Multimodal magnetic resonance imaging (MRI) quantification of hippocampal injury in acute carbon monoxide poisoning: reduced hippocampal volume as a sensitive biomarker linked to cognitive impairment and prognosis

Introduction

Carbon monoxide (CO) poisoning is one of the most prevalent contributors to accidental poisoning on a global scale (1-3). As a critical structure for spatial orientation and memory, the hippocampus is also one of the cerebral regions earliest hit by acute carbon monoxide poisoning (ACMP)-induced pathological changes. A substantial proportion of ACMP patients develop memory- and orientation-related cognitive impairments (4-8). However, there remains a relative paucity of neuroimaging studies focusing on the hippocampus and its associated cognitive impairments during the acute phase of ACMP. Automated atlas-based magnetic resonance imaging (MRI) volumetry enables objective and reproducible quantification of regional brain volume changes. Among available tools, FreeSurfer is one of the most established and widely used software solutions for automated brain segmentation and volumetry, and it has been extensively applied in neurodegeneration research to provide robust volumetric estimates (9). This study aimed to evaluate structural and microcirculatory changes in the hippocampus of ACMP patients during the acute phase using multimodal MRI, with components such as T2 fluid-attenuated inversion recovery (T2-FLAIR), three-dimensional (3D) T1-weighted spoiled gradient-echo structural imaging (BRAVO on the GE platform), and 3D pseudocontinuous arterial spin labeling (3D PCASL), and analyze their correlation with cognitive function, with the objectives of deepening the understanding of neural mechanisms underlying ACMP-related acute cognitive impairment and providing imaging evidence to facilitate early clinical intervention. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2553/rc).

Methods

This study was designed as a nonrandomized prospective cohort study without any active interventions.

Study participants recruitment

A total of 63 patients with ACMP and 60 healthy controls (HC) were initially screened. In the ACMP group, four patients were excluded due to insufficient image quality, and two were excluded due to a history of old cerebral infarction. The remaining 57 ACMP patients were included in the final analysis and further classified, based on a 3-month clinical follow-up, into a delayed encephalopathy after acute carbon monoxide poisoning (DEACMP) group (n=15) and a non-DEACMP group (n=42). In the HC group, three participants were excluded due to insufficient image quality, resulting in 57 HC being included in the final analysis. Figure 1 illustrates the participant inclusion and exclusion process.

Figure 1 Flowchart illustrating participant inclusion and exclusion. A total of 63 patients with ACMP and 60 HCs were initially screened. In the ACMP group, four patients were excluded due to insufficient image quality, and two were excluded due to a history of old cerebral infarction. The remaining 57 ACMP patients were included in the final analysis and were further classified, based on a 3-month clinical follow-up, into the DEACMP group n=15 and the non-DEACMP group n=42. In the HC group, three participants were excluded due to insufficient image quality, resulting in 57 HCs included in the final analysis. ACMP, acute carbon monoxide poisoning; DEACMP, delayed encephalopathy after acute carbon monoxide poisoning; HC, healthy controls.

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of North Sichuan Medical College (No. NSMC2021-05). Informed consent was provided by all the participants.

ACMP group

ACMP patients who presented to Affiliated Hospital of North Sichuan Medical College were recruited, and all underwent MRI within 3 days of poisoning.

The inclusion criteria for the ACMP group were as follows:

• Diagnosis of ACMP was confirmed based on: a history of exposure to high-concentration CO; acute onset of central nervous system impairment symptoms and signs; timely measurement of blood carboxyhemoglobin (COHb) levels; on-site hygienic investigation and determination of ambient CO concentration; and exclusion of other potential etiologies.
• Onset of poisoning within 3 days before enrollment.
• Right-handedness.

The exclusion criteria for the ACMP group were as follows:

• Baseline MRI was performed in the acute stage; therefore, patients already in the DEACMP stage at the time of MRI were excluded.
• A history of encephalitis, brain trauma, cerebrovascular pathologies, metabolic disorders, epilepsy, psychiatric disorders, or other severe neurological symptoms/signs.
• Presence of severe systemic diseases.
• Prior dependence on alcohol or drugs.
• Patients who met the clinical definition of DEACMP at the time of baseline MRI acquisition.
• Pediatric patients (infants and young children).

In line with the above inclusion and exclusion criteria, the total number of ACMP patients was 57.

Control group

The control group was individually aligned with the ACMP group (n=57) at a 1:1 ratio, with 57 healthy volunteers finally included. The following key matching variables were selected: (I) sex; (II) age (±3 years); (III) years of education (≤2 years difference); and (IV) right-handedness. For each ACMP patient, controls were recruited in the order of patient enrollment, and their eligibility was confirmed by strictly adhering to the above criteria. Controls with systemic diseases or neurological symptoms/signs, obvious structural abnormalities on routine brain MRI, mental disorders, or alcohol or drug dependence were excluded.

DEACMP group and non-DEACMP group

All ACMP patients were followed clinically for more than 3 months without loss to follow-up, and no repeat MRI was acquired. Patients who, after regaining consciousness from coma, experienced a “latent period” of 2–60 days followed by recurrence of the following clinical manifestations were classified into the DEACMP group: disturbance of consciousness, abnormal mental behavior, pyramidal tract damage, extrapyramidal dysfunction, focal cerebral cortical dysfunction, and cranial/peripheral nerve injury-related brain dysfunction. During the 3-month follow-up, 15 ACMP patients developed DEACMP (thus included in the DEACMP group), and the remaining 42 patients with complete recovery were assigned to the non-DEACMP group.

Assessment of activities of cognitive function in ACMP group

To assess cognitive status, the Montreal Cognitive Assessment (MoCA) scale was used for all enrolled ACMP patients. Two subscores of the MoCA were extracted: the delayed recall subscore (total score: 5 points) and the orientation subscore (total score: 6 points).

MRI scanning protocol

All participants underwent MRI scanning upon admission using a GE MR750 3.0 T MRI scanner (GE Healthcare, Chicago, IL, USA) with a 32-channel head coil. The scanning protocol included axial T2-FLAIR, 3D T1-weighted structural imaging, and 3D PCASL sequence.

During scanning, participants’ heads were fixed with a soft pad, and earplugs were supplied to mitigate noise. All participants were required to remain awake throughout the scanning process.

AX T2-FLAIR sequence parameters

The scanning range covered the entire hippocampal structure. The sagittal localizer was aligned parallel to the long axis of the hippocampus, and the coronal localizer was parallel to the line connecting the bilateral temporal lobes. Repetition time (TR) =7,975 ms, echo time (TE) =150 ms, field of view (FOV) =24×24 cm, slice thickness =3.0 mm, matrix =256×256.

3D T1-weighted structural imaging parameters

TR =8 ms, TE =3 ms, FOV =24 × 24 cm, slice thickness =1 mm, flip angle =11°, voxel size =0.47 mm × 0.47 mm × 1.00 mm, matrix =256×256, with 140 slices acquired per participant.

3D PCASL sequence parameters

3D spiral k-space filling-integrated fast spin-echo (FSE) acquisition, requirement of M0, background signal suppression, and 3D spiral FSE readout mode, acceleration factor =8, effective resolution =3.75×3.75×4 mm, data points =512, TR =4,632 ms, TE =10.5 ms, FOV =24×24 cm, slice thickness =4.0 mm, bandwidth =62.5 kHz, number of excitations (NEX) =3, post-label delay (PLD) =2,025 ms, axial slices =36. Cerebral blood flow (CBF) data automatically generated during 3D PCASL scanning were used for subsequent preprocessing.

Data processing and statistical calculation

All quantitative imaging analyses—including hippocampal volumetry, CBF, and T2-FLAIR-based hippocampal signal intensity measurements—were performed using the baseline MRI acquired in the acute stage (within 3 days after CO poisoning). The 3-month follow-up was clinical and used for prognosis classification (DEACMP vs. non-DEACMP), and no follow-up MRI was obtained for this study.

Preprocessing of T1-weighted 3D images

FreeSurfer version 7.0 (http://surfer.nmr.mgh.harvard.edu/fswiki/) was used to preprocess T1-weighted structural images using the standard recon-all pipeline, including motion correction, removal of non-brain tissue, Talairach registration, intensity normalization, and segmentation of cortical and subcortical structures. Hippocampal segmentation was subsequently performed using FreeSurfer’s HippocampalSubfieldsAndNucleiOfAmygdala module (implemented via the segmentHA_T1.sh script), which applies a Bayesian, atlas-based approach derived from ultra-high-resolution ex vivo MRI. The original hippocampal subfield labels generated by FreeSurfer were then merged into three longitudinal subregions—head, body, and tail—according to the standard head-body-tail (HBT) parcellation scheme, as shown in Figure 2A. Intracranial volume (ICV) was also extracted from FreeSurfer outputs. This segmentation framework has been extensively validated and widely used in neuroimaging studies (10,11).

Figure 2 Illustration of multimodal MRI processing at the hippocampal level in a representative patient with acute carbon monoxide poisoning. (A) Three-dimensional reconstruction of the hippocampus from T1-weighted structural MRI following FreeSurfer segmentation, with hippocampal subfield labels merged into head, body, and tail regions according to the HBT parcellation scheme. (B) Axial T2-weighted FLAIR image demonstrating hippocampal regions of interest used for signal intensity measurement. (C) Multiplanar structural MRI views illustrating bilateral hippocampal localization used for volumetric and perfusion analyses. This figure illustrates the processing workflow by which hippocampal volume, signal intensity, and CBF metrics were derived from multimodal MRI in individual patients. CBF, cerebral blood flow; FLAIR, fluid-attenuated inversion recovery; HBT, head-body-tail; MRI, magnetic resonance imaging.

All FreeSurfer-derived hippocampal segmentations were subjected to visual quality control by an experienced neuroradiologist following standard FreeSurfer guidelines. Segmentations were inspected for boundary mislabeling, over- or under-segmentation, and topological defects. Scans with evident segmentation errors that could not be resolved by reprocessing were excluded from subsequent analyses. Only segmentations passing visual quality control were included.

Measurement of hippocampal signal intensity on T2-FLAIR

Bilateral hippocampal signal intensities on T2-FLAIR images were measured using RadiAnt DICOM Viewer software (https://www.radiantviewer.com/). For each hippocampus, three regions of interest (ROIs) were manually delineated avoiding cerebrospinal fluid in the hippocampal fissure and cystic lesions, and the mean signal intensity was calculated. To standardize measurements across participants, the mean signal intensity of the hippocampus was normalized by dividing it by the signal intensity measured in an ipsilateral extracranial background region, as shown in Figure 2B.

All measurements were performed independently by two experienced radiologists, and the average value of their results was used as the final data.

Extraction of hippocampal CBF

First, Digital Imaging and Communications in Medicine (DICOM) files of CBF data underwent conversion to Neuroimaging Informatics Technology Initiative (NIfTI) format via the dcm2nii software (https://people.cas.sc.edu/rorden/mricron/index.html). Preprocessing and statistical analysis were then performed using the Statistical Parametric Mapping 8 (SPM8) toolbox (https://www.fil.ion.ucl.ac.uk/spm) on the MATLAB 2013b platform (MathWorks, Inc., Natick, MA, USA). After preprocessing, hippocampal CBF values were extracted using SPM8, DPABI software (12), and MRIcro software on the MATLAB platform.

The segmentation of hippocampal subfield volumes, measurement of hippocampal T2-FLAIR signals, and the AAL template used for hippocampal CBF extraction are shown in Figure 2C. Preprocessing of ASL-derived CBF data followed a standard pipeline including motion correction, spatial normalization to Montreal Neurological Institute (MNI) space, global mean CBF normalization, and spatial smoothing. Hippocampal CBF values were then extracted using SPM8, DPABI, and MRIcro software. Detailed preprocessing steps and parameters are provided in Appendix 1.

Outcome measures

The primary outcome measures were as follows: (I) acute-phase hippocampal volume changes (total and subregional volumes, measured by 3D T1-weighted structural imaging); (II) occurrence of DEACMP during 3-month follow-up; (III) cognitive function assessed by MoCA delayed recall and orientation subscores.

The secondary outcome measures were as follows: (I) hippocampal T2-FLAIR signal intensity (normalized by ipsilateral extracranial blank area signal); (II) hippocampal CBF measured by 3D PCASL sequence.

Statistical analysis

The software SPSS 19.0 (IBM Corp., Armonk, NY, USA) was used for statistical analysis of demographic and clinical data. The continuous variables’ normality was tested using the Kolmogorov-Smirnov test. They were presented as mean ± standard deviation if normally distributed, and as median (interquartile range) if not. Two-group comparisons of normally distributed continuous variables were conducted with independent samples t-tests. For non-normally distributed continuous variables with unequal variances, group comparisons employed the nonparametric Mann-Whitney U test. The Chi-square test was used for categorical variable comparisons. A two-tailed P value <0.05 indicated statistical significance. Detailed demographic and clinical data appear in Table 1.

One-way analysis of covariance (ANCOVA) was utilized for comparing hippocampal volume, signal intensity, and CBF among the HC, DEACMP, and non-DEACMP groups (after testing for normality and homogeneity of variances). Covariates included sex, age, years of education, and ICV for total and subregional (head, body, tail) hippocampal volume analyses; for signal intensity and CBF comparisons, covariates were sex, age, and years of education. Overall group differences were assessed using ANCOVA, followed by post hoc pairwise comparisons with Bonferroni correction (P<0.05). Given multiple volumetric outcomes, whole-hippocampus volumes (left and right whole hippocampus) were prespecified as the primary volumetric outcomes. Hippocampal subregional volumes (head, body, and tail) were analyzed as exploratory measures to localize potential effects. We conducted partial correlation analysis to examine associations between hippocampal imaging indices (volume, signal intensity, CBF) and cognitive function (MoCA total score, delayed recall, orientation). When analyzing the correlation between hippocampal volume and cognitive scores, covariates were sex, age, years of education, and ICV. For other correlations (signal intensity/CBF with cognitive scores), only sex, age, and years of education were adjusted. To control for multiple testing in the partial correlation analyses, we applied the Benjamini-Hochberg false discovery rate (BH-FDR) procedure within each cognitive outcome column (MoCA total score, delayed recall subscore, and orientation subscore). For each outcome, 13 correlations were tested (bilateral whole hippocampal volume and left/right whole, head, body, tail volumes, hippocampal CBF, and T2-FLAIR signal intensity), and FDR-adjusted P values (P-FDR) < 0.05 were considered statistically significant.

COHb and arterial blood gas (ABG)-derived bicarbonate (HCO₃⁻) were summarized and compared between prognosis groups as ancillary clinical variables; they were not included as primary predictors in the imaging–cognition association models.

Results

Clinical and demographic variables

A total of 57 participants were enrolled in the ACMP group, and 57 were assigned to the HC group. There were no statistically significant differences in sex, age, or educational years between the ACMP group and the HC group (P=0.178, 0.058, and 0.587, respectively).

All ACMP patients completed >3 months of follow-up without loss to follow-up; 15 developed DEACMP (DEACMP group) and 42 fully recovered (non-DEACMP group). Between the DEACMP group and the non-DEACMP group: no significant differences were observed in terms of sex or educational years. Compared with patients in the non-DEACMP group, those in the DEACMP group were significantly older. COHb and ABG HCO₃⁻ were collected for all patients. No significant differences in COHb or HCO₃⁻ were observed between the DEACMP and non-DEACMP groups. The DEACMP group had significantly lower MoCA total scores, delayed recall subscores, and orientation subscores compared with the non-DEACMP group. Detailed results are presented in Table 1.

All hippocampal volumetric findings reported below were derived from baseline acute-stage MRI (within 3 days); prognosis groups were defined based on clinical outcomes observed during the 3-month follow-up.

Analysis of total hippocampal volume and subregional volumes

ANCOVA (adjusted with consideration of sex, age, educational duration, and ICV) revealed the following:

• Total hippocampal volume: a significant difference in total right hippocampal volume was observed among the HC, DEACMP, and non-DEACMP groups (F=6.374, P=0.003), whereas no significant difference was noted in total left hippocampal volume across the three groups.
• Hippocampal subregional volumes: significant differences were detected for the left hippocampal body (F=3.769, P=0.028), right hippocampal head (F=3.459, P=0.024), and right hippocampal body (F=12.586, P<0.001). No significant differences were observed in the left hippocampal head or bilateral hippocampal tails. Detailed results are presented in Table 2.

Post hoc pairwise comparisons further showed:

• Right hippocampal head volume: the DEACMP group showed a significantly lower volume than the control group (P=0.019; Figure 3A);
• Right hippocampal body volume: both the non-DEACMP group and the DEACMP group showed significantly lower volumes than the control group (both P<0.001), and the DEACMP group additionally showed a significantly lower volume than the non-DEACMP group (P=0.044; Figure 3B);
• Total right hippocampal volume: both the non-DEACMP group and the DEACMP group showed significantly lower volumes than the control group (P=0.011 and P=0.004, respectively; Figure 3C);
• Left hippocampal body volume: both the non-DEACMP group and the DEACMP group showed significantly lower volumes than the control group (P=0.038 and P=0.040, respectively; Figure 3D).

Figure 3 Group differences in hippocampal volumes. Volumetric measures were derived from baseline acute-stage MRI. Values are mean ± SD (HC, n=57; non-DEACMP, n=42; DEACMP, n=15). (A) Right hippocampal head volume. (B) Right hippocampal body volume. (C) Right whole hippocampal volume. (D) Left hippocampal body volume. Group differences were tested using ANCOVA controlling for age, sex, education years, and intracranial volume, with Bonferroni-corrected post hoc comparisons (P values shown). ANCOVA, analysis of covariance; DEACMP, delayed encephalopathy after acute carbon monoxide poisoning; HC, healthy controls; MRI, magnetic resonance imaging; SD, standard deviation.

Detailed post hoc results are presented in Table 3.

For significant findings in the ANCOVA models, effect sizes are reported as partial eta-squared (ηp²), with values of 0.01, 0.06, and 0.14 considered small, medium, and large, respectively. For post hoc comparisons, Cohen’s d is reported, with values of 0.2, 0.5, and 0.8 considered small, medium, and large, respectively.

Analysis of hippocampal signal intensity on T2-FLAIR

Subsequent to controlling for sex, age, and the number of years of education, ANCOVA showed no significant differences in bilateral hippocampal signal intensity on T2-FLAIR among the HC, DEACMP, and non-DEACMP groups (all P>0.05). Detailed results are presented in Table 4.

Analysis of hippocampal CBF

After adjusting for sex, age, and educational duration, ANCOVA revealed no significant differences in bilateral hippocampal CBF among the HC, DEACMP, and non-DEACMP groups (P>0.05). Table 5 provides details.

Correlations of hippocampal imaging indices with cognitive function

Partial correlation analyses adjusted for sex, age, educational years, and ICV showed that right whole, head, and body hippocampal volumes were positively correlated with the delayed recall subscore (all FDR-corrected P<0.05). In addition, bilateral whole hippocampal volume and bilateral hippocampal body volume were positively correlated with the orientation subscore (all FDR-corrected P<0.05). The participant-level association between bilateral whole hippocampal volume and orientation performance is shown in Figure S1. No significant correlations were found between hippocampal volume and MoCA total score after FDR correction. In addition, no significant correlations were detected between bilateral hippocampal T2-FLAIR signal intensity or CBF and any cognitive function indices (MoCA total score, delayed recall, or orientation) after adjustment for sex, age, and years of education (all FDR-corrected P>0.05). Detailed results are presented in Table 6.

Discussion

This study explored the changing patterns of hippocampal structure and perfusion in patients with ACMP during the acute phase and their relationship with cognitive prognosis, using multimodal MRI techniques. The main findings are as follows:

Significant reductions in hippocampal volume were observed in the entire right hippocampus as well as in partial hippocampal subregions of both hemispheres. Notably, the reduction in right hippocampal body volume was more pronounced in patients who subsequently developed DEACMP.

A significant correlation was identified between reduced hippocampal volume and impairments in delayed recall and orientation.

No significant intergroup differences in hippocampal T2-FLAIR signal intensity or CBF were detected during the acute phase.

These results suggest that early hippocampal structural damage, particularly in the right hippocampal body, might represent a key neuroimaging marker associated with cognitive dysfunction following ACMP and with the subsequent occurrence of DEACMP.

This study found that the hippocampus is extremely sensitive to the hypoxic toxicity of CO, and reduced hippocampal volume can be measurable on MRI acquired in the acute stage (within 3 days); however, given this acute timeframe, the finding should be interpreted as an early volume difference rather than definitive evidence of de novo atrophy developing within days. The hippocampus, a key brain structure, performs a crucial role in information storage and memory (13). Acute CO exposure induces neuronal apoptosis (14,15) and significantly inhibits hippocampal neurogenesis (16), both of which play a role in reducing hippocampal volume. From our results, the damage to the right hippocampus was more severe than that to the left, indicating that hippocampal damage induced by ACMP exhibits asymmetry. This finding aligns with previous reports showing that hippocampal abnormalities exhibit laterality—specifically, a significant reduction in the degree centrality (DC) of the right hippocampus has been observed in DEACMP patients. This observation indicates a disruption in the functional interactions of the hippocampus as a node within the brain network, leading to a diminished influence of hippocampal network interactions on other components within the brain network (17). Importantly, laterality of volume reduction does not necessarily imply that the right hippocampus is uniquely functionally indispensable for cognition. Evidence from epilepsy research indicates that unilateral mesial temporal interventions (including selective amygdalohippocampectomy and minimally invasive laser interstitial thermal therapy targeting the hippocampal region) are clinically performed and can be accompanied by network-level reorganization and heterogeneous structural effects beyond the ablated tissue (18). In addition, high-resolution volumetry studies in non-lesional epilepsy demonstrate that subtle, lateralized volume alterations (including hippocampal-region/subfield findings) can be detected even when no overt lesion is visible on routine MRI, supporting the concept that laterality may reflect regional vulnerability rather than a simple “critical versus non-critical” functional dichotomy (19). Therefore, our right-predominant hippocampal volume reduction in ACMP may reflect selective vulnerability and/or reserve differences of the right mesial temporal system to CO-related hypoxic injury, rather than implying that right hippocampal tissue alone determines cognitive outcome. We propose that reduced hippocampal volume may reflect pre-existing structural vulnerability and/or early injury-related influences, which could contribute to the prominent amnesia and disorientation observed in ACMP patients. Longitudinal imaging is warranted to clarify temporal trajectories and causality.

In acute ischemic events, an initial volume increase is frequently observed during the first days due to cytotoxic and vasogenic edema, and apoptosis may also be accompanied by transient tissue expansion rather than immediate macroscopic volume loss. Therefore, the apparent discrepancy highlights the need for cautious interpretation of acute-stage volumetry. In our cohort, the observed reduced hippocampal volume measured in the acute stage may reflect pre-existing smaller hippocampal volume (structural vulnerability/reduced reserve) and/or acute-stage influences on MRI-derived volumetry (e.g., tissue water/signal alterations, segmentation uncertainty, and partial-volume effects), especially in a small structure with complex boundaries such as the hippocampus. Longitudinal imaging at multiple time points is warranted to determine the true trajectories of hippocampal structural change after ACMP.

In the present study, a quantitative analysis of the volumes of hippocampal subregions was conducted. The results demonstrated that the hippocampal body is the most severely affected subregion, which indicates that the neurotoxic effect of CO possesses distinct regional selectivity. The hippocampus is a functionally heterogeneous region, with distinct functional subregions that exhibit differential susceptibility to pathology (20-25). To date, structural imaging studies that partition the hippocampus into the head, body, and tail have reported associations between alterations in hippocampal subregional volume and specific pathology-related issues (13,26,27). Previous studies have reported an association between reduced volume of the right hippocampal body and an increased likelihood of Parkinson’s disease (28). Patients with subjective memory complaints (SMC) exhibit decreased fractional anisotropy (FA) and increased mean diffusivity (MD) within the hippocampal body, indicating the presence of microstructural alterations (29). Subregional analysis can provide more precise localization information and higher sensitivity, thereby serving as a reference for accurate diagnosis.

Our partial correlation analysis further confirmed specific associations between hippocampal structural measurements and clinical cognitive performance. Delayed recall was primarily correlated with structural measures of the right hippocampus, including right whole, head, and body volumes. In contrast, orientation performance was associated with bilateral whole hippocampal volume as well as bilateral hippocampal body volume. These findings suggest that different aspects of cognitive dysfunction after ACMP may relate to distinct, although partially overlapping, hippocampal structural patterns. Reduced hippocampal volume has been associated with memory disorders and other cognitive deficit (28,30,31). Researchers have identified an association between memory impairment and hippocampal volumetric measures in patients with CO intoxication through longitudinal assessments (31). In addition, a sequence of structural and functional studies have indicated that the anterior body of the hippocampus may serve as a critical hippocampal hub for cognition related to scenes (32). Therefore, structural damage in specific hippocampal subregions may act as a potential biomarker to evaluate cognitive performance of ACMP patients, and represents a potential imaging target for effective management and intervention at the early phase of the illness. Although reduced hippocampal body volume showed significant group differences and was associated with subsequent development of DEACMP, formal predictive modeling was not performed in this study. Therefore, the present findings should not be interpreted as establishing individual-level prediction or validated clinical risk stratification. Recent studies emphasize that robust prediction from neuroimaging requires dedicated modeling frameworks, adequate sample sizes, and external validation (33,34). Accordingly, future studies with larger, multicenter cohorts and longitudinal imaging are warranted to integrate hippocampal structural measures into predictive models of DEACMP risk.

No notable disparities in hippocampal T2-FLAIR signal or CBF were observed among the three groups in this study. It should be noted that, in the acute stage, the absence of clear T2-FLAIR hyperintensity does not exclude microstructural injury, and MRI-derived volumetry can be influenced by acute physiological changes and segmentation/partial-volume effects; therefore, acute-stage volume estimates should be interpreted cautiously. Meanwhile, no intergroup difference was found in hippocampal perfusion levels, indicating that there may be no extensive, persistent hypoperfusion in this region that can be captured by 3D PCASL technology. Therefore, for the detection of early hippocampal injury, structural volume analysis based on T1-weighted images may be more sensitive than analysis of conventional signal intensity or perfusion imaging.

This study has several limitations. First, this was a single-center cross-sectional study with a relatively small sample size. Second, MRI was acquired in the acute stage (within 3 days after CO poisoning), during which brain volume estimates may be influenced by acute physiological changes (e.g., edema) as well as image-derived segmentation and partial-volume effects; therefore, our findings should be interpreted as reduced hippocampal volume measured in the acute stage, rather than definitive evidence of de novo atrophy developing within a few days. Third, because pre-exposure MRI was unavailable, we cannot exclude the possibility that reduced hippocampal volume predated CO poisoning and represented a structural vulnerability factor with reduced compensatory capacity. Fourth, the lack of longitudinal follow-up structural MRI at multiple post-poisoning time points prevents delineation of the dynamic evolution of hippocampal injury; consequently, we cannot determine whether hippocampal volume demonstrates progressive volumetric decline over time, particularly in patients who develop DEACMP, nor can we relate volumetric trajectories to the latent period and subsequent deterioration. In addition, the DEACMP and non-DEACMP groups differed significantly in age, and age is closely related to hippocampal volume and cognitive performance; thus, residual confounding may remain despite covariate adjustment. Furthermore, subregional volumetric analyses were exploratory and involved multiple tests, and although BH-FDR correction was applied for the imaging-cognition correlation analyses, the number of tests and the relatively small sample size may still have yielded chance findings; therefore, independent validation in larger, ideally longitudinal cohorts is warranted. Finally, although COHb and ABG-derived HCO₃⁻ were collected for all patients, these measures can be influenced by clinical management (e.g., oxygen therapy) and reflect systemic status at the time of sampling; accordingly, they were reported as ancillary clinical context and were not included as primary predictors in the imaging-cognition association models, which focused on neuroimaging markers. Future research may focus on the following directions: (I) increasing the sample size and conducting multi-center studies; (II) conducting prospective longitudinal studies with repeated structural MRI at multiple post-poisoning time points to characterize hippocampal volumetric trajectories and determine whether patients who develop DEACMP exhibit progressive hippocampal volume decline, and to relate these trajectories to cognitive outcomes; and (III) achieving a more thorough understanding of how the hippocampus functions in the pathophysiological changes of ACMP from perspectives such as network connectivity.

Conclusions

This study indicated that in the acute phase of ACMP, reduced right hippocampal body volume is already detectable, and this early volume reduction is associated with long-term cognitive outcomes. The injury in this subregion acts as a key imaging index for the early evaluation of cognitive impairment following ACMP and may help identify patients at increased risk for unfavorable neurological outcomes.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2553/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2553/dss

Funding: This work was supported by the Doctoral Research Start-up Fund Project of Affiliated Hospital of North Sichuan Medical College (No. 2023-2GC009), the School Level Program of North Sichuan Medical College (No. CBY21-QA43), the Project of Taking on the Challenge and Leading the Way of Affiliated Hospital of North Sichuan Medical College (No. 2022JB001), the Natural Science Foundation of Sichuan Province (No. 2024NSFSC1791), and the Special Project for the Construction of Provincial-level Platform under the Special Program of Science and Technology Strategic Cooperation between the City and the University (No. 20SXPTJS0001).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1-2553/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of North Sichuan Medical College (No. NSMC2021-05). Informed consent was taken from all the participants.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Savioli G, Gri N, Ceresa IF, Piccioni A, Zanza C, Longhitano Y, Ricevuti G, Daccò M, Esposito C, Candura SM. Carbon Monoxide Poisoning: From Occupational Health to Emergency Medicine. J Clin Med 2024;13:2466. [Crossref] [PubMed]
2. Weaver LK. Carbon monoxide poisoning. Undersea Hyperb Med 2020;47:151-69. [Crossref] [PubMed]
3. Dent MR, Rose JJ, Tejero J, Gladwin MT. Carbon Monoxide Poisoning: From Microbes to Therapeutics. Annu Rev Med 2024;75:337-51. [Crossref] [PubMed]
4. Li W, Meng J, Lei J, Li C, Yue W. Carbon monoxide poisoning with hippocampi lesions on MRI: cases report and literature review. BMC Neurol 2024;24:159. [Crossref] [PubMed]
5. Kim DM, Lee IH, Park JY, Hwang SB, Yoo DS, Song CJ. Acute carbon monoxide poisoning: MR imaging findings with clinical correlation. Diagn Interv Imaging 2017;98:299-306. [Crossref] [PubMed]
6. Jiang W, Wu Q, Zhou C, Zhao Z, Tan Y. Gray matter nuclei damage in acute carbon monoxide intoxication assessed in vivo using diffusion tensor MR imaging. Radiol Med 2020;125:80-6. [Crossref] [PubMed]
7. Nañagas KA, Penfound SJ, Kao LW. Carbon Monoxide Toxicity. Emerg Med Clin North Am 2022;40:283-312. [Crossref] [PubMed]
8. Gale SD, Hopkins RO, Weaver LK, Bigler ED, Booth EJ, Blatter DD. MRI, quantitative MRI, SPECT, and neuropsychological findings following carbon monoxide poisoning. Brain Inj 1999;13:229-43. [Crossref] [PubMed]
9. Khadhraoui E, Nickl-Jockschat T, Henkes H, Behme D, Müller SJ. Automated brain segmentation and volumetry in dementia diagnostics: a narrative review with emphasis on FreeSurfer. Front Aging Neurosci 2024;16:1459652. [Crossref] [PubMed]
10. Iglesias JE, Augustinack JC, Nguyen K, Player CM, Player A, Wright M, Roy N, Frosch MP, McKee AC, Wald LL, Fischl B, Van Leemput KAlzheimer's Disease Neuroimaging Initiative. A computational atlas of the hippocampal formation using ex vivo, ultra-high resolution MRI: Application to adaptive segmentation of in vivo MRI. Neuroimage 2015;115:117-37. [Crossref] [PubMed]
11. Schwarzer J, Khadhraoui E, Einspänner E, Kukhlenko O, Behme D, Büntjen L, Schmitt FC, Müller SJ. Comparison of brain volumetry in patients with non-lesional epilepsy on 3 and 7 T MPRAGE. Quant Imaging Med Surg 2025;15:11408-26. [Crossref] [PubMed]
12. Yan CG, Wang XD, Zuo XN, Zang YF. DPABI: Data Processing & Analysis for (Resting-State) Brain Imaging. Neuroinformatics 2016;14:339-51. [Crossref] [PubMed]
13. Tzilivaki A, Tukker JJ, Maier N, Poirazi P, Sammons RP, Schmitz D. Hippocampal GABAergic interneurons and memory. Neuron 2023;111:3154-75. [Crossref] [PubMed]
14. Zhou XD, Wang JL, Guo DD, Jiang WW, Li ZK, Wang L, Zou Y, Bi MJ, Li Q. Neuroprotective effect of targeted regulatory Nrf2 gene on rats with acute brain injury induced by carbon monoxide poisoning. Environ Toxicol 2021;36:1742-57. [Crossref] [PubMed]
15. Guan L, Zhang YL, Wen T, Wang XF, Zhu MX, Zhao JY. Dynamic changes of heme oxygenase-1 in the hippocampus of rats after acute carbon monoxide poisoning. Arch Environ Contam Toxicol 2011;60:165-72. [Crossref] [PubMed]
16. Liu WC, Yang SN, Wu CW, Chen LW, Chan JY. Hyperbaric Oxygen Therapy Alleviates Carbon Monoxide Poisoning-Induced Delayed Memory Impairment by Preserving Brain-Derived Neurotrophic Factor-Dependent Hippocampal Neurogenesis. Crit Care Med 2016;44:e25-39. [Crossref] [PubMed]
17. Wu K, Liu M, He L, Tan Y. Abnormal degree centrality in delayed encephalopathy after carbon monoxide poisoning: a resting-state fMRI study. Neuroradiology 2020;62:609-16. [Crossref] [PubMed]
18. Müller SJ, Khadhraoui E, Kukhlenko O, Schwarzer J, Voges J, Sandalcioglu IE, Behme D, Schmitt F, Büntjen L. Brain Volume Loss After Stereotactic Laser Interstitial Thermal Therapy in Patients With Temporal Lobe Epilepsy. J Neuroimaging 2025;35:e70039. [Crossref] [PubMed]
19. Schwarzer J, Khadhraoui E, Einspänner E, Kukhlenko O, Behme D, Büntjen L, Schmitt FC, Müller SJ. Local brain volume reductions in patients with non-lesional epilepsy on 7T MRI. Neuroradiology 2025; Epub ahead of print. [Crossref]
20. Daugherty AM, Yu Q, Flinn R, Ofen N. A reliable and valid method for manual demarcation of hippocampal head, body, and tail. Int J Dev Neurosci 2015;41:115-22. [Crossref] [PubMed]
21. Pronier É, Morici JF, Girardeau G. The role of the hippocampus in the consolidation of emotional memories during sleep. Trends Neurosci 2023;46:912-25. [Crossref] [PubMed]
22. Vanrobaeys Y, Peterson ZJ, Walsh EN, Chatterjee S, Lin LC, Lyons LC, Nickl-Jockschat T, Abel T. Spatial transcriptomics reveals unique gene expression changes in different brain regions after sleep deprivation. Nat Commun 2023;14:7095. [Crossref] [PubMed]
23. Oliva A, Fernandez-Ruiz A, Karaba LA. CA2 orchestrates hippocampal network dynamics. Hippocampus 2023;33:241-51. [Crossref] [PubMed]
24. Sepahvand T, Carew SJ, Yuan Q. The ventral hippocampus is activated in olfactory but not auditory threat memory. Front Neural Circuits 2024;18:1371130. [Crossref] [PubMed]
25. Karat BG, DeKraker J, Hussain U, Köhler S, Khan AR. Mapping the macrostructure and microstructure of the in vivo human hippocampus using diffusion MRI. Hum Brain Mapp 2023;44:5485-503. [Crossref] [PubMed]
26. Meier TB, Savitz J, España LY, Goeckner BD, Kent Teague T, van der Horn HJ, Tugan Muftuler L, Mayer AR, Brett BL. Association of concussion history with psychiatric symptoms, limbic system structure, and kynurenine pathway metabolites in healthy, collegiate-aged athletes. Brain Behav Immun 2025;123:619-30. [Crossref] [PubMed]
27. Allebone J, Kanaan R, Maller J, O'Brien T, Mullen SA, Cook M, Adams SJ, Vogrin S, Vaughan DN, Connelly A, Kwan P, Berkovic SF, D'Souza WJ, Jackson G, Velakoulis D, Wilson SJ. Bilateral volume reduction in posterior hippocampus in psychosis of epilepsy. J Neurol Neurosurg Psychiatry 2019;90:688-94. [Crossref] [PubMed]
28. Ning C, Jin M, Cai Y, Fan L, Hu K, Lu Z, et al. Genetic architectures of the human hippocampus and those involved in neuropsychiatric traits. BMC Med 2024;22:456. [Crossref] [PubMed]
29. Ryu SY, Lim EY, Na S, Shim YS, Cho JH, Yoon B, Hong YJ, Yang DW. Hippocampal and entorhinal structures in subjective memory impairment: a combined MRI volumetric and DTI study. Int Psychogeriatr 2017;29:785-92. [Crossref] [PubMed]
30. DeMaster D, Johnson C, Juranek J, Ewing-Cobbs L. Memory and the hippocampal formation following pediatric traumatic brain injury. Brain Behav 2017;7:e00832.
31. Tamura T, Sugihara G, Takahashi H. Memory Impairment and Hippocampal Volume after Carbon Monoxide Poisoning. Arch Clin Neuropsychol 2021;36:145-8. [Crossref] [PubMed]
32. Leelaarporn P, Dalton MA, Stirnberg R, Stöcker T, Spottke A, Schneider A, McCormick C. Hippocampal subfields and their neocortical interactions during autobiographical memory. Imaging Neurosci (Camb) 2024;2:imag-2-00105.
33. Wang Y, Wang Q, Zhou M, Liang J, You L, Asken B, Zhou X, Song Q. Integration of Genetic and Imaging Data for Alzheimer's Disease Diagnosis and Interpretation. Adv Sci (Weinh) 2025;12:e07629. [Crossref] [PubMed]
34. Wang Q, Chen WJ, Su J, Wang G, Song Q. HECLIP: histology-enhanced contrastive learning for imputation of transcriptomics profiles. Bioinformatics 2025;41:btaf363. [Crossref] [PubMed]