Traditional diagnostic methods in biomedicine are often constrained by subjectivity, misdiagnosis, and the inability to efficiently process large, complex, and multimodal datasets. Recent advances in deep learning and hybrid architectures have enabled the automated learning of discriminative representations from high‐dimensional biomedical data, supporting robust classification and integrative analysis across imaging, physiological signals, and clinical texts. This review synthesizes the current progress in applying deep learning to neuroimaging, cardiovascular disease diagnosis, functional connectivity analysis, and biomedical text mining, focusing on hybrid and ensemble strategies that combine complementary modeling strengths. These approaches show improved accuracy, scalability, and adaptability while extending situational awareness by incorporating patient‐generated and clinical textual data. Despite promising outcomes, challenges remain in data availability, computational efficiency, interpretability, and clinical integration. Future directions emphasize the development of explainable, multimodal, and generalizable frameworks capable of supporting precision medicine and advancing patient‐centered care.
Li-Feng Li, Ashfaque Khowaja, Yu-Cheng Song et al.· Medicine Advances· 0 citations
Major depressive disorder (MDD) is characterized by dysfunction in higher-order cortical regions involved in emotional and cognitive processes; however, its neurobiological basis remains unclear. Pharmacological treatments, including esketamine and sertraline, produce rapid antidepressant effects. We investigated the local intrinsic neural dynamics underlying these antidepressant effects using intrinsic neural timescales (INT), which measure the duration and capacity of information integration in localized brain regions.
A total of 57 healthy controls and 41 patients with MDD were included. All patients with MDD received six intravenous infusions of esketamine (0.25 mg/kg) and two weeks of sertraline treatment. All participants underwent resting-state fMRI to quantify INT alterations at the voxel and brain network levels, followed by spatial correlation analyses linking autocorrelation metrics of INT signals to transcriptomic data from the Allen Human Brain Atlas and PET-derived neurotransmitter receptor maps. The spatial correlation between PLS2 scores and case–control t-statistic maps did not pass rigorous spin permutation testing (
r
= 0.484,
p
= 0.068) and did not meet the conventional significance threshold of 0.05. Accordingly, all subsequent enrichment analyses based on PLS2 results are presented as exploratory preliminary observations for hypothesis generation.
Compared with healthy controls, patients with MDD exhibited significantly elevated INT levels in the left and right precuneus. Following treatment, patients with MDD showed significantly increased INT in the left occipital midline region and left calcarine cortex. At the brain network level, INT levels were significantly reduced in the default mode network (DMN) in patients with MDD compared with healthy controls. Compared with the pre-treatment state, the cerebellar network (CN) showed significantly elevated INT levels after treatment. Partial least squares regression analysis suggested potential associations between INT alterations and spatial gene expression gradients particularly those associated with immune responses, hormonal regulation, neutrophils, regulatory T cells, and glutamatergic synapses. Cell-type enrichment analysis identified excitatory and inhibitory neurons as key cellular contributors. Alterations in INT also correlated with cortical 5-HT1b and NAT receptor density, suggesting a role for inhibitory neurotransmission in temporal integration deficits.
This study advances understanding of treatment-related brain abnormalities in patients with MDD from the perspective of local neural dynamics. These findings support a multiscale pathophysiological framework involving brain connectivity dynamics, molecular architecture, and neurochemical regulation in MDD.
Xiang Liu, Yuanzhi He, Lifeng Li et al.· BMC Psychiatry· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
