Adaptive Memory Research Articles

Using deep learning to quantify neuronal activation from single-cell and spatial transcriptomic data

Neuronal activity-dependent transcription directs molecular processes that regulate synaptic plasticity, brain circuit development, behavioral adaptation, and long-term memory. Single cell RNA-sequencing technologies (scRNAseq) are rapidly developing and allow for the interrogation of activity-dependent transcription at cellular resolution. Here, we present NEUROeSTIMator, a deep learning model that integrates transcriptomic signals to estimate neuronal activation in a way that we demonstrate is associated with Patch-seq electrophysiological features and that is robust against differences in species, cell type, and brain region. We demonstrate this method’s ability to accurately detect neuronal activity in previously published studies of single cell activity-induced gene expression. Further, we applied our model in a spatial transcriptomic study to identify unique patterns of learning-induced activity across different brain regions in male mice. Altogether, our findings establish NEUROeSTIMator as a powerful and broadly applicable tool for measuring neuronal activation, whether as a critical covariate or a primary readout of interest.

BCG immunization induces CX3CR1hi effector memory T cells to provide cross-protection via IFN-γ-mediated trained immunity.

After a century of using the Bacillus Calmette-Guérin (BCG) vaccine, our understanding of its ability to provide protection against homologous (Mycobacterium tuberculosis) or heterologous (for example, influenza virus) infections remains limited. Here we show that systemic (intravenous) BCG vaccination provides significant protection against subsequent influenza A virus infection in mice. We further demonstrate that the BCG-mediated cross-protection against influenza A virus is largely due to the enrichment of conventional CD4+ effector CX3CR1hi memory αβ T cells in the circulation and lung parenchyma. Importantly, pulmonary CX3CR1hi T cells limit early viral infection in an antigen-independent manner via potent interferon-γ production, which subsequently enhances long-term antimicrobial activity of alveolar macrophages. These results offer insight into the unknown mechanism by which BCG has persistently displayed broad protection against non-tuberculosis infections via cross-talk between adaptive and innate memory responses.

Adaptive Sparse Memory Networks for Efficient and Robust Video Object Segmentation.

Recently, memory-based networks have achieved promising performance for video object segmentation (VOS). However, existing methods still suffer from unsatisfactory segmentation accuracy and inferior efficiency. The reasons are mainly twofold: 1) during memory construction, the inflexible memory storage mechanism results in a weak discriminative ability for similar appearances in complex scenarios, leading to video-level temporal redundancy, and 2) during memory reading, matching robustness and memory retrieval accuracy decrease as the number of video frames increases. To address these challenges, we propose an adaptive sparse memory network (ASM) that efficiently and effectively performs VOS by sparsely leveraging previous guidance while attending to key information. Specifically, we design an adaptive sparse memory constructor (ASMC) to adaptively memorize informative past frames according to dynamic temporal changes in video frames. Furthermore, we introduce an attentive local memory reader (ALMR) to quickly retrieve relevant information using a subset of memory, thereby reducing frame-level redundant computation and noise in a simpler and more convenient manner. To prevent key features from being discarded by the subset of memory, we further propose a novel attentive local feature aggregation (ALFA) module, which preserves useful cues by selectively aggregating discriminative spatial dependence from adjacent frames, thereby effectively increasing the receptive field of each memory frame. Extensive experiments demonstrate that our model achieves state-of-the-art performance with real-time speed on six popular VOS benchmarks. Furthermore, our ASM can be applied to existing memory-based methods as generic plugins to achieve significant performance improvements. More importantly, our method exhibits robustness in handling sparse videos with low frame rates.

Tunable insulator-metal transition in epitaxial VO2 thin films via strain and defect engineering.

The Metal to Insulator Transition (MIT) in materials, particularly vanadium dioxide (VO2), has garnered significant research interest due to its potential applications in smart windows, memristors, transistors, sensors, and optical switches. The transition from an insulating, monoclinic phase to a conducting, tetragonal phase involves changes in optical and electrical properties, opening avenues in adaptive radiative coolers, optical memories, photodetectors, and optical switches. VO2 exhibits MIT close to 68 °C, thereby requiring tuneable transition temperatures (T c) in VO2 thin films for practical device applications. In this work, we explore the role of strain and defect engineering in tuning the MIT temperature in epitaxial VO2 thin films deposited on c-cut sapphire using Pulsed Laser Deposition (PLD). The study involves tuning the metal-to-insulator transition (MIT) by varying growth parameters, mainly temperature and oxygen partial pressure. Strain engineering along the b-axis helped tune the transition temperature from 65 °C to 82 °C with the out-of-plane b-strain varying from -0.71% to -0.44%. Comprehensive structural and property analyses, including X-ray diffraction (XRD), Reciprocal Space Mapping (RSM), X-ray Photoelectron Spectroscopy (XPS), Raman spectroscopy, and resistivity-temperature (R-T) measurements, were performed to correlate structural properties with T c. Additionally, density functional theory (DFT) calculations were performed using Quantum Espresso within the generalized gradient approximation of the revised Perdew-Burke-Ernzerhof (PBEsol) functional to provide theoretical validity to the experimentally obtained results. Our study provides critical insights into the interplay between strain and oxygen vacancies and their effect on the physical properties of VO2 thin films with DFT calculations supporting the experimental findings.

Biomarkers-Aware Asymmetric Bibranch GAN with Adaptive Memory Batch Normalization for Prediction of Anti-VEGF Treatment Response in Neovascular Age-related Macular Degeneration.

The emergence of anti-vascular endothelial growth factor (anti-VEGF) therapy has revolutionized neovascular age-related macular degeneration (nAMD). Post-therapeutic optical coherence tomography (OCT) imaging facilitates the prediction of therapeutic response to anti-VEGF therapy for nAMD. Although the generative adversarial network (GAN) is a popular generative model for post-therapeutic OCT image generation, it is realistically challenging to gather sufficient pre- and post-therapeutic OCT image pairs, resulting in overfitting. Moreover, the available GAN-based methods ignore local details, such as the biomarkers that are essential for nAMD treatment. To address these issues, a Biomarkers-aware Asymmetric Bibranch GAN (BAABGAN) is proposed to efficiently generate post-therapeutic OCT images. Specifically, one branch is developed to learn prior knowledge with a high degree of transferability from large-scale data, termed the source branch. Then, the source branch transfer knowledge to another branch, which is trained on small-scale paired data, termed the target branch. To boost the transferability, a novel Adaptive Memory Batch Normalization (AMBN) is introduced in the source branch, which learns more effective global knowledge that is impervious to noise via memory mechanism. Also, a novel Adaptive Biomarkers-aware Attention (ABA) module is proposed to encode biomarkers information into latent features of target branches to learn finer local details of biomarkers. The proposed method outperforms traditional GAN models and can produce high-quality post-treatment OCT pictures with limited data sets, as shown by the results of experiments.

Adaptive Memory Broad Learning System for Unsupervised Time Series Anomaly Detection.

Time series anomaly detection is the process of identifying anomalies within time series data. The primary challenge of this task lies in the necessity for the model to comprehend the characteristics of time-independent and abnormal data patterns. In this study, a novel algorithm called adaptive memory broad learning system (AdaMemBLS) is proposed for time series anomaly detection. This algorithm leverages the rapid inference capabilities of the broad learning algorithm and the memory bank's capacity to differentiate between normal and abnormal data. Furthermore, an incremental algorithm based on multiple data augmentation techniques is introduced and applied to multiple ensemble learners, thereby enhancing the model's effectiveness in learning the characteristics of time series data. To bolster the model's anomaly detection capabilities, a more diverse ensemble approach and a discriminative anomaly score are recommended. Extensive experiments conducted on various real-world datasets demonstrate that the proposed method exhibits superior inference speed and more accurate anomaly detection compared to the existing competitors. A detailed experimental investigation is presented to elucidate the effectiveness of the proposed method and the underlying reasons for its efficacy.

Adaptive Memory Saving Strategy in Shielding Calculation of RMC

RMC(Reactor Monte Carlo) is a Monte Carlo simulation software used in reactor design and has been verified by different benchmarks. The former data structure and algorithm have been optimized for criticality and burn-up calculation making it inconvenient to be implemented into shielding simulation and variance reduction. In order to expand its availability in shielding calculation, several optimizations have been made in RMC code design to elevate efficiency in variance reduction. Considering parallel computation has been implemented in Monte Carlo simulation for long time, in this paper we proposed an adaptive memory saving strategy specifically designed for boosting parallel variance reduction calculation in RMC. We further found that this strategy can be very effective in certain condition without sacrificing any calculation performance.

An adaptive long memory conditional correlation model

We propose a conditional correlation model with long memory dependence and smooth structural change. Previous literature has considered correlation and covariance models with structural change or long memory, but this is the first paper to jointly model both features. The correlation matrix is decomposed into long and short run components. Short run correlations converge hypergeometrically towards a slow moving long run correlation matrix that evolves according to one or more flexible Fourier forms. The model is applied to two data sets: a US equity portfolio; and a US equity, bond, gold and oil portfolio. Model fit and out of sample forecasts over 1 to 60 day horizons support the proposed approach.

Mammalian Target of Rapamycin Complex 1 Activation in Macrophages Contributes to Persistent Lung Inflammation following Respiratory Tract Viral Infection

Respiratory virus infections cause millions of hospitalizations worldwide each year. Severe infections lead to lung damage that coincides with persistent inflammation and a lengthy repair period. Vaccination and anti-viral therapy help to mitigate severe infections prior to or during the acute stage of disease, but there are currently limited specific treatment options available to individuals suffering from the long-term sequelae of respiratory viral infection. C57BL/6 mice were infected with influenza A/PR/8/34 as a model for severe viral lung infection and allowed to recover for 21 days. Mice were treated with rapamycin, a well characterized mTORC1 inhibitor on days 12 to 20 post infection; a time period after viral clearance. Persistent inflammation following severe influenza infection in mice was primarily driven by macrophages and T cells. UMAP analysis of flow cytometry data revealed that lung macrophages had high activation of mTORC1, an energy sensing kinase involved in inflammatory immune cell effector functions. Rapamycin treatment reduced lung inflammation and the frequency of exudate macrophages, T cells, and B cells in the lung, while not impacting epithelial progenitor cells or adaptive immune memory. These data highlight mTORC1’s role in sustaining persistent inflammation following clearance of a viral respiratory pathogen and suggest a possible intervention for post-viral chronic lung inflammation.

Multi-Defect Detection Network for High-Voltage Insulators Based on Adaptive Multi-Attention Fusion

Insulators find extensive use across diverse facets of power systems, playing a pivotal role in ensuring the security and stability of electrical transmission. Detecting insulators is a fundamental measure to secure the safety and stability of power transmission, with precise insulator positioning being a prerequisite for successful detection. To overcome challenges such as intricate insulator backgrounds, small defect scales, and notable differences in target scales that reduce detection accuracy, we propose the AC-YOLO insulator multi-defect detection network based on adaptive attention fusion. To elaborate, we introduce an adaptive weight distribution multi-head self-attention module designed to concentrate on intricacies in the features, effectively discerning between insulators and various defects. Additionally, an adaptive memory fusion detection head is incorporated to amalgamate multi-scale target features, augmenting the network’s capability to extract insulator defect characteristics. Furthermore, a CBAM attention mechanism is integrated into the backbone network to enhance the detection performance for smaller target defects. Lastly, improvements to the loss function expedite model convergence. This study involved training and evaluation using publicly available datasets for insulator defects. The experimental results reveal that the AC-YOLO model achieves a notable 5.1% enhancement in detection accuracy compared to the baseline. This approach significantly boosts detection precision, diminishes false positive rates, and fulfills real-time insulator localization requirements in power system inspections.

Reward conditioning may not have an effect on category-specific memory

Behavioural tagging facilitates the temporary storage of seemingly insignificant episodic events, which may later become salient and enhanced in memory. Human behavioural studies have demonstrated selective memory enhancement for neutral stimuli from one category when this category is subsequently paired with reward. Although this phenomenon has implications for the role of reward conditioning on emotional and adaptive memory, its generalisability is underexplored. We conducted four experiments to investigate whether pairing items from a semantic category, animals or objects, with high or low rewards resulted in preferential memory for the high-reward category. Three of these experiments also aimed to replicate the category-specific retrospective enhancement effect reported by Patil et al. and two explored the corresponding prospective memory effect. None of our experiments showed consistent evidence for an effect of reward on category-specific memory enhancement, despite employing the same reward paradigm and incidental encoding protocol as in the original study. Consequently, we found no evidence for category-specific retrospective or prospective enhancement effects. Our experiments were conducted online which is an equally relevant method for assessing behavioural phenomenon as the in-person studies conducted by Patil et al. Overall, our results question the generalisability of previously reported category-specific memory enhancement effects due to reward.

Nanovaccines for Advancing Long-Lasting Immunity against Infectious Diseases.

Infectious diseases, particularly life-threatening pathogens such as small pox and influenza, have substantial implications on public health and global economies. Vaccination is a key approach to combat existing and emerging pathogens. Immunological memory is an essential characteristic used to evaluate vaccine efficacy and durability and the basis for the long-term effects of vaccines in protecting against future infections; however, optimizing the potency, improving the quality, and enhancing the durability of immune responses remains challenging and a focus for research involving investigation of nanovaccine technologies. In this review, we describe how nanovaccines can address the challenges for conventional vaccines in stimulating adaptive immune memory responses to protect against reinfection. We discuss protein and nonprotein nanoparticles as useful antigen platforms, including those with highly ordered and repetitive antigen array presentation to enhance immunogenicity through cross-linking with multiple B cell receptors, and with a focus on antigen properties. In addition, we describe how nanoadjuvants can improve immune responses by providing enhanced access to lymph nodes, lymphnode targeting, germinal center retention, and long-lasting immune response generation. Nanotechnology has the advantage to facilitate vaccine induction of long-lasting immunity against infectious diseases, now and in the future.

Adaptive Memory Networks With Self-Supervised Learning for Unsupervised Anomaly Detection

Unsupervised anomaly detection aims to build models to effectively detect unseen anomalies by only training on the normal data. Although previous reconstruction-based methods have made fruitful progress, their generalization ability is limited due to two critical challenges. First, the training dataset only contains normal patterns, which limits the model generalization ability. Second, the feature representations learned by existing models often lack representativeness which hampers the ability to preserve the diversity of normal patterns. In this paper, we propose a novel approach called Adaptive Memory Network with Self-supervised Learning (AMSL) to address these challenges and enhance the generalization ability in unsupervised anomaly detection. Based on the convolutional autoencoder structure, AMSL incorporates a self-supervised learning module to learn general normal patterns and an adaptive memory fusion module to learn rich feature representations. Experiments on four public multivariate time series datasets demonstrate that AMSL significantly improves the performance compared to other state-of-the-art methods. Specifically, on the largest CAP sleep stage detection dataset with 900 million samples, AMSL outperforms the second-best baseline by 4%+ in both accuracy and F1 score. Apart from the enhanced generalization ability, AMSL is also more robust against input noise.

Abstract A070: Olaparib enhances radiation-induced Type I interferon and sensitizes pancreatic cancer to PD-L1 immune checkpoint inhibition

Abstract Inhibitors of PARP have been shown to sensitize pancreatic ductal adenocarcinoma (PDAC) to radiation therapy through mechanisms involving DNA damage, replication stress, and PARP trapping independent of BRCA1/2 mutation status. These mechanisms have the potential to activate innate immune signaling leading to Type I Interferon (T1IFN) production. Thus, we hypothesized that the clinical PARP1/2 inhibitor olaparib would potentiate radiation-induced T1IFN to activate an anti-tumoral immune response and sensitize resistant PDAC to immunotherapy. To test this hypothesis, we assessed the effects of olaparib and radiation on T1IFN production and signaling, tumor growth, sensitivity to PD-L1 immunotherapy, and the tumor immune microenvironment in mouse models of PDAC. We found that clinically achievable concentrations of olaparib significantly enhanced radiation-induced T1IFN production measured by both an IFNβ promoter-driven reporter and Ifnβ mRNA levels. Additionally, we found that the downstream effects of radiation-induced T1IFN signaling, including PD-L1 and MHC-I cell surface expression and induction of interferon stimulated genes including Cxcl9 and Cxcl10, were enhanced by olaparib. Co-treatment with a T1IFN receptor blocking antibody diminished these effects supporting the requirement for T1IFN-dependent signaling for the innate immune effects of radiation and olaparib. In vivo treatment with olaparib and radiation sensitized tumors to anti-PD-L1 therapy with a 40% complete response rate. These therapeutic responses led to an adaptive long term immunogenic memory evidenced by reduced engraftment following tumor rechallenge. Single-cell RNA sequencing revealed that treatment with olaparib, radiation and anti-PD-L1 augmented an anti-tumoral CD8+ T cell immune response, which is supported by an increase in CD8+ effector T cells and a decrease in CD8+ dysfunctional T cells. The necessity of functional CD8+ T cells to therapy efficacy in vivo was further reflected by a significantly reduced therapeutic combinational efficacy using a CD8+ T cell blocking antibody. Taken together, the results of this study show that the combination of olaparib, radiation and anti-PD-L1 induce an anti-tumoral immune response mediated by both T1IFN and CD8+ T cells. The results of this study provide the rationale for a phase 1 clinical trial (NCT05411094) combining dose escalated olaparib with durvalumab and radiotherapy in patients with locally advanced pancreatic cancer. Citation Format: Victoria M Valvo, Qiang Zhang, Long Jiang, Ashkay Tate, Michael D Green, Meredith A Morgan. Olaparib enhances radiation-induced Type I interferon and sensitizes pancreatic cancer to PD-L1 immune checkpoint inhibition [abstract]. In: Proceedings of the AACR-NCI-EORTC Virtual International Conference on Molecular Targets and Cancer Therapeutics; 2023 Oct 11-15; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2023;22(12 Suppl):Abstract nr A070.

The relationship of intraindividual variability on mobile cognitive testing with age and frontotemporal dementia disease severity

AbstractBackgroundThere has been a rise in remote cognitive testing for neurodegenerative diseases. The ALLFTD Mobile App is designed to capture clinical features of frontotemporal dementia (FTD) through repeated remote assessment on smartphones. This is a departure from traditional neuropsychological practice in which a single examiner‐led assessment occurs in a controlled setting. When participants complete tasks on their own, those with greater disease severity may be more variable in their performance for a variety of reasons (e.g. greater distractibility, less engagement). Here we study whether intraindividual cognitive variability (IIV) is associated with older age and increasing disease severity.MethodParticipants were 279 (mean age = 54.9) healthy controls and participants with sporadic FTD or from familial FTD kindreds with a range of disease severities based on FTLD Clinical Dementia Rating scale (CDR®+NACC‐FTLD Global) scores: 0 (n = 132), 0.5 (n = 54), and ≥1 (n = 47). Participants completed three cognitive tasks at least two times across twelve days. Measures included an adaptive memory task and gamified versions of Flanker and Go/no‐go tasks. IIV was operationalized as the standard deviation in performance on each task. Regression analyses investigated the relationship of IIV with CDR®+NACC‐FTLD group, and with age in those without symptoms. Education and gender were included as covariates in all analyses.ResultGreater disease severity (CDR®+NACC‐FTLD) was associated with greater variability on Flanker (p<.001) and Go/No‐go (p = .02) tasks; a positive but non‐significant association was seen with memory testing (p = .16). Post‐hoc comparisons found that those with CDR®+NACC‐FTLD of 0.5 had significantly more IIV on the Flanker task (𝛃 = .39, p = .004) than those with CDR®+NACC‐FTLD’s of 0. Moreover, those with a CDR®+NACC‐FTLD ≥1 had significantly greater IIV on the Flanker (𝛃 = .85, p<.001) and Go/No‐go tasks (𝛃 = .51, p = .007) than those with a CDR®+NACC‐FTLD of 0. In those without symptoms (CDR®+NACC‐FTLD = 0), age was associated with increased IIV on the Flanker (𝛃 = .27, p<.001) and Go/no‐go (𝛃 = .23, p<.001) tasks.ConclusionIncreased IIV, particularly on processing speed/executive functioning tasks, was associated with greater age and greater functional impairment, even in the mildest stages. Future investigations exploring predictors of IIV and the prognostic value of IIV are warranted.

Current Understanding of Bacillus Calmette-Guérin-Mediated Trained Immunity and Its Perspectives for Controlling Intracellular Infections.

The bacillus Calmette-Guérin (BCG) is an attenuated bacterium derived from virulent Mycobacterium bovis. It is the only licensed vaccine used for preventing severe forms of tuberculosis in children. Besides its specific effects against tuberculosis, BCG administration is also associated with beneficial non-specific effects (NSEs) following heterologous stimuli in humans and mice. The NSEs from BCG could be related to both adaptive and innate immune responses. The latter is also known as trained immunity (TI), a recently described biological feature of innate cells that enables functional improvement based on metabolic and epigenetic reprogramming. Currently, the mechanisms related to BCG-mediated TI are the focus of intense research, but many gaps are still in need of elucidation. This review discusses the present understanding of TI induced by BCG, exploring signaling pathways that are crucial to a trained phenotype in hematopoietic stem cells and monocytes/macrophages lineage. It focuses on BCG-mediated TI mechanisms, including the metabolic-epigenetic axis and the inflammasome pathway in these cells against intracellular pathogens. Moreover, this study explores the TI in different immune cell types, its ability to protect against various intracellular infections, and the integration of trained innate memory with adaptive memory to shape next-generation vaccines.

Do normative data specific to Greek Australian older adults improve validity of neuropsychological assessment results?

This study aimed to compare Greek Australian and English language normative data with regard to impairment rates yielded within a healthy Greek Australian older adult sample. We also examined whether optimal cut scores could be identified and capable of sensitively and specifically distinguishing between healthy Greek Australians from those with a diagnosis of Alzheimer's disease (AD). Ninety healthy Greek Australian older adults and 20 demographically matched individuals with a diagnosis of AD completed a range of neuropsychological measures, including the Wechsler Adult Intelligence Scale-Fourth Edition, Greek Adaptation (WAIS-IV GR), verbal and visual memory, language and naming, and executive functions. Impairment rates derived from the use of either Greek Australian or English language normative data were calculated and compared, using a 1.5 standard deviation criterion to denote impairment. Receiver operating characteristics curve analysis was used to investigate the sensitivity and specificity of alternate cut scores. Impairment rates derived from the Greek Australian normative data showed that rates of impairment generally fell within the expected 7% range. In contrast, impairment rates for all tests derived using English language normative data were significantly higher and ranged from 11%-66%. Comparisons between healthy and AD participants with moderate dementia showed significant differences across all measures. Area under the curve results ranged from .721 to .999 across all measures, with most tests displaying excellent sensitivity and specificity. English language normative data were found to be inappropriate for use with Greek Australian elders, potentially leading to erroneous diagnostic outcomes. The use of minority group specific normative data and associated cut points appear to partially ameliorate this issue. Clinical implications are discussed alongside future research directions.

4 Does Working Memory Training for Children Need to be Adaptive? A Randomised Controlled Trial

Objective:A common assumption to maximise cognitive training outcomes is that training tasks should be adaptive, with difficulty adjusted to the individual’s performance. This has only been tested once in adults (von Bastian & Eschen, 2016). We aimed to examine children’s outcomes of working memory training using adaptive, self-select and stepwise approaches to setting the difficulty of training tasks compared to an active control condition.Participants and Methods:In a randomised controlled trial (ACTRN 12621000990820), children in Grades 2-5 (7 to 11 years) were allocated to one of four conditions: adaptive working memory training, self-select working memory training, stepwise working memory training, or active control. An experimental intervention embedded in Minecraft was developed for teachers to deliver in the classroom over two weeks (10 x 20-minute sessions). The working memory training comprised two training tasks with processing demands similar to daily activities: backward span with digits and following instructions with objects. The control condition comprised creative building tasks. As part of a larger protocol, children completed at baseline and immediately post-intervention working memory measures similar to the training activities (primary outcome): backward span digits and letters versions, following instructions objects and letters versions. Primary analyses were intention-to-treat. Secondary analyses included only children who completed 10 sessions.Results:Of 204 children recruited into the study, 203 were randomised, with 95% retention at post-intervention. 76% of children completed all 10 training sessions. Comparisons between each working memory training condition and the active control on working memory measures were non-significant (f2 = 0.00), with one exception. Children in the self-select condition on average performed 1-point better than the controls on the following instructions objects measure (p = .02, f2 = 0.03). A pattern emerged that the self-select condition performed better on most measures.Conclusions:We found little evidence that an adaptive approach to setting the difficulty of training tasks maximises training outcomes for children. Findings suggest that working memory outcomes following training are limited and are not modulated by the approach to setting the difficulty of training tasks. This is consistent with findings from von Bastian & Eschen (2016), who also observed that the self-select condition (and not the adaptive condition) showed a slightly larger change in working memory performance following training than the control. It is helpful for clinicians to be aware that adaptive working memory training programs might not be superior in improving children’s working memory, and the benefits of programs are limited.

3 Stricker Learning Span criterion validity: remote self-administration of a computer adaptive word list memory test shows similar ability to differentiate PET-defined biomarker groups as in-person Rey Auditory Verbal Learning Test performance in cognitively unimpaired individuals on the Alzheimer’s continuum

Objective:The Stricker Learning Span (SLS) is a computer-adaptive word list memory test specifically designed for remote assessment and self-administration on a web-based multi-device platform (Mayo Test Drive). Given recent evidence suggesting the prominence of learning impairment in preclinical Alzheimer’s disease (AD), the SLS places greater emphasis on learning than delayed memory compared to traditional word list memory tests (see Stricker et al., Neuropsychology in press for review and test details). The primary study aim was to establish criterion validity of the SLS by comparing the ability of the remotely-administered SLS and inperson administered Rey Auditory Verbal Learning Test (AVLT) to differentiate biomarkerdefined groups in cognitively unimpaired (CU) individuals on the Alzheimer’s continuum.Participants and Methods:Mayo Clinic Study of Aging CU participants (N=319; mean age=71, SD=11; mean education=16, SD=2; 47% female) completed a brief remote cognitive assessment (∼0.5 months from in-person visit). Brain amyloid and brain tau PET scans were available within 3 years. Overlapping groups were formed for 1) those on the Alzheimer’s disease (AD) continuum (A+, n=110) or not (A-, n=209), and for 2) those with biological AD (A+T+, n=43) vs no evidence of AD pathology (A-T-, n=181). Primary neuropsychological outcome variables were sum of trials for both the SLS and AVLT. Secondary outcome variables examined comparability of learning (1-5 total) and delay performances. Linear model ANOVAs were used to investigate biomarker subgroup differences and Hedge’s G effect sizes were derived, with and without adjusting for demographic variables (age, education, sex).Results:Both SLS and AVLT performances were worse in the biomarker positive relative to biomarker negative groups (unadjusted p’s<.05). Because biomarker positive groups were significantly older than biomarker negative groups, group differences were attenuated after adjusting for demographic variables, but SLS remained significant for A+ vs A- and for A+T+ vs A-T- comparisons (adjusted p’s<.05) and AVLT approached significance (p’s .05-.10). The effect sizes for the SLS were slightly better (qualitatively, no statistical comparison) for separating biomarker-defined CU groups in comparison to AVLT. For A+ vs A- and A+T+ vs A-T- comparisons, unadjusted effect sizes for SLS were -0.53 and -0.81 and for AVLT were -0.47 and -0.61, respectively; adjusted effect sizes for SLS were -0.25 and -0.42 and for AVLT were -0.19 and -0.26, respectively. In secondary analyses, learning and delay variables were similar in terms of ability to separate biomarker groups. For example, unadjusted effect sizes for SLS learning (-.80) was similar to SLS delay (.76), and AVLT learning (-.58) was similar to AVLT 30-minute delay (-.55) for the A+T+ vs AT- comparison.Conclusions:Remotely administered SLS performed similarly to the in-person-administered AVLT in its ability to separate biomarker-defined groups in CU individuals, providing evidence of criterion validity. The SLS showed significantly worse performance in A+ and A+T+ groups (relative to A- and A-T-groups) in this CU sample after demographic adjustment, suggesting potential sensitivity to detecting transitional cognitive decline in preclinical AD. Measures emphasizing learning should be given equal consideration as measures of delayed memory in AD-focused studies, particularly in the preclinical phase.

6 Remote Smartphone Cognitive and Motor Testing in Frontotemporal Dementia Research: Feasibility, Reliability, and Validity

Objective:Therapeutics targeting frontotemporal dementia (FTD) are entering clinical trials. There are challenges to conducting these studies, including the relative rarity of the disease. Remote assessment tools could increase access to clinical research and pave the way for decentralized clinical trials. We developed the ALLFTD Mobile App, a smartphone application that includes assessments of cognition, speech/language, and motor functioning. The objectives were to determine the feasibility and acceptability of collecting remote smartphone data in a multicenter FTD research study and evaluate the reliability and validity of the smartphone cognitive and motor measures.Participants and Methods:A diagnostically mixed sample of 207 participants with FTD or from familial FTD kindreds (CDR®+NACC-FTLD=0 [n=91]; CDR®+NACC-FTLD=0.5 [n=39]; CDR®+NACC-FTLD>1 [n=39]; unknown [n=38]) were asked to remotely complete a battery of tests on their smartphones three times over two weeks. Measures included five executive functioning (EF) tests, an adaptive memory test, and participant experience surveys. A subset completed smartphone tests of balance at home (n=31) and a finger tapping test (FTT) in the clinic (n=11). We analyzed adherence (percentage of available measures that were completed) and user experience. We evaluated Spearman-Brown split-half reliability (100 iterations) using the first available assessment for each participant. We assessed test-retest reliability across all available assessments by estimating intraclass correlation coefficients (ICC). To investigate construct validity, we fit regression models testing the association of the smartphone measures with gold-standard neuropsychological outcomes (UDS3-EF composite [Staffaroni et al., 2021], CVLT3-Brief Form [CVLT3-BF] Immediate Recall, mechanical FTT), measures of disease severity (CDR®+NACC-FTLD Box Score & Progressive Supranuclear Palsy Rating Scale [PSPRS]), and regional gray matter volumes (cognitive tests only).Results:Participants completed 70% of tasks. Most reported that the instructions were understandable (93%), considered the time commitment acceptable (97%), and were willing to complete additional assessments (98%). Split-half reliability was excellent for the executive functioning (r’s=0.93-0.99) and good for the memory test (r=0.78). Test-retest reliabilities ranged from acceptable to excellent for cognitive tasks (ICC: 0.70-0.96) and were excellent for the balance (ICC=0.97) and good for FTT (ICC=0.89). Smartphone EF measures were strongly associated with the UDS3-EF composite (ß's=0.6-0.8, all p<.001), and the memory test was strongly correlated with total immediate recall on the CVLT3-BF (ß=0.7, p<.001). Smartphone FTT was associated with mechanical FTT (ß=0.9, p=.02), and greater acceleration on the balance test was associated with more motor features (ß=0.6, p=0.02). Worse performance on all cognitive tests was associated with greater disease severity (ß's=0.5-0.7, all p<.001). Poorer performance on the smartphone EF tasks was associated with smaller frontoparietal/subcortical volume (ß's=0.4-0.6, all p<.015) and worse memory scores with smaller hippocampal volume (ß=0.5, p<.001).Conclusions:These results suggest remote digital data collection of cognitive and motor functioning in FTD research is feasible and acceptable. These findings also support the reliability and validity of unsupervised ALLFTD Mobile App cognitive tests and provide preliminary support for the motor measures, although further study in larger samples is required.