Foundation Model Research Articles

Novel artificial intelligence for diabetic retinopathy and diabetic macular edema: what is new in 2024?

Given the increasing global burden of diabetic retinopathy and the rapid advancements in artificial intelligence, this review aims to summarize the current state of artificial intelligence technology in diabetic retinopathy detection and management, assessing its potential to improve care and visual outcomes in real-world settings. Most recent studies focused on the integration of artificial intelligence in the field of diabetic retinopathy screening, focusing on real-world efficacy and clinical implementation of such artificial intelligence models. Additionally, artificial intelligence holds the potential to predict diabetic retinopathy progression, enhance personalized treatment strategies, and identify systemic disease biomarkers from ocular images through 'oculomics', moving towards a more precise, efficient, and accessible care. The emergence of foundation model architectures and generative artificial intelligence, which more clearly reflect the clinical care process, may enable rapid advances in diabetic retinopathy care, research and medical education. This review explores the emerging technology of artificial intelligence to assess the potential to improve patient outcomes and optimize personalized management in healthcare delivery and medical research. While artificial intelligence is expected to play an increasingly important role in diabetic retinopathy care, ongoing research and clinical trials are essential to address implementation issues and focus on long-term patient outcomes for successful real-world adoption of artificial intelligence in diabetic retinopathy.

Reducing Training Data Using Pre-Trained Foundation Models: A Case Study on Traffic Sign Segmentation Using the Segment Anything Model.

The utilization of robust, pre-trained foundation models enables simple adaptation to specific ongoing tasks. In particular, the recently developed Segment Anything Model (SAM) has demonstrated impressive results in the context of semantic segmentation. Recognizing that data collection is generally time-consuming and costly, this research aims to determine whether the use of these foundation models can reduce the need for training data. To assess the models' behavior under conditions of reduced training data, five test datasets for semantic segmentation will be utilized. This study will concentrate on traffic sign segmentation to analyze the results in comparison to Mask R-CNN: the field's leading model. The findings indicate that SAM does not surpass the leading model for this specific task, regardless of the quantity of training data. Nevertheless, a knowledge-distilled student architecture derived from SAM exhibits no reduction in accuracy when trained on data that have been reduced by 95%.

Ukrainian adaptation of the Clance Impostor Phenomenon Scale (CIPS)

Expanding the methodological toolkit of research, particularly through adapting foreign psychometric tools, is crucial for Ukrainian psychology. This allows for conducting new research and comparing its results. One promising direction is studying the impostor phenomenon, manifested in an individual’s doubts about their achievements. The Clance Impostor Phenomenon Scale, developed by foreign researchers, is used for this purpose. This study aimed to adapt the Clance Impostor Phenomenon Scale into Ukrainian. The Clance Impostor Phenomenon Scale was translated into Ukrainian by two psychologists using the back-translation method. The research sample consisted of 297 students aged 18 to 22 (M = 20.4, SD = 1.49), including 82 males and 215 females. Descriptive statistics, reliability analysis of scales, correlation analysis (Spearman’s coefficient), and confirmatory factor analysis were used. The results of the study indicate a considerable degree of consistency between the obtained data and the foundational theoretical model. The assessment of content validity revealed a significant number of statistically significant correlations. A high level of test-retest reliability was identified for the adapted methodology. The resulting factor model is conceptually similar to the original; however, it is not identical. Other studies have also noted the emergence of new factor models, suggesting the potential for refinement of this methodology or reflecting age, temporal, and cross-cultural differences among the subjects studied. The evaluation of the psychometric characteristics of the Ukrainian version of the Clance Impostor Phenomenon Scale demonstrated high internal validity and reliability for the adapted tool. The Ukrainian version of the Clance Impostor Phenomenon Scale may be valuable for the psychodiagnosis of impostor syndrome manifestations and further standardisation and application in comparative research. The findings could contribute to optimising psychological support by taking into account the specific characteristics of impostor syndrome in the course of psychological work

Applications of Artificial Intelligence and Machine Learning in Spine MRI.

Diagnostic imaging, particularly MRI, plays a key role in the evaluation of many spine pathologies. Recent progress in artificial intelligence and its subset, machine learning, has led to many applications within spine MRI, which we sought to examine in this review. A literature search of the major databases (PubMed, MEDLINE, Web of Science, ClinicalTrials.gov) was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The search yielded 1226 results, of which 50 studies were selected for inclusion. Key data from these studies were extracted. Studies were categorized thematically into the following: Image Acquisition and Processing, Segmentation, Diagnosis and Treatment Planning, and Patient Selection and Prognostication. Gaps in the literature and the proposed areas of future research are discussed. Current research demonstrates the ability of artificial intelligence to improve various aspects of this field, from image acquisition to analysis and clinical care. We also acknowledge the limitations of current technology. Future work will require collaborative efforts in order to fully exploit new technologies while addressing the practical challenges of generalizability and implementation. In particular, the use of foundation models and large-language models in spine MRI is a promising area, warranting further research. Studies assessing model performance in real-world clinical settings will also help uncover unintended consequences and maximize the benefits for patient care.

Development of a basic evaluation model for manual therapy learning in rehabilitation students based on the Delphi method

ObjectiveManual therapy is a crucial component in rehabilitation education, yet there is a lack of models for evaluating learning in this area. This study aims to develop a foundational evaluation model for manual therapy learning among rehabilitation students, based on the Delphi method, and to analyze the theoretical basis and practical significance of this model.MethodsAn initial framework for evaluating the fundamentals of manual therapy learning was constructed through a literature review and theoretical analysis. Using the Delphi method, consultations were conducted with young experts in the field of rehabilitation from January 2024 to March 2024. Fifteen experts completed three rounds of consultation. Each round involved analysis using Dview software, refining and adjusting indicators based on expert opinions, and finally summarizing all retained indicators using Mindmaster.ResultsThe effective response rates for the three rounds of questionnaires were 88%, 100%, and 100%, respectively. Expert familiarity scores were 0.91, 0.95, and 0.95; coefficient of judgment were 0.92, 0.93, and 0.93; authority coefficients were 0.92, 0.94, and 0.94, respectively. Based on three rounds of consultation, the model established includes 3 primary indicators, 10 secondary indicators, 17 tertiary indicators, and 9 quaternary indicators. A total of 24 statistical indicators were finalized, with 8 under the Cognitive Abilities category, 10 under the Practical Skills category, and 6 under the Emotional Competence category.ConclusionThis study has developed an evaluation model for manual therapy learning among rehabilitation students, based on the Delphi method. The model includes multi-level evaluation indicators covering the key dimensions of Cognitive Abilities, Practical Skills, and Emotional Competence. These indicators provide a preliminary evaluation framework for manual therapy education and a theoretical basis for future research.

A Common Longitudinal Intensive Care Unit data Format (CLIF) to enable multi-institutional federated critical illness research.

Critical illness, or acute organ failure requiring life support, threatens over five million American lives annually. Electronic health record (EHR) data are a source of granular information that could generate crucial insights into the nature and optimal treatment of critical illness. However, data management, security, and standardization are barriers to large-scale critical illness EHR studies. A consortium of critical care physicians and data scientists from eight US healthcare systems developed the Common Longitudinal Intensive Care Unit (ICU) data Format (CLIF), an open-source database format that harmonizes a minimum set of ICU Data Elements for use in critical illness research. We created a pipeline to process adult ICU EHR data at each site. After development and iteration, we conducted two proof-of-concept studies with a federated research architecture: 1) an external validation of an in-hospital mortality prediction model for critically ill patients and 2) an assessment of 72-hour temperature trajectories and their association with mechanical ventilation and in-hospital mortality using group-based trajectory models. We converted longitudinal data from 94,356 critically ill patients treated in 2020-2021 (mean age 60.6 years [standard deviation 17.2], 30% Black, 7% Hispanic, 45% female) across 8 health systems and 33 hospitals into the CLIF format, The in-hospital mortality prediction model performed well in the health system where it was derived (0.81 AUC, 0.06 Brier score). Performance across CLIF consortium sites varied (AUCs: 0.74-0.83, Brier scores: 0.06-0.01), and demonstrated some degradation in predictive capability. Temperature trajectories were similar across health systems. Hypothermic and hyperthermic-slow-resolver patients consistently had the highest mortality. CLIF facilitates efficient, rigorous, and reproducible critical care research. Our federated case studies showcase CLIF's potential for disease sub-phenotyping and clinical decision-support evaluation. Future applications include pragmatic EHR-based trials, target trial emulations, foundational multi-modal AI models of critical illness, and real-time critical care quality dashboards.

Foundation model-driven distributed learning for enhanced retinal age prediction.

The retinal age gap (RAG) is emerging as a potential biomarker for various diseases of the human body, yet its utility depends on machine learning models capable of accurately predicting biological retinal age from fundus images. However, training generalizable models is hindered by potential shortages of diverse training data. To overcome these obstacles, this work develops a novel and computationally efficient distributed learning framework for retinal age prediction. The proposed framework employs a memory-efficient 8-bit quantized version of RETFound, a cutting-edge foundation model for retinal image analysis, to extract features from fundus images. These features are then used to train an efficient linear regression head model for predicting retinal age. The framework explores federated learning (FL) as well as traveling model (TM) approaches for distributed training of the linear regression head. To evaluate this framework, we simulate a client network using fundus image data from the UK Biobank. Additionally, data from patients with type 1 diabetes from the UK Biobank and the Brazilian Multilabel Ophthalmological Dataset (BRSET) were utilized to explore the clinical utility of the developed methods. Our findings reveal that the developed distributed learning framework achieves retinal age prediction performance on par with centralized methods, with FL and TM providing similar performance (mean absolute error of 3.57 ± 0.18 years for centralized learning, 3.60 ± 0.16 years for TM, and 3.63 ± 0.19 years for FL). Notably, the TM was found to converge with fewer local updates than FL. Moreover, patients with type 1 diabetes exhibited significantly higher RAG values than healthy controls in all models, for both the UK Biobank and BRSET datasets (P < .001). The high computational and memory efficiency of the developed distributed learning framework makes it well suited for resource-constrained environments. The capacity of this framework to integrate data from underrepresented populations for training of retinal age prediction models could significantly enhance the accessibility of the RAG as an important disease biomarker.

Enhancing representation in radiography-reports foundation model: a granular alignment algorithm using masked contrastive learning

Recently, multi-modal vision-language foundation models have gained significant attention in the medical field. While these models offer great opportunities, they still face crucial challenges, such as the requirement for fine-grained knowledge understanding in computer-aided diagnosis and the capability of utilizing very limited or even no task-specific labeled data in real-world clinical applications. In this study, we present MaCo, a masked contrastive chest X-ray foundation model that tackles these challenges. MaCo explores masked contrastive learning to simultaneously achieve fine-grained image understanding and zero-shot learning for a variety of medical imaging tasks. It designs a correlation weighting mechanism to adjust the correlation between masked chest X-ray image patches and their corresponding reports, thereby enhancing the model’s representation learning capabilities. To evaluate the performance of MaCo, we conducted extensive experiments using 6 well-known open-source X-ray datasets. The experimental results demonstrate the superiority of MaCo over 10 state-of-the-art approaches across tasks such as classification, segmentation, detection, and phrase grounding. These findings highlight the significant potential of MaCo in advancing a wide range of medical image analysis tasks.

Towards practical artificial intelligence in Earth sciences

AbstractAlthough Artificial Intelligence (AI) projects are common and desired by many institutions and research teams, there are still relatively few success stories of AI in practical use for the Earth science community. Many AI practitioners in Earth science are trapped in the prototyping stage and their results have not yet been adopted by users. Many scientists are still hesitating to use AI in their research routine. This paper aims to capture the landscape of AI-powered geospatial data sciences by discussing the current and upcoming needs of the Earth and environmental community, such as what practical AI should look like, how to realize practical AI based on the current technical and data restrictions, and the expected outcome of AI projects and their long-term benefits and problems. This paper also discusses unavoidable changes in the near future concerning AI, such as the fast evolution of AI foundation models and AI laws, and how the Earth and environmental community should adapt to these changes. This paper provides an important reference to the geospatial data science community to adjust their research road maps, find best practices, boost the FAIRness (Findable, Accessible, Interoperable, and Reusable) aspects of AI research, and reasonably allocate human and computational resources to increase the practicality and efficiency of Earth AI research.

Small data methods in omics: the power of one.

Over the last decade, biology has begun utilizing 'big data' approaches, resulting in large, comprehensive atlases in modalities ranging from transcriptomics to neural connectomics. However, these approaches must be complemented and integrated with 'small data' approaches to efficiently utilize data from individual labs. Integration of smaller datasets with major reference atlases is critical to provide context to individual experiments, and approaches toward integration of large and small data have been a major focus in many fields in recent years. Here we discuss progress in integration of small data with consortium-sized atlases across multiple modalities, and its potential applications. We then examine promising future directions for utilizing the power of small data to maximize the information garnered from small-scale experiments. We envision that, in the near future, international consortia comprising many laboratories will work together to collaboratively build reference atlases and foundation models using small data methods.

Leveraging Bayesian methods for addressing multi-uncertainty in data-driven seismic liquefaction assessment

When assessing seismic liquefaction potential with data-driven models, addressing the uncertainties of establishing models, interpreting cone penetration tests (CPT) data and decision threshold is crucial for avoiding biased data selection, ameliorating overconfident models, and being flexible to varying practical objectives, especially when the training and testing data are not identically distributed. A workflow characterized by leveraging Bayesian methodology was proposed to address these issues. Employing a Multi-Layer Perceptron (MLP) as the foundational model, this approach was benchmarked against empirical methods and advanced algorithms for its efficacy in simplicity, accuracy, and resistance to overfitting. The analysis revealed that, while MLP models optimized via maximum a posteriori algorithm suffices for straightforward scenarios, Bayesian neural networks showed great potential for preventing overfitting. Additionally, integrating decision thresholds through various evaluative principles offers insights for challenging decisions. Two case studies demonstrate the framework's capacity for nuanced interpretation of in situ data, employing a model committee for a detailed evaluation of liquefaction potential via Monte Carlo simulations and basic statistics. Overall, the proposed step-by-step workflow for analyzing seismic liquefaction incorporates multifold testing and real-world data validation, showing improved robustness against overfitting and greater versatility in addressing practical challenges. This research contributes to the seismic liquefaction assessment field by providing a structured, adaptable methodology for accurate and reliable analysis.

Masked particle modeling on sets: towards self-supervised high energy physics foundation models

Abstract We propose masked particle modeling (MPM) as a self-supervised method for learning generic, transferable, and reusable representations on unordered sets of inputs for use in high energy physics (HEP) scientific data. This work provides a novel scheme to perform masked modeling based pre-training to learn permutation invariant functions on sets. More generally, this work provides a step towards building large foundation models for HEP that can be generically pre-trained with self-supervised learning and later fine-tuned for a variety of down-stream tasks. In MPM, particles in a set are masked and the training objective is to recover their identity, as defined by a discretized token representation of a pre-trained vector quantized variational autoencoder. We study the efficacy of the method in samples of high energy jets at collider physics experiments, including studies on the impact of discretization, permutation invariance, and ordering. We also study the fine-tuning capability of the model, showing that it can be adapted to tasks such as supervised and weakly supervised jet classification, and that the model can transfer efficiently with small fine-tuning data sets to new classes and new data domains.

Comparative Analysis of Portfolio Management Models: Markowitz, CAPM, and Multi-Factor Models

This paper presents a comparative analysis of three foundational portfolio management models: the Markowitz Efficient Frontier, the Capital Asset Pricing Model (CAPM), and Multi-Factor Models. It aims to explore their theoretical foundations, empirical applicability, and practical usage to assess their strengths and limitations in financial portfolio management. The Markowitz model is appreciated for its detailed risk-return optimization framework but criticized for its reliance on historical data and computational demands. The CAPM simplifies risk assessment through market beta, yet its assumptions of market efficiency and a risk-free rate are viewed as impractical under current market conditions. Multi-Factor Models address some limitations of the CAPM by incorporating various risk factors, offering enhanced explanatory power but also adding complexity in factor selection.The study concludes that while no single model is universally superior, the choice of model should depend on the specific needs of the portfolio manager, including risk-return objectives and market environment. The paper suggests a blended or adaptive approach to model selection, highlighting the need for flexibility in response to evolving market dynamics and advances in computational finance.

1942O Application of GigaPath: An open-weight billion-parameter AI foundation model based on a novel vision transformer architecture for cancer mutation prediction and TME analysis

1942O Application of GigaPath: An open-weight billion-parameter AI foundation model based on a novel vision transformer architecture for cancer mutation prediction and TME analysis

Foundation model for generalist remote sensing intelligence: potentials and prospects

Foundation model for generalist remote sensing intelligence: potentials and prospects

Adapting Segment Anything Model for 3D Brain Tumor Segmentation With Missing Modalities

ABSTRACTThe problem of missing or unavailable magnetic resonance imaging modalities challenges clinical diagnosis and medical image analysis technology. Although the development of deep learning and the proposal of large models have improved medical analytics, this problem still needs to be better resolved.The purpose of this study was to efficiently adapt the Segment Anything Model, a two‐dimensional visual foundation model trained on natural images, to address the challenge of brain tumor segmentation with missing modalities. We designed a twin network structure that processes missing and intact magnetic resonance imaging (MRI) modalities separately using shared parameters. It involved comparing the features of two network branches to minimize differences between the feature maps derived from them. We added a multimodal adapter before the image encoder and a spatial–depth adapter before the mask decoder to fine‐tune the Segment Anything Model for brain tumor segmentation. The proposed method was evaluated using datasets provided by the MICCAI BraTS2021 Challenge. In terms of accuracy and robustness, the proposed method is better than existing solutions. The proposed method can segment brain tumors well under the missing modality condition.

Large Multi-Modal Models (LMMs) as Universal Foundation Models for AI-Native Wireless Systems

Large Multi-Modal Models (LMMs) as Universal Foundation Models for AI-Native Wireless Systems

Foundation model of neural activity predicts response to new stimulus types and anatomy.

The complexity of neural circuits makes it challenging to decipher the brain's algorithms of intelligence. Recent breakthroughs in deep learning have produced models that accurately simulate brain activity, enhancing our understanding of the brain's computational objectives and neural coding. However, these models struggle to generalize beyond their training distribution, limiting their utility. The emergence of foundation models, trained on vast datasets, has introduced a new AI paradigm with remarkable generalization capabilities. We collected large amounts of neural activity from visual cortices of multiple mice and trained a foundation model to accurately predict neuronal responses to arbitrary natural videos. This model generalized to new mice with minimal training and successfully predicted responses across various new stimulus domains, such as coherent motion and noise patterns. It could also be adapted to new tasks beyond neural prediction, accurately predicting anatomical cell types, dendritic features, and neuronal connectivity within the MICrONS functional connectomics dataset. Our work is a crucial step toward building foundation brain models. As neuroscience accumulates larger, multi-modal datasets, foundation models will uncover statistical regularities, enabling rapid adaptation to new tasks and accelerating research.

Approximate analytical algorithm for pull-out resistance-displacement relationship of combined anchor system

Building upon an evaluation of the distinct advantages and limitations of anchor rods and anchorage plates, this paper introduces an integrated anchor-tensioning system that combines these elements in series. This innovative approach significantly enhances the pullout bearing capacity of the overall anchorage system. Based on the hyperbolic model of the distribution of lateral friction resistance of the anchor solid and the Winkler′s foundation model of load transfer of the anchorage plate, the theoretical analysis method for the analysis of the pullout bearing capacity and deformation of the combined anchorage system was established, and an approximate analytical equation for the relationship between pullout force and displacement was obtained. The results from indoor model tests, numerical simulations, and theoretical analyses were compared, verifying the reliability of the theoretical methods employed. The findings of this study significantly advance the discourse on the innovation of structural support configurations, thereby enhancing the operational efficiency of anchor-pull support systems.

Foundational Models for Pathology and Endoscopy Images: Application for Gastric Inflammation.

The integration of artificial intelligence (AI) in medical diagnostics represents a significant advancement in managing upper gastrointestinal (GI) cancer, which is a major cause of global cancer mortality. Specifically for gastric cancer (GC), chronic inflammation causes changes in the mucosa such as atrophy, intestinal metaplasia (IM), dysplasia, and ultimately cancer. Early detection through endoscopic regular surveillance is essential for better outcomes. Foundation models (FMs), which are machine or deep learning models trained on diverse data and applicable to broad use cases, offer a promising solution to enhance the accuracy of endoscopy and its subsequent pathology image analysis. This review explores the recent advancements, applications, and challenges associated with FMs in endoscopy and pathology imaging. We started by elucidating the core principles and architectures underlying these models, including their training methodologies and the pivotal role of large-scale data in developing their predictive capabilities. Moreover, this work discusses emerging trends and future research directions, emphasizing the integration of multimodal data, the development of more robust and equitable models, and the potential for real-time diagnostic support. This review aims to provide a roadmap for researchers and practitioners in navigating the complexities of incorporating FMs into clinical practice for the prevention/management of GC cases, thereby improving patient outcomes.