Multimodal MRI Radiomics and Deep Learning for Brain Age Prediction: Age-Corrected Brain Age Gap Analysis in Patients With Insomnia

doi:10.21203/rs.3.rs-7949686/v1

Download PDF

Research Article

Multimodal MRI Radiomics and Deep Learning for Brain Age Prediction: Age-Corrected Brain Age Gap Analysis in Patients With Insomnia

https://doi.org/10.21203/rs.3.rs-7949686/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Objective

This study aimed to develop and validate a high-precision brain age prediction model by integrating multimodal MRI radiomics features from T1- and T2-weighted images with deep learning. The model was trained on healthy individuals for chronological age estimation and applied to patients with insomnia to calculate the Brain Age Gap (BAG), evaluating whether chronic insomnia is associated with accelerated brain aging.

Methods

A total of 1,200 participants were retrospectively included, comprising 942 healthy controls and 258 patients with insomnia. Healthy data were obtained from the IXI public dataset and Shenzhen Hospital (Futian), Guangzhou University of Chinese Medicine. All insomnia patients were recruited from the same hospital. T1- and T2-weighted MRI underwent standardized preprocessing, including resampling, gray-level discretization, and automated segmentation for radiomics feature extraction. After variance-based feature selection, multimodal features were combined to construct a deep learning regression model trained on healthy subjects and evaluated using mean absolute error (MAE), root mean square error (RMSE), and R². The model was then applied to the insomnia cohort to estimate BAG, followed by age-bias correction and group comparisons.

Results

Three models were constructed: T1-based, T2-based, and multimodal fusion. In validation, the T1 model achieved MAE of 7.58 years (R² = 0.57), the T2 model 7.90 years (R² = 0.51), and the fusion model 6.42 years (R² = 0.68; all p < 0.001). The insomnia group showed significantly higher BAG than controls both before (8.10 ± 8.57 vs. 1.26 ± 8.30 years, p < 0.00001) and after age correction (1.60 ± 6.49 vs. −2.18 ± 7.75 years, p < 0.00001).

Conclusion

The multimodal MRI radiomics–deep learning fusion model enables accurate brain age prediction and reveals evidence of accelerated brain aging in patients with insomnia.

Deep learning

Insomnia

Brain age gap

Multimodal analysis

In recent years, “Brain Age” — a biomarker based on neuroimaging data that estimates the physiological age of an individual’s brain — has garnered increasing attention. It not only reflects the structural and functional health of the brain but also provides an effective metric for assessing the progression of brain aging^[1–3]. Furthermore, the “Brain Age Gap” (BAG), defined as the difference between predicted brain age and chronological age (BAG = predicted brain age − chronological age), is regarded as a biomarker of “accelerated aging.” A positive BAG (BAG > 0) has been closely associated with various neurodegenerative diseases, psychiatric disorders, and declines in cognitive function^[4–7].

Insomnia disorder, one of the most prevalent sleep disorders, affects approximately 10% to 30% of the global adult population^{[8, 9]}. Chronic insomnia not only significantly impairs quality of life but has also been shown to be associated with cognitive deficits^[10], mood disturbances, and alterations in brain structure and function^[11]. However, the mechanisms underlying the impact of insomnia on the brain — in particular, whether it accelerates the brain’s physiological aging process — still require more objective and precise biological evidence.

From a technological perspective, medical imaging analysis is undergoing a shift from qualitative assessment to quantitative, high-dimensional feature extraction. Radiomics — an emerging technique — enables the extraction of vast amounts of high-dimensional quantitative features from conventional medical images, capturing microscopic tissue heterogeneity that may be imperceptible to the human eye. These features have been demonstrated to hold substantial potential for disease diagnosis and prognostic evaluation^[12–14].

Although current research on Brain Age and Brain Age Gap has been extensively applied to psychiatric disorders and neurodegenerative diseases, yielding substantial progress, systematic investigations into Brain Age and accelerated aging patterns in patients with chronic insomnia disorder remain relatively scarce. In particular, there is a notable lack of high-precision models that integrate the strengths of radiomics and deep learning to objectively quantify the impact of insomnia on brain aging. While previous studies have reported preliminary findings on structural and functional alterations in the brains of individuals with insomnia^[15], definitive biological evidence demonstrating whether insomnia induces an accelerated aging pattern in the brain remains insufficient. Some studies suggest that insomnia is associated with cognitive aging and may lead to cumulative effects on brain structure^[16–18], whereas others argue that primary insomnia is not linked to significant macroscopic structural changes in the brain^{[19, 20]}. This inconsistency underscores the pressing need to develop more sensitive and objective biomarkers capable of accurately assessing the impact of insomnia on brain health.

Current brain age prediction models often rely on a single feature or a limited number of data modalities, thereby constraining their ability to comprehensively capture complex pathophysiological processes, such as brain alterations associated with insomnia. To date, no high-precision brain age prediction models integrating radiomics and deep learning have been applied to populations with insomnia disorder. Consequently, effectively integrating multimodal imaging information to construct more refined and robust brain age prediction models — and thereby quantify brain age differences in insomnia patients — remains a central challenge in this field^[21–23].

This study aims to address this gap by developing a high-precision brain age prediction model that combines radiomics with deep learning, and applying it to individuals with insomnia disorder. By objectively quantifying the Brain Age Gap in insomnia patients, we seek to investigate whether chronic insomnia is associated with accelerated brain aging, thereby providing a novel perspective and tool for early intervention and treatment evaluation in insomnia.

We propose the following hypotheses: (1) The proposed fusion model can accurately predict the chronological age of healthy individuals. (2) Compared with healthy controls, the average Brain Age Gap in the insomnia disorder group will be significantly greater than zero, and significantly higher than that of the healthy control group.

To achieve these objectives, we will first build and validate a radiomics–deep learning fusion brain age prediction model based on large-scale healthy population datasets (the IXI dataset and a routine health examination cohort). The model will integrate the strengths of radiomics (by extracting localized quantitative features from regions of interest, ROI) and deep learning, enabling effective fusion of multimodal features to predict individual brain age^{[13, 14, 21–23]}. Subsequently, the trained model will be applied to an independent insomnia cohort and a matched healthy control cohort to calculate individual Brain Age Gaps. Finally, rigorous statistical analyses will be conducted to compare the Brain Age Gap between groups and to explore associations between BAG and clinical characteristics in insomnia patients^{[24, 25]}.

Basic Characteristics

A total of 1,200 participants were included in this study, comprising 942 individuals in the healthy cohort and 258 individuals in the insomnia cohort. For the brain age prediction module, the training set consisted of 753 participants, and the validation set included 189 participants. For the brain age gap analysis module, the healthy cohort corresponded to the validation set from the prediction module, while the insomnia cohort consisted of 258 participants. The mean age of the insomnia cohort was 52.90 years (range: 15–78), and the mean age of the healthy cohort was 46.54 years (range: 12–86.31).

Performance of Brain Age Prediction Model and Contribution of Multimodal Fusion

To evaluate the impact of different MRI modalities on brain age prediction performance, the model inputs were separately defined as radiomics features extracted from T1-weighted images, from T2-weighted images, and from a fusion of T1 and T2 features.

As shown in Fig. 1, the predictive performance of single-modality models varied. The T1-based model achieved a mean absolute error (MAE) of 7.58 years, a root mean square error (RMSE) of 9.67 years, and a coefficient of determination (R²) of 0.57, indicating a relatively good degree of fit. In contrast, the T2-based model yielded an MAE of 7.90 years, an RMSE of 10.36 years, and an R² of 0.51, suggesting slightly lower prediction accuracy and a reduced fit compared to the T1 model.

When features from both T1 and T2 modalities were fused, multimodal integration produced a marked improvement in predictive performance. The MAE decreased to 6.42 years, RMSE was reduced to 8.37 years, and R² increased to 0.68, demonstrating clear optimization in both error reduction and explanatory power. These results indicate that the feature information carried by T1 and T2 modalities is complementary: T1 is sensitive to structural integrity and gray matter morphology, whereas T2 is more responsive to tissue water content and white matter alterations. Their fusion provides a more comprehensive representation of brain characteristics, thereby enhancing the model’s capacity to capture individual differences in brain age.

Brain Age Gap and Accelerated Aging in Insomnia: Correlation, Age-Related Bias, and Group Comparisons

As illustrated in Fig. 2A, the proposed brain age prediction model successfully captured the age-related trends in structural brain changes. Predicted Brain Age showed a significant positive correlation with Chronological Age, with data points from both groups distributed roughly along the diagonal line. Notably, in the insomnia group (red), most data points were located above the diagonal, indicating that the predicted brain age was substantially greater than the chronological age, reflecting a clear pattern of advanced brain aging.

To further investigate this phenomenon, the Brain Age Gap (BAG = Predicted Age − Chronological Age) was calculated and its relationship with chronological age was examined. As shown in Fig. 2B, the raw BAG was significantly and positively correlated with chronological age (r = 0.599, p < 0.001), suggesting a systematic overestimation in older individuals—an age-related bias in the model. To eliminate the confounding effects of this bias on between-group comparisons, BAG was adjusted for age using a linear regression model. The results after adjustment are presented in Fig. 2C and Table 1. The insomnia group exhibited a markedly higher predicted brain age compared to chronological age (uncorrected mean BAG = 8.10 years; age-adjusted mean BAG = 1.60 years; both p < 0.001), indicating a pronounced trend of advanced brain aging and accelerated decline. In the healthy control group, the magnitude of BAG was smaller (uncorrected mean = 1.26 years, p = 0.038), and after age adjustment the mean BAG was slightly lower than the chronological age (− 2.18 years, p = 0.038). Between-group comparisons confirmed that BAG in the insomnia cohort was significantly greater than in healthy controls both before (difference ≈ 6.84 years, p < 0.00001) and after adjustment (difference ≈ 3.78 years, p < 0.00001).

Furthermore, Fig. 2D depicts the distribution of age-adjusted BAG in both groups. The curve for the insomnia group is shifted to the right, with a peak noticeably higher than that of healthy controls and concentrated within the positive BAG range. This distribution pattern aligns with the mean comparison results and visually supports the conclusion that most individuals in the insomnia group have a predicted brain age substantially exceeding their chronological age.

In summary, patients with insomnia not only exhibited a larger BAG than healthy individuals but also maintained significantly elevated predicted brain age even after removing age-related bias. These findings suggest that insomnia may be closely linked to accelerated brain aging, with structural brain characteristics potentially reflecting an earlier and steeper trajectory of decline.

Table 1
Brain Age Gap and Age-Adjusted BAG in Healthy and Insomnia Cohorts.
	n	Brain Age Gap		Age-corrected Brain Age Gap		r
	n	Mean	SD	Mean	SD
Normal Cohort	189	1.26	8.3	-2.18	7.75	0.599(p < 0.001)
Insomnia Cohort	258	8.1	8.57	1.6	6.49

Study Population

As shown in Fig. 3, we conducted a retrospective study analyzing data from Shenzhen Hospital (Futian) of Guangzhou University of Chinese Medicine, focusing on normal and insomnia cases for which brain MRI imaging data were available (Time frame: January 2019 to August 2024). This study was performed in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Shenzhen Hospital (Futian), Guangzhou University of Chinese Medicine (Approval No.: GZYLL(KY)-2025-117). Given the retrospective design, the requirement for informed consent was waived, and all patient data were handled with strict measures to ensure privacy and confidentiality.

In addition, this study utilized the publicly accessible IXI dataset, which was collected from multiple hospitals in the United Kingdom and contains multimodal brain MR images of healthy participants for scientific research purposes only. The use of the IXI dataset complied with its data usage agreement and did not involve any identifiable personal information; therefore, no additional ethical approval was required.

The exclusion criteria for this study were as follows: a documented history of neurological disorders (e.g., stroke, traumatic brain injury, epilepsy, Parkinson’s disease, multiple sclerosis); diagnosis of severe psychiatric disorders (e.g., schizophrenia, bipolar disorder, major depressive disorder); presence of severe systemic diseases (e.g., serious cardiovascular conditions, hepatic or renal insufficiency, or uncontrolled hypertension/diabetes); MRI findings indicating significant structural brain abnormalities (e.g., intracranial tumors, intracranial hemorrhage, congenital malformations); other sleep disorders known to influence sleep quality (e.g., sleep apnea, narcolepsy, restless legs syndrome); contraindications to MRI examination (e.g., metallic implants, claustrophobia); and individuals who failed to complete MRI scanning or whose imaging data did not meet quality requirements.

This study was designed to consist of two modules: the brain age prediction module and the brain age gap analysis module. The dataset for the brain age prediction module was constructed by integrating the publicly available IXI dataset with MRI data from healthy participants recruited at Shenzhen Hospital (Futian), Guangzhou University of Chinese Medicine, forming a combined healthy dataset. All imaging data underwent standardized preprocessing after acquisition and were randomly allocated into a training cohort (n = 753, 80%) and a validation cohort (n = 189, 20%) using computer-assisted algorithms. The brain age gap analysis module comprised the validation cohort from the prediction module along with an additional cohort of insomnia patients, yielding a total of 258 subjects. This design allowed the validation cohort to be used both for model performance assessment and for comparative analysis with the insomnia cohort, thereby verifying the applicability of the brain age prediction model in populations with insomnia disorder.

Radiomics Feature Detection

The study workflow is presented in Fig. 4. The first stage, the brain age prediction module, involved collecting T1- and T2-weighted brain MRI images from healthy participants and applying automated segmentation methods to extract brain regions, defined as regions of interest (ROIs). Prior to radiomics feature extraction, all images underwent preprocessing to minimize the influence of variability caused by differences in MRI scanners and scanning parameters. To reduce variability in radiomics features, image resampling and gray-level discretization techniques were applied for standardized processing. Each ROI was resampled to dimensions of 256 × 256 × 48 before feature extraction, which included first-order statistical features, shape descriptors, and texture features. In total, 107 radiomics features were extracted from each modality-specific ROI. Subsequently, a variance-based feature selection method was employed to eliminate features with extremely low variance and minimal informational value, retaining 97 features in total. This approach reduced redundancy and enhanced the stability and predictive performance of the subsequent analytical models.

Model Development

In this study, a deep learning–based feature fusion approach was employed, using the variance-filtered radiomics features to construct the brain age regression model. Specifically, radiomics features derived from T1-weighted andT2-weighted images were integrated into a unified brain age prediction framework to capture complex nonlinear relationships between inputs from different modalities. The core of the model was implemented using a multilayer perceptron (MLP), with each participant’s chronological age serving as the regression label. The optimization objective was to minimize the mean absolute error (MAE) and root mean square error (RMSE), while the coefficient of determination (R²) was used to evaluate the degree of fit between predicted and actual ages.

In the brain age gap analysis module, differences between predicted brain age and chronological age (i.e., Brain Age Gap) were compared between healthy controls and patients with insomnia, in order to investigate the potential impact of insomnia on brain age and to verify the association between disease status and deviations in brain age.

Statistical Analysis

All statistical analyses were performed in a Python environment using open-source libraries. Model performance was evaluated using mean absolute error (MAE), root mean square error (RMSE), and the coefficient of determination (R²), which quantify the degree of fit between predicted and chronological age. The performance of models based on different MRI modalities (T1, T2, and T1 + T2 fusion) was assessed by directly comparing the error metrics and goodness-of-fit measures.

To detect potential age-related bias in the model, the Brain Age Gap (BAG = Predicted Brain Age − Chronological Age) was calculated for each participant, and Pearson’s correlation analysis was used to assess the linear relationship between BAG and chronological age. If a significant correlation was identified, a linear regression model was applied to adjust BAG for age in order to remove systematic bias. Group comparisons of BAG between the insomnia and healthy control cohorts were conducted using independent-samples t-tests, performed both before and after age adjustment. For within-group comparisons of mean BAG before and after adjustment, paired t-tests were used.

Table 2
Age-Stratified Comparison of Age-Corrected Brain Age Gap Between Healthy and Insomnia Cohorts
Age Group (years)	Normal Cohort (n)	Normal Cohort (Mean)	Insomnia Cohort (n)	Insomnia Cohort (Mean)	p
< 30	27	0.35	8	−4.13	0.07586
30–40	50	−0.15	16	−0.34	0.88497
40–50	47	−2.62	65	0.43	0.03484
> 50	65	-4.48	169	2.50	< 0.001

In this study, we developed a radiomics–deep learning fusion framework to predict brain age from T1- and T2-weighted MRI and to quantify the brain age gap (BAG) in individuals with insomnia. Our results showed that multimodal integration of T1 and T2 features improved predictive accuracy compared with single-modality models, highlighting the complementary nature of structural and tissue-sensitive information. While predicted brain age was strongly correlated with chronological age, the raw BAG exhibited age-related bias, necessitating age correction to allow valid intergroup comparisons. After age correction, the insomnia cohort exhibited consistently higher BAG values than the healthy cohort, suggesting a potential link between insomnia and accelerated brain aging.

Age-stratified analysis provided preliminary (Table 2), exploratory insights into this relationship. After age correction, minimal and nonsignificant differences were observed in participants younger than 40 years. In the 40–50 year range, BAG in the insomnia group was modestly higher than in the healthy group (p = 0.03484), while in participants over 50 years of age, the difference was more pronounced (healthy mean − 4.48 years vs. insomnia mean 2.50 years; p < 0.001). However, these findings are based on cross-sectional data with uneven sample distribution and may be influenced by unmeasured confounders, limiting the strength of the conclusions. Rather than establishing causality, these results hint at a possible trend in which insomnia-related BAG may be more observable in midlife and older adults. One plausible hypothesis, still requiring verification in larger, longitudinal cohorts, is that age-related declines in neural plasticity, cumulative inflammatory burden, vascular changes, and myelin integrity might increase vulnerability to the neurobiological effects of chronic sleep disturbance. Notably, in the insomnia group, the age-corrected BAG distribution curve was slightly shifted toward positive values and showed tighter clustering compared with the healthy group. Although this observation aligns with the mean difference analysis, replication in independent datasets is needed to confirm its robustness.

From a methodological perspective, the use of variance-filtered radiomics features and standardized preprocessing enhanced model stability and reduced variability introduced by different scanners or acquisition protocols. Explicit detection and correction of age-related bias avoided a common pitfall in brain age research, ensuring that observed differences were not artifacts of residual age effects. Additionally, the comparison among T1, T2, and multimodal fusion features clarified the added value of capturing nonredundant structural information.

These findings agree with prior literature linking insomnia to alterations in gray matter morphometry, white matter microstructure, and neuroinflammatory processes. Importantly, the persistence of differences after age correction strengthens the hypothesis of an insomnia–brain aging relationship, although the age-stratified results should be interpreted with caution due to their exploratory nature. Clinically, BAG may have potential as a sensitive proxy for brain health burden in insomnia, particularly in older adults. If validated longitudinally, BAG could be used to monitor treatment effects and identify individuals at greatest risk for cognitive decline or other brain health outcomes.

Limitations of the present study include the single-center retrospective design, possible residual confounding, and lack of causal inference given the cross-sectional approach. Future research should involve multi-center prospective recruitment, longitudinal imaging follow-up, and integration with complementary biomarkers (e.g., diffusion metrics, quantitative T1/T2 imaging, functional connectivity, and blood-based indicators). Combining BAG with clinical, cognitive, and lifestyle variables may help build individualized risk models for precision sleep medicine.

In summary, using multimodal MRI radiomics fused via deep learning, coupled with age bias correction, we observed higher brain age estimates in insomnia—particularly in older adults—supporting the concept of insomnia-related accelerated brain aging while also emphasizing the preliminary nature of age-stratified findings. These results highlight the need for longitudinal, mechanistic investigations to validate BAG as a biomarker for risk stratification and therapeutic monitoring in sleep disorders.

Ethical Approval

This study was performed in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Shenzhen Hospital (Futian), Guangzhou University of Chinese Medicine (Approval No.: GZYLL(KY)-2025-117).

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Author Contribution

Dr Jingshan Gong had full access to all the data in the study and took responsibility for the integrity of the data and the accuracy of the data analysis. Jiandong Guo conceived and designed the study. Junxiong Zhao, Yue Zhou, and Yongyi Li were responsible for data collection and preprocessing. Shasha Zeng implemented data quality control and algorithm verification. All authors participated in data analysis and interpretation. Jiandong Guo and Junxiong Zhao performed the statistical analyses. Shasha Zeng developed, trained, and applied the predictive model. Shasha Zeng drafted the initial manuscript, and Jingshan Gong critically revised it for important intellectual content. All authors reviewed and approved the final version of the manuscript.

Data Availability

The data that supports the findings of this study are available from the corresponding authors with a signed data access agreement. The raw image data are not publicly available because they contain sensitive information that could compromise patient privacy.

Soumya Kumari L K, Sundarrajan R. A review on brain age prediction models[J]. Brain Research, 2024,1823:148668.
Franke K, Gaser C. Ten years of brainage as a neuroimaging biomarker of brain aging: what insights have we gained?[J]. Frontiers in Neurology, 2019,Volume 10–2019.
Baecker L, Garcia-Dias R, Vieira S, et al. Machine learning for brain age prediction: introduction to methods and clinical applications[J]. Ebiomedicine, 2021,72:103600.
Chang X, Jia X, Eickhoff S B, et al. Multi-center brain age prediction via dual-modality fusion convolutional network[J]. Medical Image Analysis, 2025,101:103455.
Ballester P L, Romano M T, de Azevedo C T, et al. Brain age in mood and psychotic disorders: a systematic review and meta-analysis[J]. Acta Psychiatr Scand, 2022,145(1):42–55.
Biondo F, Jewell A, Pritchard M, et al. Brain-age is associated with progression to dementia in memory clinic patients[J]. Neuroimage Clin, 2022,36:103175.
Stefaniak J D, Mak E, Su L, et al. Brain age gap, dementia risk factors and cognition in middle age[J]. Brain Commun, 2024,6(6):e392.
van Straten A, Weinreich K J, Fábián B, et al. The prevalence of insomnia disorder in the general population: a meta-analysis[J]. J Sleep Res, 2025,34(5):e70089.
Aernout E, Benradia I, Hazo J B, et al. International study of the prevalence and factors associated with insomnia in the general population[J]. Sleep Med, 2021,82:186–192.
Pang R, Zhan Y, Zhang Y, et al. Aberrant functional connectivity architecture in participants with chronic insomnia disorder accompanying cognitive dysfunction: a whole-brain, data-driven analysis[J]. Front Neurosci, 2017,11:259.
Joo E Y, Kim H, Suh S, et al. Hippocampal substructural vulnerability to sleep disturbance and cognitive impairment in patients with chronic primary insomnia: magnetic resonance imaging morphometry[J]. Sleep, 2014,37(7):1189–1198.
Wang W, Yang G, Liu Y, et al. Multimodal deep learning model for prognostic prediction in cervical cancer receiving definitive radiotherapy: a multi-center study[J]. Npj Digital Medicine, 2025,8(1):503.
Rogers W, Thulasi S S, Refaee T, et al. Radiomics: from qualitative to quantitative imaging[J]. Br J Radiol, 2020,93(1108):20190948.
Lambin P, Woodruff H C, Mali S A, et al. Radiomics quality score 2.0: towards radiomics readiness levels and clinical translation for personalized medicine[J]. Nature Reviews Clinical Oncology, 2025,22(11):831–846.
Guo X, Ding Y, Xu W, et al. Predicting brain age gap with radiomics and automl: a promising approach for age-related brain degeneration biomarkers[J]. Journal of Neuroradiology, 2024,51(3):265–273.
Sexton C E, Sykara K, Karageorgiou E, et al. Connections between insomnia and cognitive aging[J]. Neurosci Bull, 2020,36(1):77–84.
Stolicyn A, Lyall L M, Lyall D M, et al. Comprehensive assessment of sleep duration, insomnia, and brain structure within the uk biobank cohort[J]. Sleep, 2024,47(2).
Miao Y, Wang J, Li X, et al. Poor sleep health is associated with older brain age: the role of systemic inflammation[J]. Ebiomedicine, 2025,120:105941.
Spiegelhalder K, Regen W, Baglioni C, et al. Insomnia does not appear to be associated with substantial structural brain changes[J]. Sleep, 2013,36(5):731–737.
Weihs A, Frenzel S, Bi H, et al. Lack of structural brain alterations associated with insomnia: findings from the enigma-sleep working group[J]. J Sleep Res, 2023,32(5):e13884.
Luo N, Shi W, Yang Z, et al. Multimodal fusion of brain imaging data: methods and applications[J]. Machine Intelligence Research, 2024,21(1):136–152.
Lee J, Burkett B J, Min H K, et al. Deep learning-based brain age prediction in normal aging and dementia[J]. Nat Aging, 2022,2(5):412–424.
de Lange A G, Anatürk M, Rokicki J, et al. Mind the gap: performance metric evaluation in brain-age prediction[J]. Hum Brain Mapp, 2022,43(10):3113–3129.
de Lange A G, Cole J H. Commentary: correction procedures in brain-age prediction[J]. Neuroimage Clin, 2020,26:102229.
Yu Y, Cui H Q, Haas S S, et al. Brain-age prediction: systematic evaluation of site effects, and sample age range and size[J]. Hum Brain Mapp, 2024,45(10):e26768.

No competing interests reported.

Download PDF

Editorial decision: Revision requested
13 Dec, 2025
Reviews received at journal
01 Dec, 2025
Reviewers agreed at journal
30 Nov, 2025
Reviewers agreed at journal
28 Nov, 2025
Reviewers agreed at journal
28 Nov, 2025
Reviewers agreed at journal
05 Nov, 2025
Reviewers invited by journal
03 Nov, 2025
Editor assigned by journal
30 Oct, 2025
Submission checks completed at journal
30 Oct, 2025
First submitted to journal
26 Oct, 2025

You are reading this latest preprint version

Multimodal MRI Radiomics and Deep Learning for Brain Age Prediction: Age-Corrected Brain Age Gap Analysis in Patients With Insomnia

Status:

Version 1

Abstract

Objective

Methods

Results

Conclusion

Figures

Introduction

Results

Basic Characteristics

Performance of Brain Age Prediction Model and Contribution of Multimodal Fusion

Brain Age Gap and Accelerated Aging in Insomnia: Correlation, Age-Related Bias, and Group Comparisons

Methods

Study Population

Radiomics Feature Detection

Model Development

Statistical Analysis

Discussion

Declarations

Funding

Author Contribution

Data Availability

References

Additional Declarations

Status:

Version 1