Explore the vast range of titles available.

Assessment of individuals on the autism spectrum often includes a measure of nonverbal IQ. One such measure is the Raven’s Standard Progressive Matrices (RSPM). For large research studies with participants distributed nationally it is desirable for assessments to be available online. Because time is a premium, it is ideal that the measure produces accurate scores quickly. The Hansen Research Services Matrix Adaptive Test (HRS-MAT) addresses these needs and with similar psychometric properties of the RSPM. Scores based on the HRS-MAT correlated at r  = .81 with those of the RSPM. In adult-child pairs, HRS-MAT scores correlated at approximately r  = .50. Details from respondents in a national sample and psychometric properties including reliability and validity are discussed.

We compare the financial literacy of two groups of Koreans living in South Korea, namely, native-born South Koreans and North Korean refugees, who were born and raised in contrasting economic systems. Examining the financial literacy of North Korean refugees and its changes over time after their settlement in a capitalistic society underscores the importance of institutional environments in developing financial literacy. We find that North Korean refugees, with very limited access to financial markets in their home country, are significantly less financially literate than native-born South Koreans. The gap is significant even after controlling for cognitive ability, which is also starkly different between the two groups. The financial literacy of the refugees increases over time during their settlement in South Korea, but the magnitude of such improvement is insubstantial. Our findings suggest that financial literacy is developed at the early life stages and cannot be easily modified at the later stages.

The recent advent of large language models has reinvigorated debate over whether human cognitive capacities might emerge in such generic models given sufficient training data. Of particular interest is the ability of these models to reason about novel problems zero-shot, without any direct training. In human cognition, this capacity is closely tied to an ability to reason by analogy. Here we performed a direct comparison between human reasoners and a large language model (the text-davinci-003 variant of Generative Pre-trained Transformer (GPT)-3) on a range of analogical tasks, including a non-visual matrix reasoning task based on the rule structure of Raven’s Standard Progressive Matrices. We found that GPT-3 displayed a surprisingly strong capacity for abstract pattern induction, matching or even surpassing human capabilities in most settings; preliminary tests of GPT-4 indicated even better performance. Our results indicate that large language models such as GPT-3 have acquired an emergent ability to find zero-shot solutions to a broad range of analogy problems. Webb et al. show that new artificial intelligence language models, such as Generative Pre-trained Transformer 3, are able to solve analogical reasoning problems at a human-like level of performance.

Background The agreement between plasma Aβ42/40 and Aβ‐PET is approximately 75%, with a large portion of discrepancies due to positive plasma with negative PET results. Questions remain about whether these reflect brain Aβ changes detectable in plasma before PET‐detectable. We aimed to examine these cases over 11 years to assess the risk and timing of progression to Aβ‐PET positivity. Method Cognitively unimpaired participants from large‐scale longitudinal studies of AIBL, OASIS, and ADNI underwent baseline Aβ‐PET and plasma Aβ42/40 analysis by IPMS, followed by 1‐7 additional PET scans every 1.5‐3 years. Aβ‐PET was quantified to Centiloid (CL) using the SPM pipeline. Individuals with baseline Aβ‐PET < 20 CL (n = 507) were included, with those < 5 CL classified as PET‐, and 5‐20 CL as PETLow. Plasma ‐/+ was based on the Aβ42/40 Youden's Index threshold (0.119) corresponding to Aβ‐PET status. We used Kaplan‐Meier method and Cox proportional hazards analysis to assess the risk of progression to PET+ (> 20 CL). Result Plasma+/PET‐ (< 5 CL) individuals were at higher risk than Plasma‐/PET‐ of progressing to PET+ (hazard ratio (HR): 3.90 [95% CI: 2.00‐7.61], p <0.001), even after matching the groups’ baseline CL values (HR: 3.43 [1.43‐8.26], p = 0.010), or adjusting for age, sex, APOE ε4 and baseline CL (HR: 2.48 [1.22 ‐ 5.07], p = 0.013) (Figure 1A). Plasma+/PET‐ accumulated brain Aβ ∼8 times faster than Plasma‐/PET‐ (1.14 CL/year vs. 0.15 CL/year respectively, p <0.001). Plasma+/PET‐ progressors became PET+, on average, 2 years earlier than Plasma‐/PET‐ progressors. Plasma+/PETLow group had faster decline in survival probability than Plasma‐/PETLow (HR: 20.82 [11.28 – 38.42], p <0.001 vs. HR: 6.67 [3.51 – 12.65], p <0.001) (Figure 1B) but this was driven by higher CL in the Plasma+ group. Conclusion Cognitively unimpaired individuals with abnormal plasma Aβ42/40 but negative Aβ‐PET face a significantly increased risk of future positive Aβ‐PET. This provides supporting evidence that brain Aβ pathology can be detected in plasma with IPMS before it is PET‐detectable. Whether this also applies to plasma Aβ42/40 immunoassays warrants investigation.

Background The agreement between plasma Aβ42/40 and Aβ‐PET is approximately 75%, with a large portion of discrepancies due to positive plasma with negative PET results. Questions remain about whether these reflect brain Aβ changes detectable in plasma before PET‐detectable. We aimed to examine these cases over 11 years to assess the risk and timing of progression to Aβ‐PET positivity. Method Cognitively unimpaired participants from large‐scale longitudinal studies of AIBL, OASIS, and ADNI underwent baseline Aβ‐PET and plasma Aβ42/40 analysis by IPMS, followed by 1‐7 additional PET scans every 1.5‐3 years. Aβ‐PET was quantified to Centiloid (CL) using the SPM pipeline. Individuals with baseline Aβ‐PET < 20 CL (n = 507) were included, with those < 5 CL classified as PET‐, and 5‐20 CL as PETLow. Plasma ‐/+ was based on the Aβ42/40 Youden’s Index threshold (0.119) corresponding to Aβ‐PET status. We used Kaplan‐Meier method and Cox proportional hazards analysis to assess the risk of progression to PET+ (> 20 CL). Result Plasma+/PET‐ (< 5 CL) individuals were at higher risk than Plasma‐/PET‐ of progressing to PET+ (hazard ratio (HR): 3.90 [95% CI: 2.00‐7.61], p<0.001), even after matching the groups’ baseline CL values (HR: 3.43 [1.43‐8.26], p = 0.010), or adjusting for age, sex, APOE ε4 and baseline CL (HR: 2.48 [1.22 ‐ 5.07], p = 0.013) (Figure 1A). Plasma+/PET‐ accumulated brain Aβ ∼8 times faster than Plasma‐/PET‐ (1.14 CL/year vs. 0.15 CL/year respectively, p <0.001). Plasma+/PET‐ progressors became PET+, on average, 2 years earlier than Plasma‐/PET‐ progressors. Plasma+/PETLow group had faster decline in survival probability than Plasma‐/PETLow (HR: 20.82 [11.28 – 38.42], p <0.001 vs. HR: 6.67 [3.51 – 12.65], p <0.001) (Figure 1B) but this was driven by higher CL in the Plasma+ group. Conclusion Cognitively unimpaired individuals with abnormal plasma Aβ42/40 but negative Aβ‐PET face a significantly increased risk of future positive Aβ‐PET. This provides supporting evidence that brain Aβ pathology can be detected in plasma with IPMS before it is PET‐detectable. Whether this also applies to plasma Aβ42/40 immunoassays warrants investigation.

Background Convolutional neural networks (CNNs) excel in classifying dementia subtypes from T1‐weighted MRI scans, yet their limited interpretability hinders clinical adoption. Classical machine learning (ML) models, while typically less performant, offer greater explainability by utilising predefined anatomically‐informed features. However, a systematic comparison of the many popular T1 MRI preprocessing pipelines for feature extraction — particularly in the context of dementia — is lacking. Method This study evaluates seven widely used MRI preprocessing pipelines (FSL, SPM, CAT12 VBM, CAT12 SBM, SynthSeg, Freesurfer, FastSurfer) across five open‐source dementia imaging datasets (ADNI, OASIS, AIBLE, NIFD, NACC). Regional volumetric and surface‐based features extracted by each pipeline were input into six classical ML classifiers: support vector machines (SVM), random forest, logistic regression, naïve Bayes, k‐nearest neighbors (kNN), and XGBoost. Performance metrics, including accuracy, precision, sensitivity, and AUC‐ROC, were used to evaluate the classification of dementia subtypes (including Alzheimer’s disease and frontotemporal dementia). Additionally, raw MRI scans are classified using several 3D CNN architectures, including a modified ResNet18, to benchmark feature‐based ML models against CNN performance. The analysis pipeline is summarised in Figure 1. Result Preliminary analysis on the ADNI dataset identified Freesurfer as the best‐performing preprocessing pipeline, achieving a mean accuracy of 0.872 across ML classifiers. Random forest emerged as the top‐performing model overall, with a mean accuracy of 0.840 across pipelines. Shapley value analysis for Freesurfer features revealed consistent influential regions across ML models; for example, Figure 2 highlights similarities between SVM and logistic regression models. The CNN results are currently being generated and will be available for comparison at the conference. Conclusion This study highlights the strengths and limitations of different preprocessing‐ML combinations for dementia classification, providing a framework for optimising performance while enhancing interpretability. By systematically comparing these pipelines to CNNs, we aim to identify clinically viable alternatives that balance accuracy, explainability, and computational efficiency.

Background We evaluated the predictive performance of 18F‐flortaucipir (FTP) tau imaging within the NIA‐AA multilevel tau staging framework with respect to tau accumulation and cognitive decline in non‐demented individuals. We also tested the relationships of cognitive measures with baseline tau and tau accumulation. Methods FTP scans from 213 non‐demented participants were processed and sampled in Statistical Parametric Mapping software (SPM), version 8, using CenTauR masks. Tau accumulation and cognitive decline associations were assessed longitudinally, with respect to two timepoints, their baseline and most recent evaluations, via survival analysis. Individuals were categorized into 4 groups reflecting the NIA‐AA imaging stages: Initial, with only b‐amyloid (Ab) pathology was present in PET; Early, with Ab pathology and tau pathology in the mesial temporal region; Intermediate, with moderate tau pathology in the meta temporal region; and Advanced, with high levels of tau in the meta temporal region. A “None” group reflecting no pathology was included as a control. Linear regressions were used to compare the longitudinal effects of either baseline tau (SUVR) or tau accumulation (SUVR/year) on cognitive decline. Results While the two sets of thresholds yielded slightly different trajectories, both showed that when applying multiple levels of tau positivity, increasing stages of tau predicted both earlier tau accumulation and earlier cognitive decline. Linear regressions revealed that change in global measures of cognition (MMSE, CDR‐SB) were significantly associated with baseline tau, while decline in Delayed Recall (DR) was significantly associated with both baseline tau and tau accumulation, where tau accumulation had a greater influence in the model, and Immediate Recall (LM) decline was significantly only associated with tau accumulation. Conclusions Implementing the multiple tau stages from the new NIA‐AA biological staging framework clearly predicts distinct patterns of tau accumulation and cognitive decline. While baseline tau is predictive of global cognitive decline, tau accumulation is a better predictor of memory decline. Future work is needed to determine how the thresholds utilized here compare to visual reads and to determine the suitability of these thresholds in differentiating trajectories of individuals with cognitive impairment.

Background Convolutional neural networks (CNNs) excel in classifying dementia subtypes from T1‐weighted MRI scans, yet their limited interpretability hinders clinical adoption. Classical machine learning (ML) models, while typically less performant, offer greater explainability by utilising predefined anatomically‐informed features. However, a systematic comparison of the many popular T1 MRI preprocessing pipelines for feature extraction — particularly in the context of dementia — is lacking. Method This study evaluates seven widely used MRI preprocessing pipelines (FSL, SPM, CAT12 VBM, CAT12 SBM, SynthSeg, Freesurfer, FastSurfer) across five open‐source dementia imaging datasets (ADNI, OASIS, AIBLE, NIFD, NACC). Regional volumetric and surface‐based features extracted by each pipeline were input into six classical ML classifiers: support vector machines (SVM), random forest, logistic regression, naïve Bayes, k‐nearest neighbors (kNN), and XGBoost. Performance metrics, including accuracy, precision, sensitivity, and AUC‐ROC, were used to evaluate the classification of dementia subtypes (including Alzheimer's disease and frontotemporal dementia). Additionally, raw MRI scans are classified using several 3D CNN architectures, including a modified ResNet18, to benchmark feature‐based ML models against CNN performance. The analysis pipeline is summarised in Figure 1. Result Preliminary analysis on the ADNI dataset identified Freesurfer as the best‐performing preprocessing pipeline, achieving a mean accuracy of 0.872 across ML classifiers. Random forest emerged as the top‐performing model overall, with a mean accuracy of 0.840 across pipelines. Shapley value analysis for Freesurfer features revealed consistent influential regions across ML models; for example, Figure 2 highlights similarities between SVM and logistic regression models. The CNN results are currently being generated and will be available for comparison at the conference. Conclusion This study highlights the strengths and limitations of different preprocessing‐ML combinations for dementia classification, providing a framework for optimising performance while enhancing interpretability. By systematically comparing these pipelines to CNNs, we aim to identify clinically viable alternatives that balance accuracy, explainability, and computational efficiency.

Background Microglia can take on proinflammatory or anti‐inflammatory phenotypes, but it is unclear how these phenotypes play out along the Alzheimer's disease (AD) continuum. The purpose was to assess regional variances of microglial activation in distinct stages of the disease, and the role of microglial responses on grey matter volume and mean diffusivity in different brain areas and on cognition across the course of AD. Method 48 subjects (23 AD patients, 14 mild cognitive impairment [MCI], and 11 healthy controls [HC]) underwent TSPO‐PET, diffusion tensor imaging (DTI) and extensive neuropsychometric assessment. SPM (Statistical parametric mapping) analysis was conducted for single subject analysis. GM volume for each subject was derived from T1 volumetric MRI using the FreeSurfer pipeline, mean diffusivity (DTI), alongside voxel‐based morphometric analysis. Using postmortem brain tissue from 26 AD subjects, we evaluated CD32a and CD163 anti‐bodies, markers of proinflammatory and anti‐inflammatory microglia, respectively. Result Voxel‐wise analyses identified positive and negative clusters of associations for IRF90 with GM volume and mean diffusivity. The presence of pro‐ and anti‐inflammatory phenotypes was confirmed in human postmortem brain, in the frontal, parietal and entorhinal cortices. In addition, higher TSPO, signal in temporal lobes associated with lower hippocampal volume and mean diffusivity, and higher TSPO, signal across the cortex, correlated with poorer performance on neuropsychometric tests. Conclusion There may be protective and deleterious microglial phenotypes in an individual, region specific and disease stage dependent. Therapeutic modulation of microglia is warranted to promote anti‐inflammatory functions while suppressing pro‐inflammatory functions.

Background Specialized pro‐resolving mediators (SPMs) promote inflammatory resolution and homeostasis and are thought to have specific reprogramming effects on human microglia. Decreased SPM levels have been correlated with chronic neuroinflammation, late‐stage Alzheimer's disease (AD) and neuropathology in humans, yet few studies have explored the cellular signatures of resolution. Amyloid is though to bind one target resolution receptor, ChemR23, leading to internalization. In this study we explore if chronic administration of the specific ChemR23‐targeted SPM, Resolvin E1 (RvE1), alters memory performance and inflammation 5xFAD transgenic mice at 12 months of age when pathology burden and memory impairment is substantial. We hypothesize the RvE1 may have significant recuse potential. Method At 11 months of age we administered, RvE1 (10 per kg body weight/day), for 1.5 months with subcutaneous mini‐osmotic pumps. A 9‐day sequential series of memory performance was administered before and after treatment to quantify novel object recognition, hyperactivity, spatial context memory, and cognitive resilience. We quantified cytokine responses in blood and brain tissue using standard immuno‐dot blots, multiplexed ELISA and quantitative RNA assay. We measured RvE1 and other lipids from brain and other organs to confirm pharmacokinetics. We examined the occurrence and degree of plaque and microglial morphology in multiple brain regions using immunofluorescence microscopy. Result RvE1 significantly reduces microglial activation, cytokine expression and plaque burden in amyloid challenged 5xFAd mice. We discovered that RvE1 leads to compensatory response in specific ChemR23 receptor and other resolution‐oriented G protein coupled receptor family. We are performing single cell analysis to better ascertain how RvE1 altered cellular resolution response in different classes of cells. Conclusion We have demonstrated that RvE1 modulates inflammation in a mouse model and in a microglial cell line. Resolution pathways may represent drugable targets to reduce neuroinflammation associated with neuropathology, but a more thorough understanding of how RvE1 achieves therapeutic benefits is needed.