Loading
Dementia and Cognitive Impairment Research · Journal article
Discover Applied Sciences · August 6, 2026
Early or partial results. Treat as a signal, not a conclusion.
NeuroCognex is a multimodal deep-learning system combining fMRI and EEG that reports an AUC of 0.95 and balanced accuracy of 93.75% on the PEARL Neuro public dataset. The work is a technical proof-of-concept with promising algorithmic performance, but lacks prospective clinical validation, a clinical comparator, and independent external testing necessary to establish diagnostic utility or clinical practice relevance.
Single-arm validation study on public research dataset. PEARL Neuro dataset (EEG and fMRI recordings); specific patient cohort characteristics not described. Intervention: NeuroCognex multimodal deep-learning system using dual-stream architecture with cross-modal attention mechanisms to integrate fMRI and EEG data for Alzheimer's disease prediction.
AUC of 0.95 reported on PEARL Neuro dataset Balanced accuracy of 93.75% achieved Error rate of 6.5% reported
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
This is a computational model development report, not a clinical study. Clinicians should not use these metrics to inform diagnostic practice; prospective validation against clinical reference standards in a real patient population would be necessary before any clinical application.
Single-arm validation study on a public dataset with no control arm or comparative benchmark; demonstrates technical performance but lacks clinical validation, prospective testing, or comparison to standard diagnostic methods.
As stated by the source record.
Quoted from the source exactly as published.
This is a computational model development report, not a clinical study. Clinicians should not use these metrics to inform diagnostic practice; prospective validation against clinical reference standards in a real patient population would be necessary before any clinical application.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
Alzheimer’s disease is a progressive neurological condition marked by memory loss and cognitive decline that necessitates precise and timely diagnosis. This proposed method presents NeuroCognex (NCX), a system that uses functional magnetic resonance imaging (fMRI) and electroencephalography (EEG) to diagnose Alzheimer’s disease early by combining multimodal data processing and deep learning. PEARL Neuro dataset, a publicly available dataset consisting of EEG and fMRI recordings was used for training and validating NCX. While fMRI-derived statistical representations provide compact summaries of brain activity, the framework uses sophisticated EEG signal processing techniques, such as band-specific spectrum decomposition and power spectral density estimates, to capture spectral characteristics of EEG signals. By using cross-modal attention mechanisms to integrate these diverse signals, a dual-stream design effectively fuses haemodynamic and electrophysiological data. The system uses signal augmentation, normalisation, and class-weighted optimisation techniques to tackle issues including noise, high dimensionality, and small dataset sizes. With an AUC of 0.95 and balanced accuracy of 93.75%, the experimental results show good predictive performance while preserving physiological interpretability consistent with established brain biomarkers. Interestingly, the framework achieves a low error rate of 6.5%. Additionally, by producing clinically significant reports from intricate signal representations, the incorporation of generative AI improves interpretability. By facilitating the early detection of neurodegenerative diseases, this work advances SDG 3: Good Health and Well-Being; supports SDG 9: Industry, Innovation, and Infrastructure by creating sophisticated signal processing-driven healthcare systems; and advances SDG 10: Reduced Inequalities by making diagnostic technologies accessible and affordable.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.