET via machine learning is challenging due to the high dimensional nature of brain images relative to the generally small number of labeled examples available for training, the class imbalance among those labels, and the cooccurrence of multiple pathologies. In this study, we developed a multi‐class, multi‐label framework to address these challenges.MethodA database of clinically acquired PET/CT images from 3,000 unique patients was used to develop a latent space model using matrix decomposition. This model was then applied to images from a separate cohort of Mayo Clinic Alzheimer’s Disease Research Center participants (n=1,745) labeled as cognitively unimpaired (CU) (n=1,436) or with the following potentially co‐occurring phenotypes: Alzheimer’s disease (AD) (n=165), Lewy body dementia (DLB) (n=92), behavioral variant frontotemporal dementia (bvFTD) (n=43), semantic (svPPA) (n=10) and logopenic (lvPPA) (n=13) variant PPA, and posterior cortical atrophy (PCA) (n=17). A k‐nearest neighbors classifier that is robust to these imbalanced and overlapping labels was then trained on these examples. The resulting classifier was evaluated by area under receiver‐operator characteristic curve (ROC‐AUC) via leave one out cross validation, using clinical diagnosis as the gold standard.ResultROC curves and AUC scores for each phenotype are illustrated in Fig. 1a. Because the classifier is based on a k‐nearest neighbors connectivity matrix, it has a convenient graphical representation, where images are nodes and edges are drawn between an image and its set of nearest neighbors in latent space. A self‐organizing force directed graph constructed in this way is illustrated in Fig 1b, highlighting the strong separation of CU and degenerative images, as well as the segregation of each phenotype within the neurodegenerative region of the graph.ConclusionIn this study, we developed a machine learning framework for classification of neurodegenerative disease based on k‐nearest neighbor analysis in a low dimensional latent space projection of FDG‐PET images. By leveraging low‐dimensional representations and k‐nearest neighbors analysis, this framework is robust in multi‐class, multi‐label tasks with strong class imbalance and provides a highly interpretable graphical representation.

Abstract The external globus pallidus (GPe) guides action selection. However, it remains elusive how diverse GPe cells engage in routine formation while learning action sequences in repetitive reward-seeking conditioning. Here, we investigated the Ca2+ signals of GPe cells during routine formation. We found that the dynamics of astrocytes and prototypic neurons’ activities are associated with the refinement of the action sequence, which enables cost-effective reward-seeking.

AbstractBackgroundIn longitudinal tau (flortaucipir) positron emission tomography (PET), a reliable method to detect a tau accumulation while minimizing a random variability is crucial. Variability can be a problem especially in the early stages of tau pathology in which only small regions and low levels of accumulation may exist relative to the range of random noise fluctuation.MethodWe recruited 76 participants who had serial scans for a total of 234 tau PET scans. In this cohort we tested a PET signal measurement method we developed called the “overlap index” (OI). The OI was calculated by identifying voxels that have elevated tau PET signal that overlap on two successive tau PET scans in the same participant (Figure 1). This analysis procedure tests the hypothesis that voxel‐wise, elevated‐uptake consistency might be more sensitive for early detection of tau tangles than standard ROI based measures. We compared OI to ROI based measures for early detection of increasing tau PET signal and detection clinical progression. A meta ROI used in this analysis included the amygdala, entorhinal cortex, fusiform, parahippocampal, and inferior temporal and middle temporal gyri.ResultOI showed regional progression of tau in the temporal lobe followed by increases throughout the brain at relatively low standardized uptake value ratio (SUVr) levels (Figure 2A and B). The topographical pattern of OI was in agreement in some respects with the sequential appearance of tau aggregation (Figure 2B and C). The high‐OI group (OI>0.5) had a significantly higher ΔSUVr compared to the low‐OI group (OI<0.5) (Figure 3A and B). OI identified subjects that did or didn’t have tau accumulation within meta SUVr subgroups even when below the meta ROI SUVr threshold. OI was associated with early clinical disease stage progression vs meta‐ROI SUVr measures (Figure 3C‐E).ConclusionIn this study, we found that OI tau PET measurement was able to quantify voxel‐wise consistency using a set of serial scans. OI may detect earlier tau tangle accumulation and earlier disease progression than standard SUVr measurements in tau PET.

Abstract The external globus pallidus (GPe) is an integrative hub and gateway for behavioral flexibility in reward-related behaviors. However, it remains unknown whether enriched astrocytes in the GPe guide behavioral flexibility. Here, we trained mice to exhibit goal-directed and habitual reward-seeking behaviors using the behavior tasks with effort- and time-based reward delivery, respectively. Then, we examined the temporal dynamics of GPe astrocytes during goal-directed and habitual learning. Overall, GPe astrocytes were substantially silenced during habitual learning compared to goal-directed learning. In the timescale of action events, GPe astrocyte activities were increased immediately after termination of reward-taking behavior before the following action. However, during habitual learning, the increase of astrocyte activity was not evident. Moreover, support vector machine (SVM) analysis demonstrated that GPe astrocytes dynamics predicted whether mice perform goal-directed or habitual behaviors. Interestingly, chemogenetic activation of GPe astrocytes, which dampened GPe neuronal firings and habitual behaviors, exhibting goal-directed behaviors. Strikingly, brief and repeated attentional stimulations recapitulated the effect of chemogenetic activation of GPe in intervening the habitual reward-seeking behaviors with increased GPe astrocyte activities. Our findings reveal a novel insight that increasing GPe astrocytic activities attenuates habitual behavior and improves behavioral flexibility, which may provide a potential therapeutic target for decision-making-related disorders, such as obsessive-compulsive disorder and addiction.

Abstract Normal brain aging is accompanied by patterns of functional and structural change. Alzheimer's disease (AD), a representative neurodegenerative disease, has been linked to accelerated brain aging at respective age ranges. Here, we developed a deep learning-based brain age prediction model using fluorodeoxyglucose (FDG) PET and structural MRI and tested how the brain age gap relates to degenerative cognitive syndromes including mild cognitive impairment, AD, frontotemporal dementia, and Lewy body dementia. Occlusion analysis, performed to facilitate interpretation of the model, revealed that the model learns an age- and modality-specific pattern of brain aging. The elevated brain age gap in dementia cohorts was highly correlated with the cognitive impairment and AD biomarker. However, regions generating brain age gaps were different for each diagnosis group of which the AD continuum showed similar patterns to normal aging in the CU.

Abstract In this study, we explored the feasibility of using functional ultrasound (fUS) imaging to visualize cerebral activation associated with thalamic deep brain stimulation (DBS), in rodents. The ventrolateral (VL) thalamus was stimulated using electrical pulses of low and high frequencies of 10 and 100 Hz, respectively, and multiple voltages (1–7 V) and pulse widths (50–1500 μ s). The fUS imaging demonstrated DBS-evoked activation of cerebral cortex based on changes of cerebral blood volume, specifically at the primary motor cortex (PMC). Low frequency stimulation (LFS) demonstrated significantly higher PMC activation compared to higher frequency stimulation (HFS), at intensities (5–7 V). Whereas, at lower intensities (1–3 V), only HFS demonstrated visible PMC activation. Further, LFS-evoked cerebral activation was was primarily located at the PMC. Our data presents the functionality and feasibility of fUS imaging as an investigational tool to identify brain areas associated with DBS. This preliminary study is an important stepping stone towards conducting real-time functional ultrasound imaging of DBS in awake and behaving animal models, which is of significant interest to the community for studying motor-related disorders.

Habitual reward‐seeking behavior is a hallmark of addictive behaviors. Although habitual behavior could be reversed to goal‐directed behavior, this shifting is not well studied. Here, we identify a role of astrocyte activation in the dorsomedial striatum (DMS) on transition from habitual to goal‐directed reward‐seeking behavior. Utilizing the Inscopix endomicroscopy/fiber photometry in vivo calcium imaging and the cell‐type specific activation of designer receptors exclusively activated by designer drugs (DREADDs), we confirmed that GFAP‐driven activation of hM3Dq DREADDs increases intensity and frequency of calcium signaling in the DMS astrocytes of ALDH1L1‐GCaMP6s expressing mice. We found that this chemogenetic activation of the DMS astrocytes increase and decrease neuronal excitability of direct pathway Medium Spiny Neurons (dMSNs) and indirect pathway Medium Spiny Neurons (iMSNs), respectively. Specifically, it reduced the frequency of spontaneous excitatory postsynaptic currents (sEPSCs) in the dMSNs, whereas it increased the amplitude of the sEPSCs and decreased the frequency of spontaneous inhibitory postsynaptic currents (sIPSCs) in the iMSNs. The changes in synaptic events induced by astrocytic activation were significantly inhibited by either of the pretreatments of DPCPX (Adenosine A1R antagonist, 1 μM), PSB12379 (Inhibitor of ectonucleotidase, CD73, 10 μM), NBTI [Inhibitor of the astrocytic adenosine transporter, equilibrative nucleoside transporter 1 (ENT1, slc29a1), 10 μM], and genetic deletion of ENT1. These data suggest that astrocytic activation‐induced synaptic activities are, in part, adenosine‐dependent. In the evaluation tests of goal‐directed and habitual reward‐seeking, the chemogenetic activation of the DMS astrocytes shifted the habitual behavior to goal‐directed action in sucrose reward‐seeking for WT, but not ENT1 KO mice. The astrocytic activation‐induced shifting was rescued when ENT1 expression was normalized in the DMS astrocytes of the ENT1 KO mice. On the other hand, when mice were trained to have goal‐directed behavior, the astrocytic activation in the DMS did not change the reward‐seeking behavioral pattern. Together, our results indicate that astrocyte‐regulated adenosine signaling in the DMS determines goal‐directed and habitual behaviors by selectively regulating synapses in the MSNs. The interaction between astrocytes and neurons via adenosine signaling could be a potential therapeutic target for maladaptive reward‐seeking behaviors. Support or Funding Information This work was supported by the Samuel C. Johnson for Genomics of Addiction Program at Mayo Clinic, the Ulm Foundation, and National Institute on Alcohol Abuse and Alcoholism (K01 AA027773 to SK, R01 AA018779 to DSC).