Synthetic MRI Neuroimaging Correlates of Frailty: Correlations Between Frailty Scales and Brain Structural Integrity

ABSTRACT

Objectives

Frailty, a common condition in the elderly, is characterized by a decline in physiological reserves and an increased vulnerability to stressors, leading to negative health outcomes such as falls, fractures, and prolonged recovery from illness. Despite its clinical significance, the relationship between frailty and cerebral health, particularly the correlation of frailty evaluation scales with brain structural integrity changes, is not well‐defined. The study aims to investigate the associations between frailty and brain structure, focusing on the relationship between various frailty scales and brain volumetric and relaxometric changes as identified by synthetic MRI neuroimaging techniques.

Methods

This prospective study initially enrolled 109 participants from the neurology ward. After applying exclusion criteria, 41 participants were included in the final analysis. Data on demographics, medical history, and physical examinations were collected. Synthetic MRI was conducted using a MAGnetic resonance image Compilation (MAGiC) with images processed using SyMRI 8.0 and SPM12 software for volumetric and relaxometric analyses. Frailty was assessed using the Clinical Frailty Scale (CFS), Fried scale, FRAIL scale, and Edmonton Frailty Scale (EFS). Statistical analysis was performed by partial correlation analysis, adjusting for age, gender, and history of stroke, to examine the relationships between global and regional brain volumetric and relaxometry parameters with frailty scales. To control for multiple comparisons, the false discovery rate (FDR) method was applied, with statistical significance set at FDR‐adjusted p value < 0.05.

Results

Compared to non‐frail participants, frail individuals exhibited significantly lower gray matter volumes in the cerebellum, hippocampus, and thalamus, as well as reduced white matter volumes globally and in the temporal and parietal lobes. Brain relaxometry revealed higher hippocampal and thalamic T1 and T2 values in the frail group. The CFS and EFS were significantly correlated with global and regional gray and white matter volumes, indicating a link between frailty severity and brain structural integrity.

Conclusions

This study highlights significant correlations between frailty scales, particularly CFS and EFS, and changes in brain volumetry and relaxometry in specific brain regions. These findings suggest that the CFS and EFS may serve as sensitive frailty scales for assessing the impact of frailty on brain structure. Synthetic MRI holds promise as a valuable neuroimaging tool for diagnosing and evaluating frailty. Further research is needed to confirm these findings and enhance clinical strategies for frailty assessment and intervention.

Frailty was associated with decreased gray and white matter brain volumes in older adults. Synthetic MRI further revealed relaxometry alterations correlated with frailty severity across multiple clinical scales.

1. Introduction

Frailty is a prevalent condition in the geriatric population, affecting approximately 25% of individuals over 65 years old and nearly 50% of those over 85 [1, 2]. It is characterized by a decline in physiological reserves, making individuals more vulnerable to stressors such as falls and illness, which can increase the risk of disability, prolonged recovery, and mortality [3, 4, 5, 6, 7]. Consequently, frailty has emerged as a major health concern in aging societies.

Despite its prevalence, there is no single, universally accepted standardized diagnostic tool for frailty, and various clinical scales are employed to assess its severity. Commonly used scales include the Fried scale [8], FRAIL scale [9], Frailty Index [10], Edmonton Frailty Scale (EFS) [11], and Clinical Frailty Scale (CFS) [12, 13]. Among these, the CFS, developed during the Canadian Study of Health and Aging (CSHA), has gained recognition due to its comprehensive assessment of mobility, cognitive function, and recent changes in overall health status, with a scoring range from 1 (very fit) to 9 (terminally ill). A CFS score above 4 generally indicates frailty [14]. Its ease of use, combined with its ability to provide a rapid assessment, makes the CFS a valuable tool in clinical settings [15].

While clinical assessment tools for frailty are well‐established, research on the relationship between frailty and brain health remains limited. Brain structural decline may not only reflect systemic frailty but could also be its neurobiological underpinning. Studies investigating the correlation between frailty and brain health have produced conflicting results. Some research reports an association between frailty, assessed by the Fried scale, and reduced gray matter (GM) and white matter (WM) volumes [16, 17, 18], while other studies have found no significant association using the frailty index. These inconsistencies may be attributed to differences in study design, the sensitivity of neuroimaging techniques, and variability in frailty assessment scales across studies.

To explore the relationship between frailty and brain health, this study employs synthetic MRI, using a MAGnetic resonance image Compilation (MAGiC), to measure the global and regional brain volume and relaxometry. It provides precise measurements of GM and WM volumes, while relaxometry, which measures MRI relaxation times, offers insights into microstructural changes in the brain, including alterations in myelin and axon integrity [19].

This study aims to investigate the relationship between multiple frailty scales—including the CFS, Fried scale, FRAIL scale, and EFS—and brain volumetric and relaxometric changes using synthetic MRI. By employing these advanced neuroimaging methods, we seek to provide a clearer understanding of the link between frailty and brain structure, potentially improving diagnostic and therapeutic approaches for frailty in the elderly.

2. Methods

2.1. Subjects

This prospective study enrolled 109 consecutive patients from the neurology ward between September 2020 and February 2021. Inclusion criteria were: (1) age ≥ 65 years, and (2) voluntary participation. Exclusion criteria included: (1) inability to cooperate with frailty assessments (due to coma, unstable vital signs, deafness, blindness, severe dementia, etc.) (n = 8); (2) patients with acute ischemic stroke, Parkinson's disease, brain tumor, or other severe cerebral diseases (n = 41); (3) contraindications to MRI (n = 5) or poor MRI image quality (n = 4); and (4) inability to provide informed consent (n = 10). A total of 41 patients were ultimately included in the study.

2.2. Information Collection

Demographic information, including sex and age, was collected, along with medical history covering hypertension, diabetes mellitus, stroke, and coronary heart disease. Physical examination data included body mass index (BMI), systolic blood pressure (SBP), and diastolic blood pressure (DBP). We also recorded current medication use and laboratory blood test results (white blood cell count, hemoglobin, platelet count, glucose, and albumin), which were collected within the first three days of hospital admission.

2.3. MRI Protocol

MRI scans were conducted using a 3.0 T MRI machine (SIGNA Pioneer, GE Healthcare, Milwaukee, WI, USA) equipped with a 32‐channel head–neck coil. Routine sequences included axial and sagittal T1‐weighted imaging, axial T2‐weighted imaging, T2 fluid‐attenuated inversion recovery (FLAIR), and diffusion‐weighted imaging (DWI). A three‐dimensional fast spoiled gradient‐recalled echo (3D‐FSPGR) sequence was used to acquire T1‐weighted images through the whole head, parallel to the anterior–posterior commissure. The acquisition parameters were: Repetition time = 6.0 ms, echo time = 1.9 ms, flip angle = 11°, bandwidth = 31.25 kHz, matrix size = 256 × 256, field of view = 256 mm × 256 mm, and 200 continuous slices with a slice thickness of 1 mm. Synthetic MRI (MAGnetic resonance image Compilation, MAGiC) images were also obtained with parameters: Repetition time = 4,906 ms, echo‐train length = 12, echo time 1/echo time 2 = 22.0/87.9 ms, bandwidth = 22.73 kHz, matrix size = 320 × 256, field of view = 240 mm × 240 mm, 32 continuous slices with a slice thickness of 4 mm and 0.5 mm gap.

2.4. MRI Processing

The MRI image processing method has been introduced in our previous study [20]. Briefly, we processed the raw data from the MAGiC images using SyMRI 8.0 software (SyntheticMR AB, Linköping, Sweden) to obtain parameters for global gray matter volume (GMV), white matter volume (WMV), cerebrospinal fluid volume (CSFV), and myelin volume, along with intracranial volume (ICV, calculated as GMV + WMV + CSFV). The streamlined process involved rigid registration of relaxation maps from MAGiC to 3D‐FSPGR images using SPM12 software (http://www.fil.ion.ucl.ac.uk/spm/), followed by segmentation and normalization to obtain tissue probability maps and normalized relaxation maps with the CAT12 toolbox. Global relaxometry measurements across gray matter, white matter, and cerebrospinal fluid (CSF) were derived by averaging relaxation values from voxels with a partial volume exceeding 95% of the respective tissue type. Regional relaxometry and volumetric data were extracted based on the Anatomical Automatic Labeling version 3 (AAL3) atlas (http://www.gin.cnrs.fr/en/tools/aal/). Regional definitions were consolidated from atlas‐based parcellations according to predefined rules. First, subregional parcels belonging to the same anatomical structure (e.g., anterior, middle, and posterior subdivisions) were combined into unified regions. Second, left and right hemispheric counterparts were merged to generate bilateral regions. Third, these bilateral regions were further organized into larger composite brain regions (e.g., frontal, temporal, and parietal lobes) according to established neuroanatomical conventions. This hierarchical integration procedure was applied consistently across both volumetric and relaxometry analyses.

2.5. Frailty Assessment

Frailty was evaluated using multiple standardized scales. The CFS, developed by Rockwood et al., assessed functional performance, comorbidities, and cognitive capacity, with a score range from 1 to 9 (higher scores indicating greater frailty). Participants with a CFS score > 4 were classified as frail [14]. Additionally, the Fried scale, FRAIL scale, and EFS were used to further quantify frailty. Physical function was assessed using the Short Physical Performance Battery (SPPB), which included grip strength and 4‐m walking speed. Cognitive function was evaluated via the Mini‐Mental State Examination (MMSE), and nutritional status was measured using the short‐form mini‐nutritional assessment (MNA‐SF). The ability to perform daily tasks was assessed using the Basic Activities of Daily Living (BADL) and Instrumental Activities of Daily Living (IADL) scales.

2.6. Statistical Analysis

Participants were categorized into frail (CFS > 4) and non‐frail groups. Descriptive statistics for all participants were reported as means (± standard deviation) for continuous variables with normal distribution, or medians (inter‐quartile range) for non‐normal variables. Categorical variables were described as absolute numbers (percentages). Group comparisons for demographic and clinical variables were performed using independent sample t‐tests (for continuous variables) or χ ² tests (for categorical variables).

For neuroimaging data, Analysis of Covariance (ANCOVA) was employed to compare differences between frail and non‐frail groups to control for potential confounders. Specifically, Volumetric data comparisons were adjusted for age, gender, history of stroke, and ICV. Relaxometry data comparisons were adjusted for age, gender, and history of stroke. p‐values < 0.05 were considered statistically significant.

Partial correlation analysis was conducted to evaluate the associations between frailty scales and neuroimaging metrics (volumetry and relaxometry), while controlling for potential confounders. This was implemented using multiple linear regression models. For the volumetric analysis, the models were adjusted for age, gender, history of stroke, and total ICV. For the relaxometry analysis, the models were adjusted for age, gender, and history of stroke. The partial correlation coefficients (r) were derived from the t‐statistics of the regression coefficients using the formula: ($r=\frac{t}{\sqrt{t^2+{\mathrm{df}}_{\text{resid}}}}$) where df_resid represents the residual degrees of freedom. To account for multiple comparisons, raw p‐values were corrected using the False Discovery Rate (FDR) method (Benjamini–Hochberg procedure). Statistical significance was defined as an FDR‐ adjusted p‐values < 0.05.

All analyses were conducted using R software (version 4.0.3, R Foundation for Statistical Computing, Vienna, Austria).

3. Results

3.1. Participant Characteristics

A total of 41 participants were included in the analysis, of which 19 were classified as frail based on a CFS score greater than 4. The mean age of participants was 76.29 years (±5.49 years), with 56.10% being female. No significant differences were found between frail and non‐frail participants regarding age, gender, years of education, or the prevalence of conditions such as hypertension, coronary heart disease, diabetes, or stroke (p > 0.05). Additionally, there were no significant differences in smoking status, drinking status, BMI, SBP, DBP, white blood cell count, hemoglobin levels, platelet count, glucose levels, or albumin levels between the frail and non‐frail groups (p > 0.05). However, the frail group exhibited a significantly higher modified Fazekas score compared to the non‐frail group (χ ² = 9.102, p = 0.028).

The frail group performed significantly worse in several key frailty‐related assessments. Compared to the non‐frail group, they had lower scores in physical function (SPPB), cognitive function (MMSE), nutritional status (MNA‐SF), and activities of daily living (BADL and IADL) (all p < 0.05). Additionally, frail participants had significantly higher scores on the Fried scale, FRAIL scale, and EFS, indicating a higher degree of frailty across multiple domains (p < 0.001) (Table 1).

TABLE 1.

Demographic data, clinical data, and frailty assessment of the participants.

	Total (n = 41)	Frail (n = 19)	Non‐frail (n = 22)	t/ꭓ2 value	p
Age (years)	76.29 ± 5.49	78.00 ± 4.46	74.82 ± 5.95	1.953	0.058
Gender, female, n (%)	23 (56.10)	12 (63.16)	11 (50.00)	0.717	0.397
CFS	4.15 ± 1.41	5.47 ± 0.70	3.00 ± 0.62	11.948	< 0.001
Fried scale	2.12 ± 1.17	2.89 ± 0.94	1.45 ± 0.91	4.971	< 0.001
FRAIL scale	1.51 ± 1.16	2.47 ± 0.77	0.68 ± 0.72	7.661	< 0.001
EFS	6.85 ± 2.34	8.89 ± 1.33	5.09 ± 1.38	8.987	< 0.001
MMSE	22.90 ± 5.24	20.74 ± 4.77	24.77 ± 4.99	−2.643	0.012
SPPB	4.61 ± 3.49	1.79 ± 2.23	7.05 ± 2.36	−7.332	< 0.001
MNA‐SF	11.49 ± 2.00	10.58 ± 2.32	12.27 ± 1.28	−2.835	0.009
ADL	74.39 ± 24.37	60.26 ± 25.14	86.59 ± 15.92	−3.935	< 0.001
IADL	4.88 ± 1.55	3.95 ± 1.84	5.68 ± 0.48	−3.994	< 0.001
Education year (years)	10.73 ± 4.90	10.74 ± 5.73	10.73 ± 4.19	0.006	0.995
Hypertension, n (%)	33 (80.49)	15 (78.95)	18 (81.82)	0.054	0.817
Coronary heart disease, n (%)	14 (34.15)	5 (26.32)	9 (40.91)	0.966	0.326
Diabetes mellitus, n (%)	14 (34.15)	6 (31.58)	8 (36.36)	0.104	0.747
Stroke, n (%)	34 (82.93)	16 (84.21)	18 (81.82)	0.041	0.839
Current smoker, n (%)	6 (14.63)	2 (10.53)	4 (18.18)	0.478	0.489
Current drinker, n (%)	4 (9.76)	1 (5.26)	3 (13.64)	0.812	0.368
BMI (kg/m2)	24.44 ± 4.02	23.93 ± 5.03	24.89 ± 2.94	−0.732	0.470
SBP (mmHg)	139.09 ± 18.63	140.26 ± 20.24	138.09 ± 17.55	0.364	0.718
DBP (mmHg)	75.54 ± 10.77	76.95 ± 8.04	74.32 ± 12.73	0.801	0.428
White blood cell (109/L)	6.47 ± 1.83	7.01 ± 2.09	6.00 ± 1.47	1.752	0.089
Hb (g/L)	129.22 ± 16.47	125.00 ± 18.25	132.86 ± 14.19	−1.522	0.137
Platelet (109/L)	220.93 ± 65.51	233.58 ± 83.63	210.00 ± 43.70	1.105	0.279
Glucose (mmol/L)	5.68 ± 1.61	5.42 ± 1.27	5.91 ± 1.86	−1.003	0.322
Albumin (g/L)	38.93 ± 2.40	38.47 ± 2.67	39.32 ± 2.12	−1.107	0.276
Modified Fazekas score, n (%)				9.102	0.028
0	3 (7.32)	0 (0.00)	3 (13.64)
1	11 (26.83)	2 (10.53)	9 (40.91)
2	16 (39.02)	10 (52.63)	6 (27.27)
3	11 (26.83)	7 (36.84)	4 (18.18)

Abbreviations: ADL, activities of daily living; BMI, body mass index; CFS, clinical frailty scale; DBP, diastolic blood pressure; EFS, Edmonton frailty scale; Hb, hemoglobin; IADL, instrumental activities of daily living; MMSE, Mini‐Mental State Examination; MNA‐SF, short‐form mini‐nutritional assessment; SBP, systolic blood pressure; SPPB, short physical performance battery.

3.2. Brain Volumetry

3.2.1. Brain Volumetry: Comparison Between the Frail and Non‐Frail Groups

After adjusting for age, gender, history of stroke, and ICV, significant differences in specific brain volumes were observed between the frail and non‐frail groups (Table 2). Frail participants exhibited significantly lower cerebellar gray matter volume (GMV) (58.05 ± 4.89 mL vs. 66.19 ± 5.80 mL, p < 0.001), hippocampal GMV (6.30 ± 1.19 mL vs. 7.58 ± 1.07 mL, p = 0.011), and thalamic GMV (5.74 ± 1.18 mL vs. 7.11 ± 1.59 mL, p = 0.018) compared to the non‐frail group.

TABLE 2.

Brain volumetry comparison between frail and non‐frail groups.

	Total (n = 41)	Frail (n = 19)	Non‐frail (n = 22)	F value	p
Global GMV, mL	518.64 ± 61.34	490.40 ± 56.05	543.04 ± 55.94	−1.96	0.058
Frontal GMV, mL	76.95 ± 10.94	72.90 ± 10.79	80.45 ± 10.02	−0.99	0.329
Temporal GMV, mL	70.12 ± 10.27	66.12 ± 9.71	73.58 ± 9.65	−0.99	0.328
Parietal GMV, mL	18.44 ± 2.20	17.52 ± 2.05	19.24 ± 2.05	−1.51	0.139
Occipital GMV, mL	24.03 ± 3.38	22.73 ± 3.17	25.15 ± 3.21	−0.98	0.334
Cerebellum GMV, mL	62.42 ± 6.73	58.05 ± 4.89	66.19 ± 5.80	−4.15	< 0.001 ***
Hippocampus GMV, mL	6.99 ± 1.28	6.30 ± 1.19	7.58 ± 1.07	−2.68	0.011 *
Thalamus GMV, mL	6.48 ± 1.56	5.74 ± 1.18	7.11 ± 1.59	−2.48	0.018 *
Global WMV, mL	429.80 ± 54.58	399.79 ± 47.34	455.72 ± 47.27	−2.84	0.007 **
Frontal WMV, mL	45.22 ± 6.82	42.49 ± 6.20	47.58 ± 6.56	−1.28	0.208
Temporal WMV, mL	35.58 ± 5.21	32.99 ± 3.98	37.81 ± 5.18	−2.33	0.026 *
Parietal WMV, mL	7.80 ± 1.29	7.10 ± 1.00	8.41 ± 1.22	−2.58	0.014 *
Occipital WMV, mL	17.22 ± 2.36	16.33 ± 2.58	17.99 ± 1.88	−1.43	0.162
Cerebellum WMV, mL	22.21 ± 4.09	21.64 ± 3.82	22.71 ± 4.35	0.18	0.856
Hippocampus WMV, mL	1.89 ± 0.27	1.79 ± 0.26	1.98 ± 0.25	−1.17	0.250
Thalamus WMV	5.29 ± 0.69	5.28 ± 0.83	5.31 ± 0.55	0.55	0.588
CSF volume, mL	413.00 ± 65.44	429.49 ± 74.13	398.75 ± 54.65	2.77	0.009 **
Myelin volume, mL	134.20 ± 20.23	122.52 ± 13.97	144.30 ± 19.54	−3.14	0.003 **
ICV, mL	1402.78 ± 132.22	1368.37 ± 122.54	1432.50 ± 135.77	−1.34	0.188

Abbreviations: CSF, cerebrospinal fluid; GMV, gray matter volume; ICV, Intracranial volume; WMV, white matter volume.

^* p < 0.05.

^** p < 0.01.

^*** p < 0.001.

For white matter volume (WMV), frail participants had significantly lower global WMV (399.79 ± 47.34 mL vs. 455.72 ± 47.27 mL, p = 0.007), temporal WMV (32.99 ± 3.98 mL vs. 37.81 ± 5.18 mL, p = 0.026), and parietal WMV (7.10 ± 1.00 mL vs. 8.41 ± 1.22 mL, p = 0.014).

Myelin volume was also significantly reduced in the frail group (122.52 ± 13.97 mL vs. 144.30 ± 19.54 mL, p = 0.003). Additionally, CSFV was significantly higher in the frail group compared to the non‐frail group (429.49 ± 74.13 mL vs. 398.75 ± 54.65 mL, p = 0.009). No significant differences were found in global GMV, frontal GMV, temporal GMV, parietal GMV, occipital GMV, frontal WMV, occipital WMV, cerebellar WMV, hippocampal WMV, thalamic WMV, or ICV between the groups (p > 0.05).

3.3. Correlation Between Frailty Scales and Brain Volumetry

After adjusting for age, gender, history of stroke, and ICV, significant correlations were observed between specific frailty scales and regional brain volumes after FDR correction (Table 3). The CFS demonstrated significant negative correlations with global GMV (r = −0.430, FDR‐adjusted p = 0.019) and global WMV (r = −0.501, FDR‐adjusted p = 0.006), and was also significantly associated with cerebellar, hippocampal, and thalamic GMV, as well as temporal and parietal WMV, and myelin volume (all FDR‐adjusted p < 0.05).

TABLE 3.

Correlations between the frailty scales and global, regional GM and WM volumetry.

	CFS (FDR‐adjusted p‐values)	Fried scale (FDR‐adjusted p‐values)	FRAIL scale (FDR‐adjusted p‐values)	EFS (FDR‐adjusted p‐values)	SPPB (FDR‐adjusted p‐values)	MMSE (FDR‐adjusted p‐values)	MNA‐SF (FDR‐adjusted p‐values)
Global GMV, mL	−0.430 (0.019) *	−0.468 (0.034) *	−0.330 (0.227)	−0.440 (0.017) *	0.441 (0.037) *	0.303 (0.409)	−0.036 (0.927)
Frontal GMV, mL	−0.287 (0.113)	−0.355 (0.070)	−0.278 (0.285)	−0.338 (0.061)	0.349 (0.068)	0.316 (0.409)	0.016 (0.927)
Temporal GMV, mL	−0.350 (0.059)	−0.356 (0.070)	−0.245 (0.290)	−0.293 (0.101)	0.331 (0.081)	0.243 (0.499)	0.067 (0.927)
Parietal GMV, mL	−0.284 (0.113)	−0.234 (0.262)	−0.239 (0.290)	−0.218 (0.218)	0.406 (0.038) *	0.099 (0.720)	−0.026 (0.927)
Occipital GMV, mL	−0.254 (0.155)	−0.328 (0.096)	−0.198 (0.308)	−0.266 (0.134)	0.253 (0.181)	0.182 (0.591)	−0.136 (0.927)
Cerebellum GMV, mL	−0.582 (0.001) **	−0.380 (0.070)	−0.340 (0.227)	−0.589 (0.001) **	0.554 (0.004) **	0.367 (0.409)	−0.065 (0.927)
Hippocampus GMV, mL	−0.499 (0.006) **	−0.399 (0.070)	−0.340 (0.227)	−0.393 (0.029) *	0.550 (0.004) **	0.177 (0.591)	0.390 (0.305)
Thalamus GMV, mL	−0.449 (0.016) *	−0.465 (0.034) *	−0.324 (0.227)	−0.411 (0.023) *	0.323 (0.081)	0.278 (0.431)	0.145 (0.927)
Global WMV, mL	−0.501 (0.006) **	−0.222 (0.262)	−0.241 (0.290)	−0.604 (0.001) **	0.422 (0.038) *	0.112 (0.703)	0.140 (0.927)
Frontal WMV, mL	−0.298 (0.110)	−0.221 (0.262)	−0.102 (0.579)	−0.439 (0.017) *	0.159 (0.416)	0.146 (0.635)	−0.079 (0.927)
Temporal WMV, mL	−0.420 (0.019) *	−0.204 (0.291)	−0.210 (0.308)	−0.432 (0.017) *	0.416 (0.038) *	0.020 (0.905)	0.300 (0.643)
Parietal WMV, mL	−0.421 (0.019) *	−0.140 (0.434)	−0.103 (0.579)	−0.382 (0.032) *	0.195 (0.318)	−0.043 (0.861)	−0.017 (0.927)
Occipital WMV, mL	−0.346 (0.059)	−0.110 (0.516)	−0.227 (0.290)	−0.458 (0.016) *	0.319 (0.081)	−0.040 (0.861)	0.058 (0.927)
Cerebellum WMV, mL	0.095 (0.575)	0.174 (0.342)	0.150 (0.452)	−0.108 (0.524)	0.051 (0.762)	−0.078 (0.773)	0.222 (0.927)
Hippocampus WMV, mL	−0.220 (0.216)	−0.191 (0.309)	−0.071 (0.677)	−0.318 (0.077)	0.137 (0.459)	0.131 (0.662)	0.077 (0.927)
Thalamus WMV, mL	0.176 (0.315)	0.369 (0.070)	0.306 (0.237)	0.138 (0.438)	−0.133 (0.459)	−0.159 (0.627)	−0.156 (0.927)
CSF volume, mL	0.497 (0.006) **	0.394 (0.070)	0.233 (0.290)	0.471 (0.016) *	−0.355 (0.068)	−0.218 (0.499)	−0.085 (0.927)
Myelin volume, mL	−0.604 (0.001) **	−0.300 (0.128)	−0.205 (0.308)	−0.456 (0.016) *	0.389 (0.044) *	0.231 (0.499)	0.081 (0.927)

Abbreviations: CFS, clinical frailty scale; CSF, cerebrospinal fluid; EFS, Edmonton frailty scale; GMV, gray matter volume; MMSE, Mini‐Mental State Examination; MNA‐SF, short‐form mini‐nutritional assessment; SPPB, short physical performance battery; WMV, white matter volume.

^* p < 0.05.

^** p < 0.01.

^*** p < 0.001.

Similarly, the EFS showed significant negative correlations with both GMV and WMV across various regions (p < 0.05). The Fried scales and SPPB, while showing significant correlations with certain regions, exhibited weaker associations compared to the CFS and EFS (Figure 1).

FIGURE 1.

Correlation Analysis of Global and Regional Brain Volumes with Frailty Scales. The heatmap displays the partial correlation coefficients (r) between frailty scales and global/regional gray matter and white matter volumes, adjusted for age, gender, history of stroke, and total intracranial volume (ICV). The color scale represents the strength and direction of the correlations, ranging from −1 (blue, negative correlation) to +1 (red, positive correlation). Numeric values within the cells indicate the correlation coefficients. CFS, clinical frailty scale; CSF, cerebrospinal fluid; EFS, Edmonton frailty scale; GMV, gray matter volume; MMSE, Mini‐Mental State Examination; MNA‐SF, short‐form mini‐nutritional assessment; SPPB, short physical performance battery; WMV, white matter volume. *p < 0.05, **p < 0.01, ***p < 0.001.

3.4. Brain Relaxometry

3.4.1. Brain Relaxometry: Comparison Between the Frail and Non‐Frail Groups

Frail participants exhibited significantly higher T1 and T2 values in the hippocampus and thalamus compared to the non‐frail group (p < 0.05). Additionally, temporal lobe PD values were significantly higher in the frail group (p < 0.05). No significant differences were observed in global gray matter, global white matter, or other brain regions for T1, T2, or proton density (PD) values (p > 0.05) (Table 4).

TABLE 4.

Brain relaxometry comparison between frail and non‐frail groups.

	Total (n = 41)	Frail (n = 19)	Non‐frail (n = 22)	F value	p
Global GM, T1	1554.09 ± 77.83	1571.54 ± 78.23	1539.02 ± 76.03	0.58	0.566
Global GM, T2	126.96 ± 14.09	132.38 ± 12.72	122.28 ± 13.77	1.69	0.100
Global GM, PD	81.90 ± 0.75	82.04 ± 0.80	81.77 ± 0.70	0.41	0.682
Global WM, T1	1080.33 ± 61.92	1102.47 ± 64.14	1061.22 ± 54.32	1.61	0.116
Global WM, T2	92.81 ± 9.96	96.92 ± 11.91	89.26 ± 6.23	2.03	0.050
Global WM, PD	73.25 ± 1.43	73.61 ± 1.63	72.93 ± 1.17	1.01	0.321
Frontal, T1	1665.89 ± 99.69	1679.71 ± 111.01	1653.95 ± 89.70	0.44	0.661
Frontal, T2	157.01 ± 23.20	162.79 ± 23.54	152.02 ± 22.23	1.02	0.316
Frontal, PD	80.60 ± 1.37	80.80 ± 1.34	80.43 ± 1.41	0.64	0.527
Parietal, T1	1786.52 ± 162.87	1795.90 ± 168.81	1778.43 ± 161.10	−0.13	0.899
Parietal, T2	169.85 ± 39.62	175.09 ± 47.37	165.32 ± 31.94	0.25	0.801
Parietal, PD	82.47 ± 1.36	82.73 ± 1.51	82.25 ± 1.21	0.71	0.480
Temporal, T1	1606.83 ± 95.75	1640.11 ± 86.55	1578.10 ± 95.80	1.39	0.174
Temporal, T2	137.16 ± 19.42	144.63 ± 18.85	130.71 ± 17.87	1.91	0.064
Temporal, PD	80.62 ± 1.45	81.21 ± 1.15	80.10 ± 1.52	2.05	0.048 *
Occipital, T1	1412.36 ± 129.85	1427.67 ± 148.28	1399.14 ± 113.47	−0.12	0.908
Occipital, T2	104.90 ± 15.80	107.08 ± 18.82	103.02 ± 12.81	0.05	0.961
Occipital, PD	79.98 ± 1.58	80.23 ± 1.63	79.76 ± 1.53	0.02	0.987
Cerebellum, T1	1713.95 ± 101.28	1742.18 ± 95.23	1689.56 ± 102.07	1.02	0.314
Cerebellum, T2	144.84 ± 23.80	152.23 ± 22.15	138.45 ± 23.79	1.22	0.231
Cerebellum, PD	83.66 ± 0.99	83.78 ± 1.03	83.55 ± 0.96	−0.19	0.850
Hippocampus, T1	1714.21 ± 205.29	1816.07 ± 208.60	1626.24 ± 159.62	2.64	0.012 *
Hippocampus, T2	183.25 ± 65.18	217.39 ± 74.40	153.76 ± 37.18	2.96	0.005 **
Hippocampus, PD	82.38 ± 1.84	83.14 ± 1.78	81.73 ± 1.68	1.72	0.093
Thalamus, T1	1099.73 ± 94.87	1146.14 ± 88.16	1059.65 ± 82.76	2.65	0.012 *
Thalamus, T2	92.49 ± 12.91	98.20 ± 11.25	87.56 ± 12.42	2.04	0.048 *
Thalamus, PD	75.03 ± 1.73	75.64 ± 1.62	74.50 ± 1.68	1.75	0.089

Abbreviations: GM, gray matter; PD, proton density; WM, white matter.

^* p < 0.05.

^** p< 0.01.

3.5. Correlation Between Frailty Scales and Brain Relaxometry

After adjusting for age, gender, and history of stroke, partial correlation analysis with FDR correction revealed few significant associations between frailty scales and brain relaxometry values (Table 5). The CFS was significantly correlated with hippocampal T1 and T2 values, and the EFS showed significant positive correlations with global white matter T2 and hippocampal T2 values. The Short Physical Performance Battery (SPPB) was negatively correlated with hippocampal T1 and T2 values (FDR‐adjusted p < 0.05). No other significant correlation was observed (Figure 2).

TABLE 5.

Correlations between frailty scales and global, regional brain relaxometry.

	CFS (FDR‐adjusted p‐values)	Fried scale (FDR‐adjusted p‐values)	FRAIL scale (FDR‐adjusted p‐values)	EFS (FDR‐adjusted p‐values)	SPPB (FDR‐adjusted p‐values)	MMSE (FDR‐adjusted p‐values)	MNA‐SF (FDR‐adjusted p‐values)
Global GM T1	0.208 (0.334)	0.155 (0.731)	−0.023 (0.961)	0.080 (0.945)	−0.113 (0.922)	−0.173 (0.618)	0.055 (1.000)
Global GM T2	0.350 (0.092)	0.267 (0.532)	0.136 (0.948)	0.330 (0.192)	−0.245 (0.373)	−0.187 (0.599)	0.011 (1.000)
Global GM, PD	0.193 (0.350)	0.018 (0.916)	−0.109 (0.948)	−0.009 (0.995)	−0.101 (0.922)	−0.201 (0.599)	−0.073 (1.000)
Global WM, T1	0.366 (0.092)	0.179 (0.731)	0.167 (0.948)	0.257 (0.357)	−0.280 (0.267)	−0.390 (0.422)	0.107 (1.000)
Global WM, T2	0.376 (0.092)	0.258 (0.532)	0.285 (0.746)	0.479 (0.032) *	−0.387 (0.088)	−0.249 (0.567)	0.033 (1.000)
Global WM, PD	0.242 (0.275)	0.088 (0.853)	0.076 (0.948)	0.083 (0.945)	−0.110 (0.922)	−0.304 (0.567)	0.139 (1.000)
Frontal, T1	0.148 (0.487)	0.141 (0.731)	−0.033 (0.948)	−0.023 (0.995)	−0.004 (0.988)	−0.083 (0.836)	0.019 (1.000)
Frontal, T2	0.213 (0.334)	0.158 (0.731)	0.050 (0.948)	0.117 (0.814)	−0.022 (0.969)	−0.071 (0.836)	−0.013 (1.000)
Frontal, PD	0.141 (0.488)	0.056 (0.853)	−0.042 (0.948)	0.001 (0.995)	−0.022 (0.969)	−0.156 (0.628)	−0.135 (1.000)
Parietal, T1	0.075 (0.735)	0.064 (0.853)	−0.097 (0.948)	0.032 (0.995)	0.058 (0.969)	0.006 (0.973)	0.095 (1.000)
Parietal, T2	0.147 (0.487)	0.037 (0.858)	0.005 (0.977)	0.180 (0.538)	−0.042 (0.969)	0.073 (0.836)	0.174 (1.000)
Parietal, PD	0.193 (0.350)	0.064 (0.853)	−0.048 (0.948)	0.073 (0.945)	−0.055 (0.969)	−0.069 (0.836)	−0.034 (1.000)
Temporal, T1	0.337 (0.095)	0.328 (0.398)	0.135 (0.948)	0.191 (0.536)	−0.193 (0.555)	−0.241 (0.567)	−0.153 (1.000)
Temporal, T2	0.378 (0.092)	0.395 (0.382)	0.181 (0.948)	0.276 (0.357)	−0.203 (0.546)	−0.233 (0.567)	−0.208 (1.000)
Temporal, PD	0.363 (0.092)	0.163 (0.731)	0.098 (0.948)	0.229 (0.449)	−0.292 (0.252)	−0.228 (0.567)	−0.353 (0.796)
Occipital, T1	0.029 (0.875)	0.131 (0.731)	−0.034 (0.948)	−0.005 (0.995)	0.056 (0.969)	−0.115 (0.782)	0.076 (1.000)
Occipital, T2	0.049 (0.830)	0.052 (0.853)	−0.057 (0.948)	0.021 (0.995)	0.031 (0.969)	−0.044 (0.889)	0.059 (1.000)
Occipital, PD	−0.026 (0.875)	−0.038 (0.858)	−0.041 (0.948)	−0.029 (0.995)	0.025 (0.969)	0.022 (0.929)	0.000 (1.000)
Cerebellum, T1	0.293 (0.154)	0.144 (0.731)	0.052 (0.948)	0.161 (0.603)	−0.167 (0.656)	−0.246 (0.567)	0.065 (1.000)
Cerebellum, T2	0.221 (0.330)	0.197 (0.731)	0.033 (0.948)	0.188 (0.536)	−0.090 (0.936)	−0.159 (0.628)	−0.002 (1.000)
Cerebellum, PD	0.094 (0.676)	0.062 (0.853)	−0.007 (0.977)	0.009 (0.995)	−0.003 (0.988)	−0.041 (0.889)	−0.141 (1.000)
Hippocampus, T1	0.460 (0.049) *	0.263 (0.532)	0.308 (0.746)	0.386 (0.143)	−0.534 (0.014) *	−0.187 (0.599)	−0.248 (1.000)
Hippocampus, T2	0.495 (0.042) *	0.330 (0.398)	0.338 (0.746)	0.491 (0.032) *	−0.510 (0.014) *	−0.242 (0.567)	−0.135 (1.000)
Hippocampus, PD	0.348 (0.092)	0.165 (0.731)	0.164 (0.948)	0.259 (0.357)	−0.437 (0.051)	−0.185 (0.599)	−0.203 (1.000)
Thalamus, T1	0.416 (0.084)	0.134 (0.731)	0.248 (0.905)	0.373 (0.143)	−0.427 (0.051)	−0.131 (0.732)	−0.054 (1.000)
Thalamus, T2	0.345 (0.092)	0.102 (0.853)	0.171 (0.948)	0.354 (0.157)	−0.366 (0.108)	0.037 (0.889)	−0.055 (1.000)
Thalamus, PD	0.295 (0.154)	0.087 (0.853)	0.145 (0.948)	0.211 (0.501)	−0.304 (0.246)	−0.071 (0.836)	−0.043 (1.000)

Abbreviations: CSF, cerebrospinal fluid; EFS, Edmonton frailty scale; GM, gray matter; MMSE, Mini‐Mental State Examination; MNA‐SF, short‐form mini‐nutritional assessment; PD, proton density; SPPB, short physical performance battery; WM, white matter.

^* p < 0.05.

^** p < 0.01.

^*** p < 0.001.

FIGURE 2.

Correlation Analysis of Brain Relaxometry with Frailty Scales. The heatmap displays the partial correlation coefficients (r) between frailty scales and global/regional brain relaxometry values, adjusted for age, gender, and history of stroke. The color scale represents the strength and direction of the correlations, ranging from −1 (blue, negative correlation) to +1 (red, positive correlation). Numeric values within the cells indicate the correlation coefficients. (A) Correlations with T1 values; (B) Correlations with T2 values; (C) Correlations with PD values. CSF, cerebrospinal fluid; EFS, Edmonton Frailty Scale; MMSE, Mini‐Mental State Examination; MNA‐SF, short‐form mini‐nutritional assessment; SPPB, short physical performance battery; GM, gray matter; PD, proton density; WM, white matter. *p < 0.05, **p < 0.01, ***p < 0.001.

4. Discussion

This study demonstrated that frail individuals exhibited significant reductions in gray and white matter volumes in specific brain regions, primarily involving the thalamus, temporal lobe, parietal lobe, cerebellum, and hippocampus. In addition, relaxometry metrics in the hippocampus and thalamus showed significant correlations with frailty. These findings provide new insights into the structural and microstructural brain changes associated with frailty.

4.1. Brain Volume Reductions and Frailty

Gray and white matter volumes of specific regions were significantly reduced in frail participants, with prominent involvement of the thalamus, temporal lobe, parietal lobe, cerebellum, and hippocampus. These findings are consistent with previous studies demonstrating associations between frailty and brain volume loss as well as white matter integrity impairment, suggesting a potential neurostructural basis for frailty [17, 20, 21, 22, 23, 24, 25, 26, 27].

The strong correlations between CFS scores and both GM and WM volumes suggest that the CFS may be a particularly sensitive measure for capturing the impact of frailty on brain structural integrity. This finding is consistent with prior studies that have linked brain atrophy to frailty using various frailty scales [20, 25]. The observed reductions in brain volume highlight the importance of early detection and intervention in frail individuals to mitigate the associated cognitive and functional decline.

4.2. Brain Relaxometry and Frailty

In addition to volumetric changes, our study identified significant increases in T1 and T2 values, specifically in the hippocampus and thalamus of frail participants. These increases likely reflect increased tissue water content, microstructural disorganization, and disruptions in myelin or axonal integrity, providing complementary information to volumetric measures [19, 28]. Such microstructural alterations are consistent with age‐related neuroinflammatory processes, or “inflammaging,” which can lead to glial activation, edema, and subsequent neuronal dysfunction [29]. The strong associations between CFS scores and hippocampal and thalamic relaxometry parameters suggest that these regions are particularly vulnerable in frailty, as previously indicated by studies linking hippocampus and thalamus to both frailty and cognitive decline [30, 31].

4.3. Comparison of Frailty Scales With Brain Structure

In this study, both CFS and EFS showed significant correlations with brain volumetry, but with distinct regional patterns. EFS exhibited stronger associations with global GMV, global WMV, and cerebellar GMV, suggesting it may be more sensitive to overall brain structural integrity. In contrast, CFS demonstrated stronger correlations with hippocampal and thalamic GMV, as well as myelin volume, suggesting that limbic and thalamic structures are particularly vulnerable in individuals with higher CFS scores. These differences may reflect the multidimensional nature of frailty and the complementary value of these scales in capturing distinct neural substrates.

By contrast, the Fried scale and FRAIL scale showed weaker or non‐significant correlations with brain structure. These scales focus primarily on physical frailty, which may limit their ability to capture the full spectrum of frailty‐related neurobiological changes. Compared with solely physical assessment, CFS and EFS encompass multiple dimensions of frailty, including physical, cognitive, and nutritional functions [11, 12]. The stronger performance of CFS and EFS highlights the importance of using multidimensional frailty scales in studies investigating brain structural integrity.

4.4. Implications for Clinical Practice

Our findings suggest that neuroimaging, particularly synthetic MRI, may serve as a complementary tool for assessing brain structural changes associated with frailty. The robust correlations between multidimensional frailty scales and brain volumetry highlight the potential of synthetic MRI to provide objective biomarkers for frailty evaluation. While relaxometry findings were limited to specific regions, they suggest possible microstructural alterations in these areas associated with frailty.

4.5. Study Limitations

There are several limitations to this study. First, the relatively small sample size may limit the generalizability of our findings. Future research with larger cohorts is needed to validate these results. Additionally, the participants were recruited from hospitalized patients in the neurology ward, which may introduce selection bias, as their baseline brain structural integrity and frailty could be influenced by underlying neurological conditions. While this introduces selection bias compared to healthy community controls, this ‘symptomatic but not severe’ group actually holds greater clinical screening significance. These are precisely the individuals who are most likely to present to hospitals and benefit from opportunistic frailty screening in real‐world clinical practice. Finally, the cross‐sectional nature of this study precludes the establishment of causal relationships between brain structural changes and frailty. Longitudinal studies are necessary to clarify the temporal relationship between frailty and brain atrophy.

5. Conclusion

In conclusion, this study provides strong evidence of significant associations between frailty—particularly as measured by the CFS and EFS—and brain volumetry. Additionally, relaxometry metrics in the hippocampus and thalamus showed significant but region‐specific associations with frailty. These findings suggest that brain structural changes are closely tied to the severity of frailty, and that synthetic MRI may offer a promising tool for diagnosing and monitoring frailty in clinical settings. Further research is needed to explore the underlying neurobiological mechanisms and to assess the clinical utility of synthetic MRI in managing frailty in the elderly.

Author Contributions

Conceptualization: Yuhui Chen, Chunmei Li. Methodology: Yuhui Chen, Fan Feng. Software: Sicong Wang. Validation: Qi Wang, Hua Wang, Min Chen. Formal analysis: Yuhui Chen, Qi Wang, Sicong Wang. Investigation: Yuhui Chen, Fan Feng, Qi Wang. Data curation: Yuhui Chen, Min Chen. Writing – original draft: Yuhui Chen, Fan Feng. Writing – review and editing: Tao Gong, Chunmei Li. Visualization: Qi Wang. Supervision: Tao Gong, Chunmei Li. Project administration: Tao Gong. Funding acquisition: Yuhui Chen, Chunmei Li.

Funding

This research was funded by the National High‐Level Hospital Clinical Research Funding (BJ‐2019–200) and the National Natural Science Foundation of China (Grant no. 82071891).

Ethics Statement

This study was approved by the Institutional Review Board of Beijing Hospital (Approval no. 2018BJYYEC‐121‐02), and all participants provided written informed consent.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.