Abstract
Neuroimaging has revealed that migraine is linked to alterations in both the structure and function of the brain. However, the re-lationship of these changes with aging has not been studied in detail. Here we employ the Brain Age framework to analyze migraine, by building a machine learning model that predicts age from neuroimaging data. We hypothesize that migraine pa-tients will exhibit an increased Brain Age Gap (the difference between the predicted age and the chronological age) compared to healthy participants. We trained a machine learning model to predict Brain Age from 2,771 T1-weighted magnetic resonance imaging scans of healthy subjects. The processing pipeline in-cluded the automatic segmentation of the images, the extraction of 1,479 imaging features (both morphological and intensity-based), harmonization, feature selection and training inside a 10-fold cross-validation scheme. Separate models based only on morphological and intensity features were also trained, and all the Brain Age models were later applied to a discovery cohort composed of 247 subjects, divided into healthy controls (HC, n=82), episodic migraine (EM, n=91), and chronic migraine pa-tients (CM, n=74). CM patients showed an increased Brain Age Gap compared to HC (4.16 vs -0.56 years, P=0.01). A smaller Brain Age Gap was found for EM patients, not reaching sta-tistical significance (1.21 vs -0.56 years, P=0.19). No associa-tions were found between the Brain Age Gap and headache or migraine frequency, or duration of the disease. Brain imag-ing features that have previously been associated with migraine were among the main drivers of the differences in the predicted age. Also, the separate analysis using only morphological or intensity-based features revealed different patterns in the Brain Age biomarker in patients with migraine. The brain-predicted age has shown to be a sensitive biomarker of CM patients and can help reveal distinct aging patterns in migraine.
Introduction
Migraine is a prevalent and chronic condition known for its recurrent and debilitating headache episodes. Migraine can be classified into two categories based on the frequency of headache days per month, namely episodic migraine (EM) and chronic migraine (CM) (1). Due to the inherent char-acteristics of migraine and its widespread occurrence, it im-poses a substantial burden on both individuals and society as a whole (2).
Migraine is associated with changes in the brain. Beyond the direct effects (i.e., the experience of pain during the ictal phase), neuroimaging studies have discovered alterations in the migrainous brain during the interictal phase encom-passing both the structural and functional levels (3–5). Dif-ferences between EM and CM have also been reported (6–8). Even though the structure and function of the brain are also impacted by changes due to brain development and ag-ing, the interplay between those and changes related to mi-graine has not been explored in depth. Bell et al. (9), for instance, focused on the pediatric age range, finding age-and puberty-dependent alterations in the functional connectivity of multiple networks in children with migraine using resting-state functional Magnetic Resonance Imaging (fMRI) and showing that brain changes associated with migraine begin in infancy and are modulated by development. Chong et al. (10) studied morphological changes of EM patients along age and found that patients with migraine have age-related thin-ning of regions compared to the control group. Using fluo-rodeoxyglucose positron emission tomography (FGD-PET), M. Lisicki et al. (11) showed that episodic migraine patients exhibit specific metabolic brain modifications while aging.
Recently, the so-called Brain Age paradigm has been proposed to explore the relationship between aging and disease (12). Using machine learning techniques from neuroimaging data, chronological age can be accurately predicted in healthy individuals. After training a Brain Age model, the difference between an individual’s chronological age and the age pre-dicted by the Brain Age model is usually referred to as “Brain Age Gap”, “Brain Age Gap Estimate” or “brain-predicted age difference” (brain-PAD), and has been proposed as an age-adjusted index of structural brain health. Research has shown the Brain Age paradigm to be sensitive to many neurologi-cal, psychiatric, and metabolic disorders, showing a positive Brain Age Gap, higher age compared-to the healthy brain, in disorders such as Alzheimer’s, schizophrenia, and type II diabetes, among others (13–15). Conversely, protective so-ciological and lifestyle factors including years of education, physical exercise, playing music, or meditation have been associated with a negative Brain Age Gap (16–18). An in-creased predicted Brain Age has even been associated with higher allostatic load and elevated overall mortality risk (19). Even though other pain-related conditions have been stud-ied using the Brain Age paradigm (20–23), to the best of our knowledge, migraine has not been explored from this per-spective.
In this work, the Brain Age framework was employed to in-vestigate migraine on a dataset composed of structural T1-weighted (T1w) MRI from EM and CM patients, together with normal controls. We hypothesized that migraine patients will exhibit an increased Brain Age Gap compared to healthy participants. Furthermore, we aimed to detect possible as-sociations between the Brain Age Gap and clinical charac-teristics in the patient groups exploring the role of different imaging features.
Materials and methods
Developing a robust brain age model involves several crucial steps. Firstly, it is imperative to assemble a diverse, broad, and representative dataset that encompasses neuroimaging data alongside corresponding chronological ages. The size of the dataset plays a significant role, as a larger dataset en-ables greater precision and generalizability in the final model. Subsequently, feature extraction is performed to capture per-tinent information from the neuroimaging data. This process ensures that only informative and discriminative features are included in the model.
Once the features have been extracted, an appropriate ma-chine learning algorithm is selected for age prediction based on the neuroimaging data. Common choices include support vector machines or neural networks. The chosen algorithm is then trained using the dataset, and techniques such as reg-ularization, cross-validation, and hyperparameter tuning are employed to optimize performance and prevent overfitting. The trained model is next evaluated using a separate dataset, employing metrics such as mean absolute error (MAE) or correlation coefficients to assess accuracy and generalization capabilities. This evaluation step provides valuable insights into the model’s performance and its ability to accurately es-timate brain age.
Finally, the trained brain age model can be applied to new and unseen neuroimaging data. In our case, we apply it to a dataset composed of healthy controls, patients with episodic migraine, and patients with chronic migraine.
Brain age model
To create and evaluate our age predic-tion models, we compiled a dataset (hereinafter referred to as Model Creation Dataset) consisting of 2,771 structural T1w MRI scans of healthy adults aged 18 to 60 from differ-ent studies and databases that were either publicly available. These include: the Dallas Lifespan Brain Study (DLBS) (24); the Consortium for Reliability and Reproducibility dataset (CoRR) (25); the Neurocognitive aging data release (Neu-roCog) (26); The OASIS-1 dataset (27); the Southwest Uni-versity Adult Lifespan Dataset (SALD) (28); the Informa-tion eXtraction from Images dataset (IXI) (29); and the Cam-CAN repository (available at http://www.mrc-cbu. cam.ac.uk/datasets/camcan/) (30, 31). In addition to these, we included a set of healthy adults from the Labo-ratorio de Procesado de Imagen (LPI), our own institution. Individuals who presented neurological or psychological di-agnoses or cognitive impairments were eliminated from the OASIS-1 or CoRR databases. Table 1 depicts the basic fea-tures of the Model Creation Dataset.
Summary characteristics of the datasets used in the Model Creation Dataset, sorted by median age.
From the T1w images, FastSurfer (32) was employed to ex-tract a total of 1,479 features. Fastsurfer uses Deep Learning to perform brain segmentation based on the Desikan-Killiany atlas (33, 34). Two types of features were extracted:
624 morphological features, including whole brain fea-tures, the volume of cortical and subcortical gray mat-ter regions and white matter regions from the atlas, as well as the surface, thickness and curvature of the cor-tical regions. This feature set will be referred to as Morphological Feature Set.
855 intensity-based features extracted from the same regions. This feature set will be referred to as Intensity Feature Set.
Together, all 1,479 features make up the Combined Feature Set. The three feature sets obtained using this procedure were the basis for further analysis.
To assure their quality, segmentations were manually in-spected. In Supplementary Table 1, Supplementary Figure 1 and Supplementary Table 2, features and regions of interest are covered in greater detail.
MRI acquisitions obtained at different sites and/or using dif-ferent protocols can differ in their intensity levels, which can introduce a bias in the Brain Age-predicting models. In or-der to cope with this problem, we used ComBatGAM (35) to harmonize the features from the Intensity Feature Set and the Combined Feature Set, using age, sex, and estimated total intracranial volume (eTIV) as covariates.
Afterwards, the cases were randomly divided into an 8:1:1 ratio for training, validation, and testing. We conducted a 10-fold cross-validation training procedure over the harmonized features to predict age. We flattened outliers of each feature, defined as values on the 97.5th or 2.5th percentile. In ad-dition, each characteristic was adjusted to the range (-1, 1) using min-max normalization. Each fold underwent feature selection, defining three sets of 20, 30, and 40 characteristics. The selection of features was accomplished in two steps. Ini-tially, a filter was used to choose the first decile features based on the mutual information between features and age in the training set. Next, the final feature were chosen by employ-ing a forward feature selection approach with gaussian mix-ture models to optimize the mutual information between a subset of features and age (36). As regressors, support vector regressor (SVR), random forest (RF), and a multilayer per-ceptron (MLP) were evaluated. Figure 1 depicts the process followed. By combining these three regressors with distinct feature sets of 20, 30, and 40 characteristics for each fold, a total of 90 models were trained. Predictions were obtained for the validation and test set for each fold. Validation re-sults were used to select the Brain Age model to be selected as the best-performing, while test results were exclusively employed to report the accuracy of the Brain Age model on the Model Creation Dataset. We are aware that Brain Age models suffer from regression dilution, which causes bias in Brain Age predictions. Therefore, in order to avoid possible spurious associations, a correction for this effect was applied (37, 38). A linear regression was fitted between the real age and validation results of each of the regressors of the ensem-ble. The intercept (α) and slope (β) of each fit were then used to correct the predictions obtained for the studied groups following the equation:
Comprehensive illustration of the methodologies employed for the training of the Brain Age models and the generation of brain predicted-ages. Model Creation shows the steps taken to train the Brain Age model on the Model Creation Dataset and choose the final model applied on the Application Dataset : a) Image processing includes Fastsurfer for brain segmentation and extraction of intensity and morphological features, thus building three feature sets: the Morphological Feature Set, Intensity Feature Set and the Combined Feature Set. For each of these feature sets, a feature selection procedure is performed in a 10-fold cross-validation scheme creating feature sets of 20, 30 and 40 features to feed the machine learning models (SVR, RF and MLP) for each fold. b) Validation is performed to select the best combination of feature set size and machine learning technique. c) Test on the Model Model Creation Dataset to assess performance of the Brain Age prediction model. Model Application depicts the use of the chosen model on the patient and healthy groups. Brain Age Gap is calculated as the difference between the predicted and the actual age. Differences in Brain Age Gap are then analyzed.
This approach was repeated for each of the aforementioned feature sets. The training procedure was performed using the scikit-learn Python library for machine learning (39). The SVR and RF models were imported from the library while the MLP was implemented using PyTorch (40). Details of the MLP implementation and the hyperparameters for each model are described in Supplementary Table 3.
Participants
A total of 247 subjects were included in this study, divided into healthy controls (HC, n=82), EM (n=91), and CM patients (n=74). This dataset, on which the Brain Age model previously described was applied, will be here-inafter referred to as Application Dataset. Patients were recruited from the outpatient headache unit at the Hospital Clínico Universitario de Valladolid (Spain), a public tertiary care institution that accepts patients from both secondary care and primary care. Inclusion criteria were: a) migraine diag-nosis using the third edition of the International Classifica-tion of Headache Disorders (ICHD-3) beta and ICHD-3 cri-teria (1, 41); b) a stable clinical state in the last six months; and c) expressed willingness to partake in the study, coupled with the voluntary signing of the informed consent document. We excluded patients with the following conditions: a) high-frequency episodic migraine, with 10 to 14 headache days per month; b) other painful conditions; c) known major psy-chiatric diseases (described as anamnesis or the presence of depression or anxiety in the Hospital Anxiety and Depres-sion Scale (42)); d) other neurological diseases; e) drug or substance abuse; and f) pregnancy. At the time of inclusion, no preventive treatment was given to the patients. Partici-pants were requested to complete a headache diary and were diagnosed with EM if they experienced 10 headache days per month or less and CM if they met the ICHD-3 criteria.
Age-and sex-matched HC were recruited through hospital and university colleagues, as well as ads at these institutions, using convenience sampling and snowball sampling. No HC were included if they had a current or previous history of mi-graine, or if they had any other neurological or mental dis-order following the same exclusion criteria as for migraine patients.
We gathered sociodemographic and clinical data from all patients, including migraine illness duration (years) and headache and migraine frequency (days per month).
The study was approved by Hospital Clínico Universitario de Valladolid’s local Ethics Committee (PI: 14-197). All participants read and signed a written consent form before their participation.
Image acquisition and processing
High-resolution 3D T1w MRI data were acquired for all subjects using a Philips Achieva 3T MRI unit (Philips Healthcare, Best, the Nether-lands) with a 32-channel head coil in the MRI facility at the Universidad de Valladolid (Spain). Acquisition parameters were the following: Turbo Field Echo (TFE) sequence, rep-etition time (TR) = 8.1 ms, echo time (TE) = 3.7 ms, flip angle=8o, 256 × 256 matrix size, 1 × 1 × 1 mm3 of spatial resolution, and 160 sagittal slices covering the whole brain. Image acquisitions for migraine patients were performed dur-ing interictal periods (defined as at least 24 hours from the last migraine attack). Details about the acquisition protocols of each public dataset are described in Supplementary Table 4. If more information is required, further details can be found in each portal of the databases used.
Following the image acquisition, image segmentation, fea-ture extraction and harmonization were also performed on the Application Dataset as described for the creation of the Brain Age model. Next, Brain Age was estimated for each participant, including correction from the regression dilution. Since we conducted a 10-fold cross-validation for the train-ing, validation and testing of the Brain Age model, an ensem-ble formed with the average result of the trained model from each fold was used to obtain the final prediction. Finally, the Brain Age Gap was calculated as the difference between the corrected predicted age and the chronological age of each in-dividual.
Model interpretation
The significance of each imaging feature in the Brain Age estimation was evaluated using SHapley Additive exPlanations (SHAP) (43). SHAP is a game-theory-based model-agnostic explanation method for machine learning models that evaluates the contribution of each feature to a given prediction. By employing this ap-proach, a group-level comparison of distinct brain imaging features can be conducted to determine their significant con-tribution to age prediction. Additionally, the evaluation of the influence of individual features on each participant’s Brain Age prediction is made possible, as exemplified in the study conducted by Ballester et al. (44).
The SHAP value for a particular feature for a specific pre-diction can be interpreted as the difference in the prediction when that feature is omitted from the model. SHAP values reinterpret complex models as a linear function:
where z’ is a simplified version of the input features of the model, φ0 is a reference value of the model (in our case is a value close to the average age of the training data), and φi, the attribute effect of the feature which deviates the prediction from the reference value. In a database with N participants and M features, for example, SHAP generates an N × M ma trix, where each value represents the contribution of feature m to the prediction of participant n.
We calculated the SHAP value for each subject for a deeper understanding of the regressors. Since many features are repeated across the different regressors, we summed up the contribution of repeated features into a single value. The fi-nal matrix was divided by 10 since our ensemble model is the average of the results of the 10 regressors trained during the 10-fold cross-validation.
Once we had the final matrix, we aggregated the values for each of the groups considered (HC, EM and CM) by sum-ming up the absolute values of the matrix along the partic-ipant’s axis. The best 15 features in terms of their absolute contribution for each group were selected for each model of the ensemble. Unique features among the three groups stud-ied were selected as the most informative features.
Statistical analysis
The performance evaluation of the Brain Age models was conducted using two metrics: the MAE and Pearson’s correlation coefficient (r). The MAE was calculated as the average of the absolute values of the residuals, which were obtained by subtracting the predicted age from the actual age for each individual in the group. The MAE serves as a comprehensive measure of the prediction error across the entire group, with lower values indicating a better fit. On the other hand, Pearson’s correlation coefficient measures the strength and direction of the linear relationship between the predicted ages and the real ages. Higher values of r indicate a better fit of the model. The specific formu-las for these metrics can be found in equations Eq. (3) and Eq. (4). Further exploration of these performance metrics can be found in the work from de Lange et al. (38).
We assessed the normality and homogeneity of variance for age and duration of migraine in the Application Dataset using the Kolmogorov-Smirnov test and Levene’s test for equal-ity of variances, respectively. If the null hypothesis was not rejected in both tests, we performed a one-way analysis of variance (ANOVA) to determine significant differences in the ages of the three groups. Gender-significant differences were identified using a chi-square test. For comparing clin-ical characteristics between migraine patients (i.e., duration of migraine history in years for both groups of patients), we used a two-tailed unpaired t-test if the null hypothesis was not rejected by the Kolmogorov-Smirnov; alternatively, we used the Mann-Whitney U test.
We performed an analysis of covariance (ANCOVA) on the Brain Age Gap results for each pair of groups, adding eTIV and sex as covariates. To verify that the Brain Age Gap calcu-lated for each group was approximately normal and that the variances between groups were comparable, we performed the Kolmogorov-Smirnov test and the Levene test. In the case of a negative Levene’s test, we verified that the variance ratio did not exceed 2 (45). We reset the P value threshold correct-ing for multiple comparisons using the Bonferroni correction method (P threshold = 0.0167).
Regarding the model interpretation, We conducted a Kruskal-Wallis test on the SHAP values obtained for each of the highly important features of each regressor trained to analyze differences in feature importance among the studied groups. A non-parametric test was chosen due to the non-normality of the SHAP values. To account for multiple comparisons, we applied the Benjamini-Hochberg correction method. We per-formed pairwise comparisons using the post-hoc Connover-Iman test, correcting its p-values for multiple comparisons using the Benjamini-Hochberg method if the Kruskal-Wallis Test was significant.
Finally, we computed the Pearson’s correlation coefficient between the Brain Age Gap and the clinical characteristics of the migraine groups, correcting for multiple comparisons using the Benjamini-Hochberg method. All statistical tests were conducted in Python.
Results
Demographics
There were no significant differences be-tween the groups (HC, EM and CM) in the Application Dataset regarding age or sex. Table 2 shows the demo-graphic and clinical characteristics of the dataset, while Figure 2 shows the age distribution of the subjects, together with those in the Model Creation Dataset.
demographic and clinical characteristics for the Application Dataset. Not all patients completed the headache diary. Complete data was available from 87 EM patients and 72 CM patients.
Age distributions of studies in the Model Creation Dataset and the Application Dataset, ordered by median age.
Performance of the Brain Age model
Table 3 provides a summary of the validation results for the models tested with the three feature sets studied after age bias correction. Re-sults before the bias correction can be found in Supplementary Table 5. The ensemble model formed by MLPs with a set of 40 selected features provided the greatest performance among the evaluated models in all feature sets and was there-fore selected to perform all the Brain Age predictions whose results are described next.
Validation results for the three regressors tested. Results are given as the average and the standard deviation of the values obtained from each fold of the 10-fold cross-validation scheme after age bias correction. The values in bold show the combination with the best result.
For the test data, training on the Model Creation Dataset, the Brain Age model working with the Combined Feature Set obtained a MAE and r of 5.95 years and 0.85. On the Appli-cation Dataset, this same model yielded an MAE and Pear-son’s correlation of 6.26 years and 0.84 for the HC group. With regard to the Brain Age model working only with the Morphological Feature Set, its performance was MAE = 7.12 years and r = 0.80 on the Model Creation Dataset, and MAE = 7.83 years and r = 0.74 on the Application Dataset for the HC group. Finally, the Brain Age model working only with the Intensity Feature Set, it yielded MAE = 8.19 years and r = 0.78 on the Model Creation Dataset, and MAE = 9.27 and years r = 0.64 on the Application Dataset for the HC group.
Brain Age Gap in migraine
Using the Combined Fea-ture Set, CM patients exhibited a statistically significant in-creased Brain Age Gap (average +4.16 vs -0.52 years, P = 0.010) compared to HC. EM patients showed an intermediate Brain Age Gap (average +1.21 years), and neither compar-isons with HC nor CM yielded statistical significance. Figure 3 graphically depicts these results, together with scatter-plots showing the brain-predicted age and the chronological age in both the Model Creation Dataset and the Application Dataset.
The results of each of the regressors build are shown. The MLPs selecting 40 features demonstrated the best results in validation for all the feature sets. For each trained regressor on every feature set: I) Ensemble MLPs result in the test set of each fold of Model Creation Dataset. II) Distribution of the Brain Age Gap values obtained for each of the studied groups. III) Brain Age Gap for the three groups. Statistical significance is denoted by an asterisk (*) to indicate significant findings.
When only employing the Morphological Feature Set, CM showed an increased Brain Age Gap with respect to both HC and EM (+7.54 vs 3.15 and 3.56 years, respectively), al-though these differences were not statistically significant (P = 0.07 and P = 0.17). Interestingly, the average Brain Age Gap for HC and EM were very similar in this case.
Finally, when only employing the Intensity Feature Set, CM again showed an increased Brain Age Gap with respect to HC and EM, although differences were not statistically sig-nificant in this case either. Figure 3 and Table 4 depict these results.
ANCOVA results for Brain Age Gap calculated for each regressor trained. Normality and equality of variances were tested before applying the ANCOVA. Age, sex, and eTIV were included as covariates. The η2 effect sizes were small (η2 < 0.06) for all comparisons. Statistically significant elements are shown in bold.
Model interpretation
Following the SHAP procedure de-scribed in Methods, 16 features from the regressor trained on the Combined Feature Set were selected. Among them, SHAP values differed significantly between CM patients and HC for the left hemisphere lateral orbitofrontal cortex gray matter volume, left hemisphere superior frontal gyrus gray matter volume and the left hemisphere Insula average thick-ness (p < 0.001 for all cases). For the first two features, the left hemisphere lateral orbitofrontal cortex gray matter volume and the left hemisphere superior frontal gyrus gray matter volume, significant differences were also found in the SHAP values between EM and CM (p < 0.01 in both cases). No other significant differences were found for the remain-ing characteristics among the studied groups. These results are graphically depicted in Figure 4.
P-values derived from the Kruskal-Wallis test and the post-hoc Connover-Iman test for each of the most significant characteristics chosen for each regressor. Features are ranked from highest importance in the HC group to lower, left-right. a) A total of 16 unique features were selected for the combined regressor, from which 3 demonstrated significant differences in the pairwise comparisons. b) A total of 17 unique features were chosen for the regressor trained on the Morphological Feature Set. Up to 10 features demonstrated significant differences in importance between groups. c) shows the 17 most significant characteristics for the regressor trained on the Intensity Feature Set. No significant differences were found during the Kruskal-Wallis test.
Regarding the regressor trained on the Morphological Fea-ture Set, 17 features were selected. Among them, nine fea-tures were significantly different between HC and CM, and ten, (the previous one plus one more) were significantly dif-ferent for the EM-CM comparison. These are: the volume of the right Putamen (HC-CM p < 0.05, EM-CM p < 0.01), the volume of the superior frontal gyrus of the left hemisphere (HC-CM p < 0.001, EM-CM p < 0.01), the volume of the lateral orbitofrontal cortex of the left hemisphere (HC-CM p < 0.001, EM-CM <0.01), the volume of the subcortical gray matter (EM-CM p < 0.01), the volume of the superior frontal gyrus of the right hemisphere (HC-CM p < 0.001, EM-CM p < 0.01), the total gray matter volume (HC-CM p < 0.001, EM-CM p < 0.001), the volume of the middle temporal gyrus of the right hemisphere (HC-CM p < 0.01, EM-CM p < 0.001), the folding index of the rostral middle frontal gyrus of the right hemisphere (HC-CM p < 0.001, EM-CM p < 0.01), the volume of the supramarginal gyrus of the right hemisphere (HC-CM p < 0.01, EM-CM p < 0.05) and the volume of the insula of the right hemisphere (HC-CM p < 0.001, EM-CM p < 0.001). These features and P values are also shown in Figure 4.
Finally, no features from the regressor trained on the Inten-sity Feature Set (among the 17 that were selected) showed significant differences after the Kruskal-Wallis test. Feature importance for each ensemble studied are depicted Supplementary Figure 2, Supplementary Figure 3 and Supplementary Figure 4.
Relation between Brain Age Gap and clinical charac-teristics
Employing the regressor trained on the Combined Feature Set, and both considering EM and CM separately or together, we found no significant association between the Brain Age Gap and headache frequency (CM -p = 0.89, ME -p = 0.72, both -p = 0.30), migraine frequency (CM -p = 0.71, EM -p = 0.62, both -p = 0.62), migraine duration in years (CM -p = 0.52, ME -p = 0.52, both -p = 0.52) or chronic mi-graine duration (p = 0.32). No significant associations were found either when considering the regressors trained on the Morphological Feature Set or the Intensity Feature Set. Re-sults are shown in Supplementary Figure 5, Supplementary Figure 6 and Supplementary Figure 7.
Discussion
In this study, we explored for the first time the Brain Age paradigm in migraine patients. To that end, a machine learn-ing model was developed to predict age from brain T1w MRI acquisitions from a large sample of healthy subjects and later applied to a dataset composed of HC, EM and CM. Several important findings emerged from this investigation. First, CM patients showed an increased predicted Brain Age com-pared to their healthy counterparts. EM patients, on the other hand, appear to have a much milder increase in their predicted Brain Age, which does not reach statistical significance. Sec-ond, different behaviors seem to arise when separately con-sidering morphological and intensity-based features. Distinct difference patterns of difference appear when using either morphological or intensity-based features, furthermore, mor-phological features seem to be key in the statistically signif-icant difference found. Finally, we could not prove an addi-tional effect of headache frequency or duration of migraine disease in patients’ Brain Age. However, careful consider-ation is necessary for this point. Most patients presented extremely low (less than 5) or high (20 or more) headache frequency values, and some CM patients included in the rel-atively small sample could transition between EM and CM states annually.
It is well documented that the process of aging induces changes in both the macroscopic and microstructural prop-erties of the brain (46–48). However, much less is known re-garding whether and how disease can affect or interact with this process. A few studies have investigated age-related studies in the migraine brain. Chong et al. (10) examined if aging affects cortical thickness differently in patients with migraine compared to age-matched HC, potentially exacer-bating cortical thinning in patients with migraine. For EM patients, the study found that patients with migraine expe-rienced age-related thinning in regions that do not thin in HC. This suggests that migraine may be linked to atypical cortical aging. Lisicki et al., on the other hand, (11), em-ployed FGD-PET to investigate possible specific age-related metabolic changes in the brain. They found that for EM pa-tients advancing age was positively correlated to increased metabolism in the brainstem, hippocampus, fusiform gyrus and parahippocampus, regardless of the frequency of mi-graine or the duration of the disease. Taken together, these re-sults suggest that migraine and aging do interact in the brain, and the nature of these interactions needs to be investigated.
We found a notable difference in the Brain Age Gap between CM patients and HC. The aforementioned studies, however, did not investigate CM patients but rather focused on EM. For these patients, although we identified a trend towards an increased Brain Age Gap with respect to HC, no significant differences were identified.
Our correlation analysis showed no association between the Brain Age Gap and clinical variables such as headache or migraine frequency and duration of the disease. Whereas these types of associations have been found in many condi-tions within the Brain Age framework (49–51), in other cases no associations were found between the Brain Age Gap and disease severity or duration (52). This is also the case in (53) with migraine patients, although the Brain Age framework was not explicitly employed in that study. Nevertheless, more extensive research including individuals with high-frequency episodic migraine and greater variation in the duration of the migraine may unveil subtle variances that were eluded in our study.
As with other cerebral changes found in migraine, observa-tional studies such as the present one cannot elucidate pos-sible causalities in the increased Brain Age Gap in CM pa-tients. Interestingly, Vidal-Pineiro et al. (53) found no as-sociation between cross-sectional Brain Age and the rate of Brain Age change measured longitudinally. Instead, they found that the Brain Age Gap seems to be related to early-life factors. Although those results were obtained from large datasets composed of healthy subjects, this “stability” feature of the Brain Age Gap for an individual would be consistent with our findings. Rather, an increased Brain Age Gap could be more related to the susceptibility of an individual to suf-fer from migraine than to the actual disease. Of course, lon-gitudinal studies are needed to elucidate the nature and the evolution of the Brain Age Gap in migraine.
The literature on Brain Age prediction has utilized a vari-ety of methods for creating machine learning models, includ-ing feature-based approaches or Deep Learning techniques (12, 54). In any case, accurate models are needed in or-der to obtain reliable predictions and good sensitivity to dis-ease. We chose to compile a large dataset (Model Creation Dataset) consisting of 2,771 MRI scans from different stud-ies and public databases in order to perform training and val-idation of the model, and to report its accuracy. The accuracy of our model (MAE and Pearson’s correlation coefficient of 5.95 and 0.85, respectively) when tested on the (Model Creation Dataset) is comparable to similar approaches in the lit-erature. More importantly, only a slight accuracy reduction was obtained when applying our model to the Application Dataset (MAE and Pearson’s correlation coefficient of 6.26 and 0.84, respectively), which indicates a good generaliza-tion capability and performance of the feature harmonization method.
Combining different types of features, as we did with the Combined Feature Set, is a prevalent technique in computer vision (55–57). The combination of these features is typi-cally accompanied by an enhancement in the performance of the employed machine learning models (58), as is the case with our study in comparison to the performance obtained for the models trained using the Morphological Feature Set and the Intensity Feature Set. This is typically explained by the lack of correlation between distinct feature spaces. How-ever, the separate results obtained for the Brain Age models trained using the Morphological Feature Set and the Inten-sity Feature Set allow to gain more insight into the behav-ior of CM and EM patients, since they offer complementary viewpoints of the nature of the Brain Age Gap in migraine. When employing the Morphological Feature Set, there is vir-tually no difference between HC and EM, whereas CM shows an increased Brain Age Gap that appears to vanish for older ages (see Figure 3 (b) II). Although further investigation is required to corroborate this, it suggests that morphological changes in the brain that are associated with CM are more prominent at younger ages, but aging then absorbs these alter-ations. Conversely, the behavior of the Brain Age Gap when using the Intensity Feature Set seems more stable across ages (see Figure 3 (c) II).
The interpretation of the Brain Age predicting model through SHAP allows us to better understand which brain imaging features mostly drive the Brain Age prediction, and which are responsible for the differences found. For the regressor trained on the Combined Feature Set, we identified a total of 16 features that highly influenced the prediction of the regres-sor across the studied groups. These characteristics primarily pertain to the frontal cortex (8), which is expected since fea-tures related to the frontal and temporal cortices are common in Brain Age models due to their generalized thinning as part of the normal process of aging (48). Features related to the size of the ventricles are also common in Brain Age models, but not so much in our case (two features) since the increase in the volume of the ventricles is more pronounced after the sixth decade of life (47), while the subjects employed in our study were only between 18 and 60 years old. Ten out of the 16 most significant features were intensity-based, likely representing changes in the tissue microstructure (59).
Three of the brain imaging features found to be most relevant for the Brain Age estimation showed significant differences between CM patients and HC for the regressor trained on the Combined Feature Set. Additionally, regarding the regressor trained on the Morphological Feature Set, significant differ-ences between CM and HC were found in nine additional fea-tures (and nine more features in the comparison between EM and HC). Brain regions related to the identified features have been shown to differ morphologically (6), connectivity-wise (60–62), or both between HC and CM patients and between HC and EM patients. These regions are involved in complex cognitive functions such as information integration or work-ing memory (63, 64) and their alteration may be related to the cognitive changes and other alterations associated with migraine (65, 66).
This study, however, presents several limitations that are worth discussing. First, as a cross-sectional observational study, causality cannot be established, and therefore the in-terpretation of the results must be taken with caution. Longi-tudinal studies are needed to elucidate whether susceptibility to migraine, the course of the disease or other factors are re-sponsible for the detected Brain Age Gap between CM and HC, and how this gap evolves with time. Second, our Brain Age model only employs T1w MRI, whereas other modal-ities such as diffusion MRI have shown to be sensitive to migraine (7). However, research has demonstrated that the integration of various modalities can enhance the precision of Brain Age prediction models (67). Nevertheless, the in-clusion of new modalities, despite yielding improved out-comes, can augment the complexity of the overall pipeline. This is particularly exacerbated by the challenges posed by the availability of substantial training data from diverse pub-lic datasets and the need for harmonization. Thus, we chose to limit ourselves to T1w scans in this work, although fu-ture work will need to include additional modalities such as diffusion MRI, as discussed earlier, or T2-weighted images, given their sensitivity to white matter hyperintensities, a well-known migraine characteristic associated with aging (68). Fi-nally, high-frequency EM patients (10 to 14 headaches per month) were excluded from the study. This decision was made to prevent potentially misclassified patients from skew-ing the results of the analysis.
Conclusions
In this study, we analyzed migraine using the Brain Age framework, which consists of training a machine learning model to predict age from MRI scans and later applying the resulting model to a cohort of interest. We found that CM patients exhibit an increased Brain Age Gap (i.e., the differ-ence between the predicted age and the chronological age) compared to HC. A milder Brain Age Gap was found for EM patients, although differences did not reach statistical signif-icance.
Further analysis of the Brain Age model indicated that imaging features that have previously been associated with changes in migraine were among the main drivers of the dif-ferences in the predicted age. Also, a separate analysis us-ing only morphological or intensity-based features revealed different patterns, which could represent distinct processes within the alterations that are associated with the migraine brain.
In conclusion, the Brain Age paradigm has shown to be a promising viewpoint for the study of migraine, and future work will be needed to corroborate these findings.
Data Availability
The public datasets used during the current study are available either online or after reasonable request. The institutional dataset used is also available author upon reasonable request to the corresponding author. The code used for modeling and data analysis is accessible at https://github.com/rafaloz/MigraineBA/
https://fcon_1000.projects.nitrc.org/indi/retro/dlbs.html
http://fcon_1000.projects.nitrc.org/indi/CoRR/html/concept.html
https://openneuro.org/datasets/ds003592
http://fcon_1000.projects.nitrc.org/indi/retro/sald.html
Funding
This work was supported by research grants PID2021-124407NB-I00, funded by MCIN/AEI/10.13039/501100011033/FEDER, UE, TED2021-130758B-I00, funded by MCIN/AEI/10.13039/501100011033 and the European Union “NextGenerationEU/PRTR” and PRE2019-089176 funded by MCIN/AEI/10.13039/501100011033 and ESF “ESF invests in your future”.
Abbreviations
ANCOVA: Analysis of covariance; ANOVA: Analysis of variance; brain-PAD: brain-predicted age difference; CM: Chronic migraine; CoRR: Consortium for Reliability and Reproducibility; DLBS: Dallas Lifespan Brain Study; EM: Episodic migraine; eTIV: Estimated total intracranial vol-ume; FGD-PET: fluorodeoxyglucose positron emission tomography; fMRI: functional Magnetic Resonance Imaging; HC: Healthy controls; ICHD-3: International Classification of Headache Disorders; IXI: Information eXtraction from Images dataset; MAE: Mean absolute error. MLP: Multilayer perception; NeuroCog: Nerocognitive Aging release; RF: Random forest; SALD: Southwest University Adult Lifespan Dataset; SHAP: Shaply additive explanations; SVR: Support vector regressor; TE: Echo time; TFE: Turbo Field Echo; TR: Repetition time; T1w: T1-weighted;
Availability of data and materials
The public datasets used during the current study are avail-able either online or after reasonable request. The institu-tional dataset used is also available author upon reasonable request to the corresponding author. The code used for mod-eling and data analysis is accessible at https://github.com/rafaloz/MigraineBA/.
Ethics approval and consent to participate
The research was authorized by Hospital Clínico Universi-tario de Valladolid’s local Ethics Committee (PI: 14-197).
Authors’ contributions
RNG and RLG were the major contributors to writing the manuscript. RNG, APG, DGA, and ALG collected and an-alyzed the data. RLG, SAF, RNG, and APG designed and conceptualized the study, analysed the data and revised the manuscript for intellectual content. All authors read and ap-proved the final manuscript.
Competing interests
The authors declare that they have no competing interests.
Consent for publication
Not applicable.
Additional Files
Additional File 1
Features extracted by regions.
Segmentation example from case 110033 of the CamCAN database. On the left a), the segmentation is shown without the white matter segmentation. Subcortical and cortical regions are divided. On the right b) the segmentation includes white matter segmentation.
Features extracted from the whole brain.
Hyperparameters of the regressors trained for the study.
Acquisition parameters for each scanner employed in every database used to construct the Brain Age model.
Validation results for the three regressors tested. Results are given as the average and the standard deviation of the values obtained from each fold of the 10-fold cross-validation scheme before age bias correction. The values in bold show the combination with the best result.
The sum of the absolute SHAP values of each feature for each member of the investigated groups, calculated for the regressor trained on the Combined Feature Set. The order of features varies between groups, but the 16 designated features are shared by the groups’ most pertinent features.
The sum of the absolute SHAP values of each feature for each member of the investigated groups, calculated for the regressor trained on the Morphological Feature Set. The order of features varies between groups, but the 17 designated features are shared by the groups’ most pertinent features.
The sum of the absolute SHAP values of each feature for each member of the investigated groups, calculated for the regressor trained on the Intensity Feature Set. The order of features varies between groups, but the 17 designated features are shared by the groups’ most pertinent features.
No correlations were found between Brain Age Gap calculated with the regressor trained on the Combined Feature Set and the clinical variables studied, a) Brain Age Gap change along with headache frequency, b) Brain Age Gap change along with migraine frequency, c) Brain Age Gap change along with migraine duration, and d) Brain Age Gap change along with chronic migraine duration.
No statistically significant correlation was found between clinical variables and Brain Age Gap when calculated with the regressor trained on the Morphological Feature Set, a) Brain Age Gap change along with headache frequency, b) Brain Age Gap change along with migraine frequency, c) Brain Age Gap change along with migraine duration, and d) Brain Age Gap change along with chronic migraine duration.
No statistically significant correlation was found between the clinical variables and the Brain Age Gap when calculated with the regressor trained on the Intensity Feature Set, a) Brain Age Gap change along with headache frequency, b) Brain Age Gap change along with migraine frequency, c) Brain Age Gap change along with migraine duration, and d) Brain Age Gap change along with chronic migraine duration.
Graphical Abstract
Acknowledgements
The authors are profoundly thankful to the owners of the Neurocognitive aging data release, which is avail-able at OpenNeuro under the identifier “OpenNeuro Dataset ds003592.”, the IXI dataset available at https://brain-development.org/ixi-dataset/, the DLBS dataset and the International Neuroimaging Data-Sharing Initiative Group and the participants of the Consortium for Reliability and Reproducibility (CoRR) for sharing their data publicly so this research was made possible.
Data were provided in part by OASIS OASIS-1: Cross-Sectional: Principal Investigators: D. Marcus, R, Buck-ner, J, Csernansky J. Morris; P50 AG05681, P01 AG03991, P01 AG026276, R01 AG021910, P20 MH071616, U24 RR021382;
Data collection and sharing in part for this project was pro-vided by the Cambridge Centre for Ageing and Neuroscience (CamCAN). CamCAN funding was provided by the UK Biotechnology and Biological Sciences Research Council (grant number BB/H008217/1), together with support from the UK Medical Research Council and University of Cam-bridge, UK.
The SALD data repository was supported by the National Natural Science Foundation of China (31470981; 31571137; 31500885), National Outstanding young people plan, the Program for the Top Young Talents by Chongqing, the Fundamental Research Funds for the Central Universities (SWU1509383,SWU1509451,SWU1609177), Natural Science Foundation of Chongqing (cstc2015jcyjA10106), Fok Ying Tung Education Foundation (151023), General Finan-cial Grant from the China Postdoctoral Science Founda-tion (2015M572423, 2015M580767), Special Funds from the Chongqing Postdoctoral Science Foundation (Xm2015037, Xm2016044), Key research for Humanities and social sci-ences of Ministry of Education (14JJD880009).
In addition, we are extremely grateful to all the patients and participants who volunteered to take part in this study.
Footnotes
The previous version represented an abstract presented at the conference of the International Society for Magnetic Resonance in Medicine in Toronto in 2023. This version is an extended work that includes more data and evaluates the significance of the features used.
