Adverse Childhood Experiences (ACEs) encompass physical, sexual, and emotional abuse, neglect, and other maltreatment during childhood. Such experiences are associated with lasting negative effects, including heightened risk for chronic illnesses and reduced life expectancy [1,2]. Neuroimaging research in adults links ACEs to structural and functional alterations in corticolimbic brain regions. Notably, childhood physical abuse correlates with diminished functional connectivity—measured as cross-correlated activity—between limbic structures (amygdala, hippocampus) and cortical areas (Prefrontal Cortex, Anterior Cingulate Cortex) in humans and animals. This weakened corticolimbic connectivity is tied to adverse emotional states and increased vulnerability to neuropsychiatric conditions, such as major depression, social anxiety, and bipolar disorder [3,4].

Voxel-Based Morphometry (VBM) research shows that adults with Adverse Childhood Experiences (ACE) often have reduced hippocampal and prefrontal cortex (PFC) volume, along with heightened activation of the Hypothalamic-Pituitary-Adrenal (HPA) axis, changes associated with increased inflammation and chronic stress. Results for amygdala alterations remain inconsistent [[5], [6], [7]]. ACE has also been linked to reduced grey matter and disrupted white matter connectivity in the PFC [8,9]. Diffusion Tensor Imaging (DTI) studies reveal that ACE impacts white matter (WM) integrity in psychiatric conditions like schizophrenia and bipolar disorder. Higher ACE exposure correlates with lower Fractional Anisotropy (FA) and higher Mean Diffusivity (MD) in WM tracts connecting cortical and subcortical regions [10,11]. Despite the above, a specific MR spectral signature for ACE remains unidentified.

Recent studies indicate that ACE-related epigenetic effects change with age, not persisting uniformly. Variations in DNA methylation and aging across life stages suggest a critical intervention window in young adulthood [12]. Research also reports notable neurochemical differences between young, middle-aged, and older adults without psychiatric disorder histories [12,13].

Emerging evidence shows ACE affects brain development differently in males and females, with region-specific patterns. Animal models of separation and deprivation reveal sex-based differences in stress responses [15]. In humans, women experience ACEs more often, highlighting notable gender disparities in exposure and neurodevelopmental impact [14].

Magnetic Resonance Spectroscopy (MRS) is a non-invasive technique that measures brain metabolite levels, detecting biochemical changes even when structural scans appear normal. At 3.0 Tesla, standard MRS resolves ten spectral peaks for six key metabolites, including Glu, Gln, mI, sI, tNAA (NAA + NAAG), tCho. Research on ACE-related brain chemistry is limited, but most studies focus on glutamate (Glu) and Glx, showing that higher ACE exposure often links to lower Glu, particularly in mood disorders [8,9,15].

Conventional MRS has difficulty detecting low-concentration metabolites like GABA due to spectral overlap [16] [17,18]. To overcome this limitation, J-difference editing with Hadamard encoding employs a multi-frequency acquisition scheme; the resulting spectra are processed using Hadamard linear combinations to resolve individual signals [16,19]. HERCULES (Hadamard Editing Resolves Chemicals Using Linear Combination Estimation of Spectra) [22] advances this further, allowing simultaneous quantification of up to 12 metabolites in an 11-min scan [[20], [21], [22]]. By separating overlapping signals from GABA, glutathione (GSH), ascorbate (Asc), and N-acetylaspartate (NAA) through four sub-experiments and three Hadamard spectra, HERCULES enhances resolution, providing a robust tool for studying neurochemical changes linked to mood disorders and ACE exposure [23,24].

Machine learning (ML) is a powerful tool for analyzing complex datasets, enabling computers to detect patterns and make predictions from prior examples. It begins with a training phase, where algorithms learn relationships between input features and labeled outcomes. Once trained, models can predict outcomes for new data [25]. In biomedical research, ML helps identify biomarkers or predictors, including those linked to ACEs. Common classifiers include Logistic Regression (Logit), modeling binary outcomes; Random Forest (RF), combining multiple decision trees to improve accuracy and rank feature importance; and Support Vector Machines (SVM), which find optimal boundaries to separate classes. These models often use inputs such as Magnetic Resonance Spectroscopy (MRS) data to predict outcomes such as ACE exposure, revealing biological mechanisms underlying disease [26,27].

Hence, there is a growing need for efficient and reliable identification of ACE-related neurochemical markers through ACE classifiers, as a potential non-invasive tool for assessing ACE exposure in individuals. Such an approach could help prevent the long-term negative effects of ACE by enabling timely and personalized interventions.

This cross-sectional observational study aimed to identify neurochemical patterns linked to ACE exposure by simultaneously assessing self-reported ACE histories and brain metabolite concentrations, using parallel enrollment alongside magnetic resonance spectroscopy (MRS) for data collection and analysis [28,29].