AI-supported clinical decision systems optimizing multimodal PTSD treatment selection for veterans with co-occurring traumatic brain injury and depression

 Corresponding author: Babatunde Stephen Adedeji Copyright © 2025 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution License 4.0. AI-supported clinical decision systems optimizing multimodal PTSD treatment selection for veterans with co-occurring traumatic brain injury and depression Babatunde Stephen Adedeji 1, *, Idowu R Adeyemo 2 and Azeez Elesho 3 1 Mental Health and Addiction Counselor at The Home for Little Wanderers, Roslindale, Massachusetts, USA. 2 Social Work, Ohio University, USA. 3 LCSW, Clinician at the Home for Little Wanderers, USA. GSC Biological and Pharmaceutical Sciences, 2025, 33(02), 436-452 Publication history: Received 16 October 2025; revised on 22 November 2025; accepted on 24 November 2025 Article DOI: https://doi.org/10.30574/gscbps.2025.33.2.0471 Abstract Post-traumatic stress disorder (PTSD) in military veterans frequently co-occurs with traumatic brain injury (TBI) and major depression, creating a complex clinical profile that challenges traditional treatment-selection approaches. Multimodal care including psychotherapy, pharmacotherapy, neurocognitive rehabilitation, and emerging neuromodulation techniques can improve outcomes, but determining the optimal combination for individual patients remains difficult due to heterogeneous symptom trajectories, overlapping neurobiological signatures, and varied treatment responses. This complexity has led to growing interest in artificial intelligence (AI)–supported clinical decision systems capable of integrating diverse data streams to guide personalized care pathways. AI-enabled platforms can analyze multimodal inputs such as electronic health records, neuroimaging markers, psychometric scales, behavioral data, and longitudinal treatment outcomes to generate predictive models for individualized therapy selection. Machine learning algorithms, particularly ensemble and deep learning architectures, are capable of identifying latent patterns linking neurocognitive impairment, affective dysregulation, trauma exposure profiles, and prior treatment responses to future clinical improvement. For veterans with co-occurring TBI and depression, these systems may help distinguish between PTSD-driven symptoms and TBI-related cognitive deficits, improving the precision of treatment assignments and reducing trial-and-error prescribing. Clinical decision systems also support multimodal treatment optimization by recommending combinations such as cognitive processing therapy with SSRIs, TBI-focused cognitive rehabilitation, or adjunctive neuromodulation based on predicted synergistic effects and individual risk factors. When integrated into clinical workflows, AI-supported tools can enhance early intervention, reduce care variability, and support measurement-based monitoring. By aligning clinical expertise with data-driven insights, AI-supported decision systems offer a scalable pathway for improving PTSD treatment outcomes in veterans with complex comorbidities. Their use has the potential to reduce symptom chronicity, enhance functional recovery, and support more efficient allocation of healthcare resources. Keywords: PTSD; Traumatic brain injury; Clinical decision support; Machine learning; Multimodal treatment; Veteran mental health 1. Introduction 1.1. Clinical burden of PTSD in veteran populations Post-traumatic stress disorder (PTSD) represents one of the most persistent and disabling conditions affecting veteran populations, with prevalence rates significantly higher than those observed in civilian groups [1]. Combat exposure, repeated traumatic events, and prolonged operational stress increase vulnerability to chronic PTSD trajectories, often GSC Biological and Pharmaceutical Sciences, 2025, 33(02), 436-452 437 leading to long-term psychological, social, and occupational impairment [2]. Veterans frequently present with intrusive memories, hyperarousal, sleep disturbances, emotional numbing, and impaired interpersonal functioning, all of which contribute to reduced quality of life and heightened healthcare utilization [3]. In many cases, PTSD is further associated with elevated risks of substance use, suicidal ideation, and cardiovascular morbidity, intensifying the clinical management burden placed on healthcare systems serving military personnel [4]. Moreover, delayed treatment-seeking behavior is common among veterans, driven by stigma, perceived weakness, or institutional barriers to mental-health services [5]. As a result, symptoms often become entrenched, reducing responsiveness to standard therapies and necessitating more complex clinical interventions [6]. Many veterans also struggle with reintegration into civilian life, where stressors such as unemployment, unstable housing, and limited social support exacerbate PTSD symptomatology [7]. The cumulative burden underscores the need for precise, individualized treatment decisions that consider both clinical and contextual determinants of recovery [8]. 1.2. Overlap of PTSD, TBI, and depression: diagnostic and therapeutic complexity PTSD rarely exists in isolation within veteran populations; instead, it frequently co-occurs with traumatic brain injury (TBI) and major depressive disorder (MDD), forming a clinical triad that complicates diagnosis and treatment [2]. Symptoms such as concentration difficulties, irritability, sleep issues, and memory impairment overlap across these conditions, making differential assessment challenging even for experienced clinicians [1]. This diagnostic ambiguity often results in misclassification or delayed identification of underlying neuropsychiatric drivers, thereby reducing the effectiveness of subsequent treatment pathways [4]. TBI adds another layer of complexity because its neurological effects may mimic or exacerbate PTSD symptoms, particularly in veterans with blast-related injuries [6]. Depression further amplifies functional impairment, contributing to emotional dysregulation, anhedonia, and motivational deficits that interfere with therapy engagement [3]. These intertwined symptom clusters often produce atypical response patterns to psychotherapy and pharmacological interventions, making standardized care models less effective for multimorbid veterans [8]. Therapeutic complexity is heightened by variability in severity, chronicity, and biological underpinnings across individuals [7]. For instance, patients with mild TBI may display cognitive vulnerabilities that hinder processing-based therapies, whereas those with severe depressive features may require integrated behavioral and pharmacological support [9]. This overlapping pathology reinforces the need for personalized approaches capable of disentangling symptom interactions and predicting optimal treatment responses [9]. 1.3. Limitations of conventional treatment-selection frameworks Traditional PTSD treatment-selection frameworks rely heavily on clinician judgment, broad diagnostic categories, and trial-and-error prescribing, making them poorly suited for veterans with complex comorbidities [3]. Evidence-based guidelines typically recommend trauma-focused psychotherapies and first-line medications, yet these recommendations assume homogeneity within diagnostic groups and do not account for neurocognitive differences, injury history, or symptom-interaction patterns common among veterans [1]. As a result, many patients experience incomplete remission, high dropout rates, and prolonged cycles of ineffective treatment attempts [4]. Conventional frameworks also overlook granular multimodal data such as neuroimaging findings, TBI severity markers, sleep metrics, and behavioral-engagement indicators that could meaningfully improve treatment matching [6]. In practice, clinicians may lack the time or resources to analyze complex datasets, resulting in underutilization of potentially predictive biomarkers [8]. Furthermore, standard protocols rarely incorporate longitudinal feedback loops that adapt treatment plans based on evolving symptom profiles, limiting responsiveness to clinical change [5]. The fragmentation between mental-health, neurological, and rehabilitation services further contributes to suboptimal care coordination, especially for veterans navigating multiple comorbidities [7]. Consequently, one-size-fits-all treatment strategies often fail to address heterogeneous needs, reinforcing the necessity of AI-driven models capable of integrating multimodal data to optimize individualized PTSD treatment pathways [9]. 2. Pathophysiology and symptom interaction in PTSD–TBI–depression profiles 2.1. Neurobiological mechanisms linking trauma exposure, neural injury, and affective dysregulation Trauma exposure initiates a cascade of neurobiological alterations involving the amygdala, hippocampus, and prefrontal cortex regions responsible for threat detection, memory consolidation, and emotional regulation [11]. In GSC Biological and Pharmaceutical Sciences, 2025, 33(02), 436-452 438 veterans, repeated combat stress amplifies these effects, producing hyper-reactive limbic circuits and weakened topdown inhibitory control, which contribute to persistent hyperarousal and maladaptive threat responses [14]. Traumatic brain injury (TBI) further disrupts neural pathways by damaging white-matter tracts and frontotemporal networks, impairing executive function, emotional processing, and attentional regulation [12]. These disruptions intensify PTSD symptomatology by impairing cognitive flexibility and increasing susceptibility to intrusive recollections [17]. Neuroendocrine dysregulation also plays a crucial role, as trauma alters hypothalamic–pituitary–adrenal (HPA) axis functioning, resulting in abnormal cortisol patterns that affect stress reactivity and mood stability [15]. Microstructural injury from blast exposure can additionally impair the default mode network, leading to disturbances in self-referential processing and emotional integration [18]. These combined neural changes explain why many veterans experience difficulty distinguishing between current environmental cues and trauma-linked memories, reinforcing cycles of anxiety and avoidance. Neuroinflammatory responses triggered by both TBI and psychological trauma further contribute to mood dysregulation, cognitive slowing, and increased depressive features [19]. Ultimately, the convergence of neural injury and trauma-driven alterations produces a complex neurobiological phenotype that is difficult to treat using conventional, single-modal therapeutic strategies [13]. 2.2. Cognitive, emotional, and behavioral symptom overlap: differential diagnosis challenges PTSD, TBI, and depression share numerous overlapping symptoms, making differential diagnosis in veterans particularly challenging for clinicians [16]. Cognitive impairments such as memory deficits, slowed processing, poor concentration, and executive dysfunction appear in all three conditions, often masking underlying neurological injury or obscuring trauma-specific drivers of distress [11]. Emotional symptoms including irritability, anhedonia, guilt, and emotional numbing similarly overlap, complicating evaluation of whether mood disturbances stem from neurobiological injury, trauma-induced dysregulation, or depressive pathology [18]. Behavioral manifestations add another layer of complexity. Sleep disturbances, avoidance behaviors, impulsivity, and reduced motivation are core features of PTSD but also commonly follow concussive injuries or depressive episodes [15]. Such cross-cutting symptoms often lead to ambiguous clinical presentations, requiring detailed longitudinal assessment and multimodal diagnostic tools to distinguish between interrelated etiologies [12]. Misdiagnosis or partial diagnosis is common, especially when TBI-related symptoms mimic emotional dysregulation or when depressive features overshadow underlying trauma cues [19]. Moreover, psychological distress may interact with neurological deficits, creating hybrid symptom patterns not fully captured by standard diagnostic categories [14]. For example, a veteran with mild TBI may struggle with cognitive fatigue that reduces engagement in trauma-focused therapy, while depressive rumination may heighten sensitivity to PTSD triggers [17]. These interactions demonstrate that symptom clusters are rarely isolated; instead, they manifest synergistically, making single-diagnostic frameworks insufficient for accurate assessment [13]. Consequently, integrated diagnostic models capable of parsing symptom interdependencies are essential for effective clinical decisionmaking [16]. 2.3. Implications of comorbidity for treatment planning and therapeutic sequencing Comorbidity among PTSD, TBI, and depression significantly complicates treatment planning, as overlapping symptoms influence responsiveness to both psychotherapy and pharmacological interventions [11]. Veterans with cognitive impairment may struggle with the sustained attention required for exposure-based therapies, while those with depressive features often exhibit reduced motivation, impairing engagement and treatment persistence [15]. Similarly, neurological deficits may limit the effectiveness of cognitive restructuring techniques, requiring modifications in pacing, modality, or therapeutic intensity [18]. Pharmacological treatment becomes equally complex because medications targeting PTSD symptoms can worsen cognitive deficits associated with TBI or interact unfavorably with antidepressants prescribed for co-occurring mood disturbances [16]. Treatment sequencing is therefore critical: addressing severe depressive symptoms may be necessary before initiating trauma-processing interventions, while stabilizing neurological impairments may improve therapy readiness [12]. Comorbidity also increases the likelihood of atypical response profiles, necessitating adaptive, individualized treatment plans rather than linear, guideline-driven approaches [19]. These patterns highlight the limitations of traditional, siloed GSC Biological and Pharmaceutical Sciences, 2025, 33(02), 436-452 439 treatment frameworks and reinforce the need for integrated care strategies capable of capturing dynamic interactions among trauma, neural injury, and emotional dysregulation [17]. 3. Multimodal data inputs for AI-driven treatment optimization 3.1. Electronic health records, psychometrics, and clinical trajectory data Electronic health records (EHRs) provide foundational longitudinal data for understanding how PTSD, TBI, and depression evolve over time in veteran populations [20]. They capture diagnostic codes, medication histories, therapy participation, symptom fluctuations, and clinical encounters, enabling large-scale temporal analyses that can reveal patterns invisible in isolated assessments [23]. EHR-derived trajectory mapping is particularly valuable because veterans often cycle through multiple care settings, producing a diverse set of observations that AI systems can integrate into predictive analytics frameworks [24]. Psychometric assessments including PTSD severity scales, depression inventories, and cognitive-screening tools supply standardized quantitative markers that complement EHR narratives and enhance model precision [22]. Longitudinal symptom tracking is critical for multimorbidity, as PTSD symptoms may stabilize while cognitive impairments worsen or depressive features intensify, requiring dynamic analytic models that can adapt to shifting clinical states [28]. EHR data also contain comorbidity markers such as substance-use history, sleep disorders, and chronic pain, which commonly co-occur in veterans and influence treatment responsiveness [25]. Additionally, therapy engagement metrics including session attendance, completion rates, and dropout patterns provide behavioral signals that help machine-learning systems estimate adherence likelihood and treatment sustainment [26]. Combined, these EHR and psychometric datasets form a core multimodal foundation for AI-enabled clinical decision systems. They provide structured and semi-structured information needed to generate individualized predictions, helping clinicians navigate complexities inherent in PTSD–TBI–depression cases while reducing reliance on trial-anderror treatment selection [27]. 3.2. Neuroimaging, biomarker profiles, and cognitive-function assessments Neuroimaging constitutes one of the most informative modalities for characterizing the neurobiological disruption present in veterans with overlapping PTSD, TBI, and depression [21]. Structural MRI can identify cortical thinning, volumetric loss, and white-matter abnormalities associated with blast exposure or repeated concussive events [24]. Functional MRI and PET imaging further reveal altered connectivity patterns within emotion-regulation and attentional networks, providing quantitative markers that correlate with symptom clusters [20]. These imaging-derived signals enrich AI models by adding objective physiological measurements that differentiate trauma-driven dysregulation from injury-driven cognitive impairment [27]. Biomarker data including inflammatory cytokines, cortisol dysregulation profiles, and neurodegenerative markers offer another level of biological specificity [22]. Inflammatory signatures, for instance, can help explain persistent fatigue, cognitive slowing, and mood instability, while stress-hormone irregularities can indicate chronic hyperarousal linked to PTSD [28]. Molecular markers may also detect subtle neuronal injury even when imaging yields ambiguous results, strengthening diagnostic confidence [26]. Cognitive-function assessments encompassing executive function testing, memory evaluations, attention measures, and processing-speed tasks add essential behavioral correlates to neurobiological findings [25]. These measures help differentiate TBI-related cognitive deficits from trauma-induced disorganization or depressive slowing, providing critical context for AI-supported treatment stratification [23]. When integrated, neuroimaging, biomarker panels, and cognitive testing produce a multidimensional physiological profile that enhances model accuracy and supports individualized therapeutic targeting [21]. 3.3. Behavioral, environmental, and digital phenotyping signals Beyond clinical and biological data, behavioral and environmental signals play an increasingly important role in multimodal PTSD–TBI–depression modeling [28]. Wearable sensors and mobile devices can capture sleep patterns, physical activity, heart-rate variability, geolocation stability, and daily routine consistency indicators closely linked to symptom severity and relapse risk [24]. These digital phenotyping markers provide real-time, high-frequency observations that complement slower-moving clinical data streams, enabling AI systems to detect early warning patterns before clinical deterioration becomes apparent [27]. GSC Biological and Pharmaceutical Sciences, 2025, 33(02), 436-452 440 Environmental variables such as social-support availability, occupational stressors, and exposure to high-stimulation settings also influence symptom trajectories but are often missing from traditional assessments [20]. Integrating these contextual factors helps AI systems model how external pressures interact with underlying neurobiological vulnerabilities [23]. Behavioral data including communication patterns, engagement with digital platforms, and passive monitoring metrics can further reveal subtle functional impairments not captured during clinical visits [25]. Together, these behavioral and environmental streams complete the multimodal data ecosystem required for precision treatment selection. Figure 1 Multimodal Data Streams Used in AI-Enabled PTSD–TBI–Depression Decision Systems 4. AI architectures for personalized PTSD treatment pathways 4.1. Machine learning models for predicting treatment response Machine learning (ML) has emerged as a powerful tool for predicting individualized treatment response in veterans with overlapping PTSD, TBI, and depression, largely due to its ability to uncover nonlinear patterns in complex, multimodal datasets [29]. Traditional statistical models often struggle to capture such heterogeneity, whereas ML approaches such as random forests, gradient-boosted trees, and support-vector machines can integrate symptom trajectories, psychometric scores, neurocognitive test results, and medication histories to infer which interventions are most likely to succeed for a specific patient [27]. These models identify subtle associations across datasets, detecting early predictors of psychotherapy dropout or pharmacological intolerance that clinicians may overlook when relying solely on experience or static guidelines [31]. ML treatment-response prediction models improve over time as additional patient data accumulate, enabling continuous recalibration of probability estimates and supporting adaptive personalization [33]. For example, ensemble methods can incorporate therapy engagement metrics, comorbidity patterns, and sleep-quality indicators to predict which veterans will benefit from cognitive processing therapy versus prolonged-exposure-based modalities [30]. Similarly, ML models trained on medication-response data can forecast the likelihood of adverse reactions or insufficient symptom reduction, facilitating earlier adjustments in care pathways [36]. These capabilities make ML essential for guiding individualized treatment selection in cases where multimorbidity generates atypical symptom presentations that defy traditional guideline-based approaches [34]. 4.2. Deep learning for multimodal fusion of neurocognitive and affective datasets GSC Biological and Pharmaceutical Sciences, 2025, 33(02), 436-452 441 Deep learning (DL) techniques excel in multimodal data fusion, making them particularly valuable for modeling the combined neurocognitive, emotional, and behavioral disturbances observed in PTSD–TBI–depression comorbidity [28]. Convolutional neural networks (CNNs) can extract high-resolution features from neuroimaging inputs such as fMRI connectivity maps or DTI-derived white-matter profiles capturing structural and functional disruptions linked to trauma exposure and neural injury [32]. Meanwhile, recurrent neural networks (RNNs) and long short-term memory (LSTM) architectures can model temporal changes in symptom severity, emotional variability, and sleep disruptions, providing dynamic representations of clinical evolution over time [37]. Multimodal DL frameworks integrate imaging, biomarkers, psychometrics, cognitive-performance data, and digital phenotyping signals into unified latent representations that capture relationships spanning biological, psychological, and behavioral domains [35]. This fusion enables the system to identify treatment-response signatures that emerge only when diverse data sources are jointly analyzed, such as patterns linking amygdala hyperconnectivity, inflammatory markers, and avoidance behavior to therapy non-response [27]. DL-based models also accommodate incomplete datasets by learning shared latent structures, allowing predictions even when certain modalities such as biomarkers or imaging are missing for a given individual [38]. This robustness is critical for real-world veteran care environments where data completeness varies widely. Ultimately, DL multimodal fusion supports deeper mechanistic understanding and more precise treatment stratification compared to traditional single-source analytical methods [31]. 4.3. Reinforcement-learning and adaptive decision systems for dynamic treatment sequencing Reinforcement learning (RL) provides a framework for modeling treatment planning as a sequential decision process in which the system learns optimal therapy strategies through iterative evaluation of patient outcomes [33]. RL-based models consider treatment selection not as a one-time choice but as an evolving sequence of decisions shaped by symptom changes, side-effect profiles, and engagement behaviors over time [29]. For veterans with multimorbid PTSD, TBI, and depression, this dynamic approach is especially important because symptom trajectories frequently shift in non-linear patterns influenced by neurological injury, environmental stressors, and fluctuating mood states [36]. RL systems use reward functions that encode therapeutic goals such as symptom reduction, cognitive improvement, or reduced relapse risk and update treatment strategies based on observed outcomes [28]. This allows the model to learn which interventions should be introduced, intensified, or discontinued at each stage of care. For example, RL may determine that initiating behavioral activation before trauma-focused psychotherapy produces better long-term outcomes in individuals with severe depressive features, whereas others may benefit from early cognitive rehabilitation due to TBI-related impairments [37]. Adaptive RL systems can also incorporate uncertainty and risk by factoring in the likelihood of dropout or adverse medication reactions, enabling more resilient treatment pathways [35]. Table 1 Comparison of AI Modeling Approaches for Multimodal PTSD Treatment Optimization AI Approach Primary Function Strengths in PTSD– TBI–Depression Modeling Limitations / Risks Best Clinical Use Cases Machine Learning (ML): Random Forests, Gradient Boosting, SVMs Predicts treatment response and relapse risk using structured clinical and behavioral data. • Interpretable models (relative to deep learning) • Handles nonlinear interactions well • Effective with limited or heterogeneous datasets • Good for early-stage triage and therapymatching • Limited ability to model highdimensional imaging or biomarker data • Performance depends on feature engineering • Risk of bias if training datasets lack representativeness • Predicting SSRI/SNRI response • Estimating risk of therapy dropout • Identifying candidates for trauma-focused psychotherapy Deep Learning (DL): CNNs, RNNs, Learns latent patterns across • Excellent for highdimensional and • Requires large, balanced datasets • Identifying neural signatures of GSC Biological and Pharmaceutical Sciences, 2025, 33(02), 436-452 442 LSTMs, Multimodal Fusion Networks neuroimaging, biomarkers, psychometrics, and digital phenotyping. multimodal integration • Learns complex neurocognitive– affective patterns • Outperforms ML in imaging, EEG, and behavioral-stream modeling • Captures temporal dynamics of symptom evolution • Low interpretability without XAI tools • Susceptible to overfitting with small or noisy datasets therapy response • Detecting early neurocognitive decline • Integrating sleep, behavior, and imaging to refine treatment plans Reinforcement Learning (RL) Learns optimal, personalized treatment sequences over time based on outcome feedback. • Models treatment as a dynamic, multistage process • Adjusts therapy recommendations as symptoms evolve • Incorporates sideeffects, adherence patterns, and relapse signals • Supports long-term precision adaptation • Requires longitudinal outcome data • Difficult to validate prospectively in clinical settings • Ethical concerns if recommendations appear opaque • Sequencing psychotherapy + neuromodulation • Adjusting treatment intensity dynamically • Long-term followup planning and relapse prevention Explainable AI (XAI) Provides interpretability, transparency, and clinician-aligned reasoning for ML/DL/RL outputs. • Enhances clinician trust and accountability • Identifies key drivers of treatment recommendations • Supports shared decision-making with veterans • XAI tools vary in clarity and granularity • Partial explanations may oversimplify underlying m 4.4. Explainable AI: transparency, interpretability, and clinician trust Explainable AI (XAI) is essential for ensuring that clinicians trust and adopt ML, DL, and RL systems in PTSD–TBI– depression care pathways [30]. Many high-performing models especially deep architectures operate as “black boxes,” making it difficult for clinicians to understand why a certain therapy or medication is recommended [34]. XAI tools such as SHAP values, feature-importance maps, and saliency visualizations help translate complex model outputs into human-interpretable insights that align with clinical reasoning [27]. Interpretability is especially critical in military and veteran-care settings, where transparency influences regulatory acceptance and ethical compliance [38]. By showing which factors such as cognitive scores, imaging findings, or behavioral patterns drove a recommendation, XAI strengthens clinician confidence and supports shared decisionmaking with patients [33]. This transparency bridges the gap between AI capabilities and practical clinical workflows, enabling precision treatment selection grounded in both data and clinician judgment [36]. 5. Evidence-based multimodal treatment options for complex PTSD cases 5.1. Psychotherapies: CPT, PE, EMDR, CBT-I, and trauma-integrated modalities Psychotherapy remains the therapeutic cornerstone for PTSD, especially in veteran populations presenting with overlapping TBI and depressive symptoms. Cognitive Processing Therapy (CPT) provides structured cognitive restructuring to address maladaptive trauma-related beliefs and has demonstrated strong efficacy even when depressive symptoms are prominent [38]. Prolonged Exposure (PE) therapy reduces avoidance patterns through systematic engagement with trauma memories and cues, although cognitive impairments associated with TBI may require pacing adjustments or auxiliary cognitive-rehabilitation supports [41]. Eye Movement Desensitization and GSC Biological and Pharmaceutical Sciences, 2025, 33(02), 436-452 443 Reprocessing (EMDR) integrates bilateral stimulation with trauma recall, offering benefits for veterans who struggle with the high cognitive demands of PE or CPT [37]. Cognitive Behavioral Therapy for Insomnia (CBT-I) is particularly relevant, as sleep disturbances represent a central symptom cluster across PTSD, TBI, and depression, and sleep normalization often enhances responsiveness to other modalities [44]. Trauma-integrated modalities, including acceptance-based therapies and resilience-building interventions, offer flexible therapeutic structures that accommodate cognitive fluctuations common in mild TBI cases [39]. Clinical complexity arises when multimorbidity alters the sequence or intensity of psychotherapy. For example, severe cognitive fatigue may require shortened sessions, while persistent depressive features may necessitate behavioral activation prior to trauma-focused work [43]. Across modalities, individualized adaptation remains essential to sustaining engagement and improving long-term outcomes in complex veteran populations [45]. 5.2. Pharmacotherapies: SSRIs, SNRIs, prazosin, novel investigational agents Pharmacotherapy plays a central adjunctive role in treating PTSD–TBI–depression overlap, particularly when severe affective dysregulation or neurobiological instability limits engagement in psychotherapy. Selective serotonin reuptake inhibitors (SSRIs) remain first-line agents due to their broad efficacy across anxiety, depressive, and intrusive-thought symptoms, though treatment response may be attenuated in veterans with significant TBI-related network disruption [40]. Serotonin–norepinephrine reuptake inhibitors (SNRIs) provide an alternative mechanism of action and may offer enhanced benefits for comorbid pain syndromes, which frequently co-occur in this population [38]. Prazosin remains widely used for trauma-related nightmares and hyperarousal, addressing autonomic dysregulation that often exacerbates sleep disturbances and daytime irritability [42]. Investigational agents such as ketamine, MDMAassisted psychotherapy, and neurosteroids have shown promise for refractory cases, particularly where traditional medications fail to achieve remission [37]. However, TBI-related vulnerabilities require careful consideration of dissociative or cognitive-altering effects when evaluating these treatments [45]. 5.3. Neuromodulation and neurorehabilitation for PTSD with TBI comorbidity Figure 2 Multimodal Treatment Pathways for PTSD–TBI–Depression Overlap Neuromodulation approaches have gained increasing interest for veterans whose PTSD symptoms coexist with significant TBI-related network dysfunction. Repetitive transcranial magnetic stimulation (rTMS) can modulate GSC Biological and Pharmaceutical Sciences, 2025, 33(02), 436-452 444 prefrontal–limbic circuitry implicated in emotional dysregulation and has shown meaningful reductions in hyperarousal and depressive symptoms in patients who have not responded to first-line therapies [41]. Transcranial direct-current stimulation (tDCS) and neurofeedback further target cortical connectivity abnormalities, offering adaptable, non-pharmacological strategies for addressing neurocognitive deficits [44]. Neurorehabilitation encompassing cognitive-remediation therapies, memory-training protocols, and compensatorystrategy coaching plays an essential role in enhancing executive function and improving capacity for psychotherapy engagement among individuals with TBI-linked impairments [39]. These modalities provide synergistic support when layered with trauma-focused interventions [43]. 5.4. Combining modalities: sequencing, augmentation, and synergy models Effective management of PTSD–TBI–depression overlap often requires combining modalities to match evolving symptom profiles [38]. Treatment sequencing is critical: stabilizing sleep with CBT-I or prazosin may improve readiness for trauma-focused therapy, while cognitive remediation may precede CPT for veterans with TBI-related impairments [45]. Augmentation strategies such as pairing SSRIs with neuromodulation or integrating EMDR with cognitiveenhancement protocols can achieve synergistic symptom reduction [40]. Multimodal pathways allow clinicians to adjust intensity, pacing, and therapeutic targets in response to dynamic clinical changes, improving long-term functional recovery in complex veteran populations [42]. 6. AI-supported clinical decision systems in practice 6.1. Decision-support interfaces: risk scoring, therapy recommendations, and uncertainty ranges AI-enabled clinical decision-support interfaces serve as the translation layer between complex predictive models and frontline clinicians treating veterans with PTSD–TBI–depression comorbidity. These interfaces typically combine riskstratification dashboards with personalized treatment-recommendation modules, enabling rapid synthesis of multimodal data that clinicians would otherwise need extensive time to manually interpret [41]. Risk scoring incorporates neurocognitive profiles, symptom trajectories, biomarker patterns, and behavioral-adherence indicators, producing patient-specific severity and relapse-risk estimates that update dynamically with new information [40]. Treatment-recommendation modules provide prioritized therapy options such as CPT, EMDR, or rTMS based on predictive modeling of expected clinical response for a given individual [45]. These systems often include probability ranges, confidence levels, and uncertainty intervals, helping clinicians understand the strength of supporting evidence and model reliability [43]. Uncertainty quantification is particularly crucial when dealing with veterans who present with high symptom variability, incomplete diagnostic histories, or fluctuating cognitive capacity, as overly deterministic outputs could misguide decision-making [48]. Modern interfaces allow drill-down functionality, enabling clinicians to inspect the underlying factors that contributed to a recommendation, such as neuroimaging anomalies, sleep disruptions, or depressive-burden indices [50]. This transparency strengthens clinical trust and supports treatment planning that is both personalized and evidence-aligned [47]. Ultimately, intuitive decision-support tools help clinicians manage complex comorbidity patterns with improved precision and reduced cognitive workload [44]. 6.2. Integrating AI outputs into psychiatry, neurology, and rehabilitation workflows Integrating AI-generated insights into clinical workflows requires alignment with the operational realities of psychiatry, neurology, and physical or cognitive rehabilitation services. In psychiatric settings, AI-driven treatment suggestions must fit into standard diagnostic interviews, psychotherapy planning, and medication-monitoring routines without disrupting therapeutic rapport or workflow cadence [40]. For instance, risk-scoring outputs may be reviewed during case planning, while trajectory predictions may inform therapy sequencing for individuals with severe depressive or avoidance patterns [42]. In neurology, integration emphasizes harmonizing AI outputs with imaging interpretations, cognitive test results, and TBI-severity classifications. Clinicians may use model-based alerts to identify patients who could benefit from neuromodulation or neurorehabilitation interventions, particularly when traditional diagnostic markers are ambiguous [46]. Rehabilitation specialists benefit from AI tools that map cognitive-performance data, motor-function metrics, and behavioral signals to personalized rehabilitation plans, enabling more adaptive pacing and progress monitoring [49]. GSC Biological and Pharmaceutical Sciences, 2025, 33(02), 436-452 451 [27] G. F. Farayola et al., "Reducing False Ventricular Tachycardia Alarms in ICU Settings: A Machine Learning Approach," 2025 10th International Conference on Machine Learning Technologies (ICMLT), Helsinki, Finland, 2025, pp. 63-68, doi: 10.1109/ICMLT65785.2025.11193329 [28] Emi-Johnson O, Fasanya O, Adeniyi A. Predictive crop protection using machine learning: A scalable framework for U.S. agriculture. International Journal of Science and Research Archive. 2024;12(02):3065-3083. doi:10.30574/ijsra.2024.12.2.1536. [29] Chinedu J. Nzekwe, Seongtae Kim, Sayed A. Mostafa, Interaction Selection and Prediction Performance in HighDimensional Data: A Comparative Study of Statistical and Tree-Based Methods, J. data sci. 22(2024), no. 2, 259279, DOI 10.6339/24-JDS1127 [30] Kayode B. Qiskit-based simulation of quantum attacks against RSA encryption. Int J Latest Technol Eng Manage Appl Sci (IJLTEMAS). 2025;14(5):592-6. [31] Otoko J. Microelectronics cleanroom design: precision fabrication for semiconductor innovation, AI, and national security in the U.S. tech sector. Int Res J Mod Eng Technol Sci. 2025;7(2) [32] Tetteh C. Reading against the Colonial Archive: Using AI to Recover the Hidden Economic Contributions of Market Women in Ghana (1890–1957) and African American Washerwomen in the U.S. South (1865–1920). International Journal of Research Publication and Reviews. 2024;5(12):6153-6168. Available from: https://ijrpr.com/uploads/V5ISSUE12/IJRPR36944.pdf [33] Onyechi VN. Managing large-scale capital projects in oil and gas: Cross-functional engineering leadership, cost performance and execution excellence. Global Journal of Engineering and Technology Advances. 2024;21(3):210223. doi:10.30574/gjeta.2024.21.3.0233 [34] Tetteh C. Voices of Continuity: Creating an AI-Enhanced Digital Oral Archive of Ghanaian Queen Mothers and African American Church Mothers as Custodians of Community Power. International Journal of Science and Research Archive. 2025;15(02):1923-1940. doi: https://doi.org/10.30574/ijsra.2025.15.2.1643 [37] Nweke Obinna, Adelusi Oluwatosin. Utilizing AI Driven Forecasting, Optimization, and Data Insights to Strengthen Corporate Strategic Planning. International Journal of Research Publication and Reviews. 2025 Mar;6(3):4260-4271. doi: 10.55248/gengpi.6.0325.1209. [38] Kayode B, Takpah NEA, Adeleke O, Oriakhi VN. Enhancing healthcare IoT (H-IoT) resilience: a comprehensive review. Open Access Res J Eng Technol. 2025;8(1):95-105. doi:10.53022/oarjet.2025.8.1.0033. [39] Chinedu Jude Nzekwe and Christopher J. Ozurumba. Advanced modelling techniques for anomaly detection: A proactive approach to database breach mitigation. International Journal of Science and Research Archive. 2024;13(02):2893–2909. doi:10.30574/ijsra.2024.13.2.2511. Available from: https://doi.org/10.30574/ijsra.2024.13.2.2511 [40] Ayankoya MB, Omotoso SS, Ogunlana AA. Data Driven Financial Optimization for Small and Medium Enterprises (SMEs): A Framework to Improve Efficiency and Resilience in U.S. Local Economies. International Journal of Management and Organizational Research. 2025 Jul;4(4):90–97. https://doi.org/10.54660/IJMOR.2025.4.4.9097 [41] Chinedu Jude Nzekwe and Christopher J. Ozurumba. Integrated strategies for database protection: Leveraging anomaly detection and predictive modelling to prevent data breaches. World Journal of Advanced Research and Reviews. 2024;24(03):1451–1466. doi:10.30574/wjarr.2024.24.3.3843. Available from: https://doi.org/10.30574/wjarr.2024.24.3.3843 [42] Emi-Johnson O, Fasanya O, Adeniyi A. Explainable AI for pesticide decision-making: Enhancing trust in datadriven crop protection models. International Journal of Computer Applications Technology and Research. 2025;14(01):147-161. doi:10.7753/IJCATR1401.1013. [43] Obinna Nweke. STRATEGIC DATA UTILIZATION FOR MINORITY-OWNED BUSINESSES: ENHANCING MARKET PENETRATION, CUSTOMER INSIGHTS, AND REVENUE GROWTH. International Journal of Engineering Technology Research & Management (IJETRM). 2025Mar29;09(03). [44] Adeyanju BE, Loto AO, Bello M, Adebisi T. Food Safety and Hygienic Practices Among Street Food Vendors in Ife East Local Government Area, Osun State, Nigeria. International Journal of Research and Scientific Innovation (IJRSI). 2025 Mar;12(15):468. doi: 10.51244/IJRSI.2025.121500042P. [45] Kayode B, Adebola NT, Akerele S. The State of AI-Driven Cybersecurity: Trends, Challenges, and Opportunities. J Artif Intell Mach Learn & Data Sci. 2025;3(2):2731-2739. doi:10.51219/JAIMLD/Bisola-Kayode/577 GSC Biological and Pharmaceutical Sciences, 2025, 33(02), 436-452 452 [46] Mostafa, S.A.; Smith, K.; Nelson, K.; Elbayoumi, T.; Nzekwe, C. A Learning Strategy Intervention to Promote SelfRegulation, Growth Mindset, and Performance in Introductory Mathematics Courses. Eur. J. Investig. Health Psychol. Educ. 2025, 15, 198. https://doi.org/10.3390/ejihpe15100198 [47] Nosakhare VO, Kayode B, Akerele S, et al. Machine Learning in Cybersecurity: A Multi-Industry Case Study Analysis for Enhanced Threat Detection and Response. J Artif Intell Mach Learn & Data Sci. 2025;3(2):2684-2691. DOI: doi.org/10.51219/JAIMLD/Victor-Oriakhi-Nosakhare/568 [48] Mol SS, Arntz A, Metsemakers JF, Dinant GJ, Vilters-Van Montfort PA, Knottnerus JA. Symptoms of post-traumatic stress disorder after non-traumatic events: Evidence from an open population study. The British Journal of Psychiatry. 2005 Jun;186(6):494-9. [49] Kayode B. A comparative study of SSDLC adoption: challenges and best practices in small and medium-sized enterprises and large organizations. Int J Res Appl Sci Eng Technol (IJRASET). 2025;13(5). [50] Olanlokun Y, Taiwo M. Policy and regulatory reforms needed to accelerate local pharmaceutical manufacturing in Nigeria: A comparative analysis with India and Bangladesh. Magna Scientia Advanced Biology and Pharmacy. 2025;14(01):53–68. doi:10.30574/msabp.2025.14.1.0010. [51] Kayode B. Assessing the impact of threat modeling on cyber threat prevention. Int J Prog Res Eng Manage Sci (IJPREMS). 2025;5(5):725-33. doi:10.58257/IJPREMS41272.