Neurobiological insights into psychotic depression from magnetic resonance imaging: a systematic review

Abstract

Psychotic depression (PD) is increasingly recognized as a distinct psychiatric condition with unique neurobiological features. To identify neuroimaging biomarkers of PD, we systematically reviewed evidence across structural magnetic resonance imaging (MRI), functional MRI, diffusion tensor imaging, and magnetic resonance spectroscopy, following PRISMA guidelines. PubMed, MEDLINE, and Embase were searched through December 2025; 28 of 1195 records met PICOS-based inclusion criteria. PD showed distinct structural and functional abnormalities versus healthy controls and non-psychotic depression, particularly involving lower gray matter volume in prefrontal, insular, anterior cingulate, and temporal regions; aberrant temporoparietal and frontoparietal function; altered dorsal anterior cingulate metabolism; and default mode network-cortical hypoconnectivity. Pharmacological and electroconvulsive therapies may partially reverse frontotemporal and visual network abnormalities. Evidence for subcortical volume changes was inconsistent. Despite small, heterogeneous studies precluding meta-analysis, symptom-correlated fronto-insular-limbic, and precuneus alterations may represent potential biomarkers warranting further investigation.

1. Introduction

Psychotic depression (PD) is recognized as a subtype of depression characterized by psychotic symptoms with varying severity, including mood-congruent delusions and hallucinations, as well as mood-discordant features.¹ Compared to individuals with non-psychotic depression (NPD), those with PD experience more severe depressive and psychotic symptoms, relapse more often, and bear a greater disease burden.² Notably, PD patients are among those at the highest risk for suicidal ideation across severe mental illnesses.³

Early identification and diagnosis of PD are crucial, as treatment recommendations are often diagnosis-dependent: in addition to antidepressants, patients with PD typically require combined antipsychotic treatment to relieve symptoms and prevent disease progression.⁴ In clinical practice, PD is primarily diagnosed based on the Diagnostic and Statistical Manual of Mental Disorders (DSM) and the International Classification of Diseases and Related Health Problems¹ (Table S1¹). Symptom dimensions and severity are commonly assessed using scales such as the Hamilton Depression Rating Scale (HAMD), the Mini-Mental State Examination scale (MMSE), and the Scales for the Assessment of Negative and Positive Symptoms (SANS/SAPS). However, substantial inter-individual heterogeneity and variations across diagnostic criteria remain major challenges.² Furthermore, longitudinal studies indicate that psychotic-affective diagnoses like PD can be highly unstable over time and difficult to distinguish from other severe mental illnesses, such as bipolar disorder or schizophrenia-spectrum disorders, based solely on early clinical presentations.⁵ Consequently, there is an urgent need to identify biomarkers for PD to assist clinical diagnosis and improve prediction of disease course and treatment response.

The non-invasive neuroimaging methodology of magnetic resonance imaging (MRI) has been widely used in PD. Studies have used structural MRI (sMRI), functional MRI (fMRI), diffusion tensor imaging (DTI), and magnetic resonance spectroscopy (MRS) to investigate differences among individuals with PD and NPD, and healthy controls (HC), exploring the potential for diagnosis of PD⁶ and the neural correlates of symptom severity and treatment outcomes.⁷ This systematic review summarizes recent MR-based findings in PD, with a focus on structural and functional alterations that may improve understanding of its neurobiology.

2. Methods

The study protocol was registered in PROSPERO (CRD420261426285) under the title: “Neurobiological Insights into Psychotic Depression from Magnetic Resonance Imaging: A Systematic Review”.

2.1. Search strategy

This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) guidelines⁸ (details are provided in the Supplementary Materials). A comprehensive literature search was performed in PubMed, MEDLINE, and Embase for peer-reviewed articles published up to December 2025. The key search terms were “psychotic depression”, “psychotic affective disorders”, “affective psychosis”, or “depressive psychosis”; “magnetic resonance imaging” or “MRI” (details are provided in the Supplementary Methods). The search strategy identified 1157 studies (PubMed: 731, Medline: 79, and Embase: 347). The reference lists of relevant articles were manually screened to identify further eligible studies.

2.2. Inclusion and exclusion criteria

The review includes articles that met the following criteria: published in English in a peer-reviewed journal; include participants with a primary diagnosis of PD; report MRI-based structural or functional brain findings. Exclusion criteria were: duplicate publications; non-original research articles (e.g., case reports, meta-analyses, reviews); unavailable full-text; study not involving PD patients; no quantitative MRI analysis. Following the defined criteria, one reviewer performed the primary screening of titles, abstracts, and full texts. In cases of uncertainty regarding study eligibility, a second reviewer was consulted, and discrepancies were resolved through discussion. To further validate the overall screening process and minimize selection bias, this second reviewer also blindly screened a random subset of 120 records from the total pool of excluded studies. A 100% absolute agreement rate was achieved (all 120 records being confirmed as correctly excluded), making it highly likely that no eligible studies were erroneously omitted.

2.3. Data extraction and presentation

For each study, three categories of variables were extracted: sample characteristics (sample size, sex, age, symptom severity, comorbidities, medication status, and non-pharmacological treatments), neuroimaging features (imaging modality, scanning parameters, image analysis software, coordinate type, whole-brain or region-of-interest analyses, and statistical methods), and findings (baseline characteristics, treatment-related outcomes, correlation analyses findings, study limitations, and future directions). Data extraction was primarily performed by one reviewer. Subsequently, to ensure maximum accuracy, a second reviewer independently verified the extracted data for all 28 included studies: the overall inter-extractor agreement was initially 78%. Any uncertainties or minor discrepancies during the data extraction process were cross-checked and discussed between the reviewers to reach a final consensus for the purposes of this review.

2.4. Risk of bias assessments

The methodological quality of case-control and cohort studies was assessed using the Newcastle-Ottawa Scale (NOS), which evaluates three domains: selection of study populations (4 points), comparability between groups (2 points), and result evaluation (3 points). Maximum score is 9 points; studies scoring 0–4 are considered low quality, and those scoring 5–9 are considered high quality.

Randomized controlled trials were assessed using version 2 of the Cochrane Risk of Bias tool for randomized trials (ROB 2.0), which evaluates five domains: randomization process; deviations from intended interventions; missing outcome data; measurement of outcomes; and selection of reported results. Overall risk of bias is deemed low if all domains are rated “low risk”; as of some concern if at least one domain is evaluated as “having a certain level of risk” but none are “high risk”; and as high if at least one domain is rated “high risk”.

Initially, risk of bias assessments were conducted by a single reviewer. To ensure methodological rigor and align with PRISMA guidelines, a second reviewer independently evaluated all 28 included studies. Inter-rater reliability for the case-control and cohort studies (using NOS) was calculated using Cohen’ s Kappa,⁹ while percentage agreement was calculated for randomized controlled trials (using RoB 2.0) due to the limited number of data points.¹⁰ Any disagreements between the reviewers were resolved by consensus for the purposes of this review.

3. Results

3.1. Included studies

The screening and selection process is summarized in Fig. 1. Of 1157 records identified through the literature search, 570 were excluded for duplication or because they were not original research articles. Based on titles and abstracts, a further 552 records were excluded because the study populations did not meet the diagnostic criteria for PD. During the full-text review of the remaining articles, nine studies were excluded for the following specific reasons: use of non-MRI modalities (e.g., PET or blood biomarkers), lack of available results (e.g., terminated trials or protocols), or absence of extractable group-level coordinate/region-of-interest (ROI) data (e.g., machine learning classification studies). Additionally, manual screening of the reference lists of included studies and relevant reviews identified 38 additional records, of which 2 additional studies were included. Ultimately, 28 studies were included in the analysis. Their participant demographic and clinical characteristics are summarized in Table 1.

Fig. 1.

PRISMA flowchart for literature screening included in the review.

Table 1.

Demographic and clinical characteristics of subjects in the studies included in the review.

	No. participants (female): group			Age, years (mean): group			Severity score (mean): group						Nondrug therapy
Study	PD	NPD	HC	PD	NPD	HC	PD	NPD	HC	Scale	Comorbidity	Drug status
Structural studies
Bijanki et al. 201411	20 (13)	–	20 (13)	46	–	45.4	38.5/12.9/3.3/0.5	–	–	HAMD-28/SANS/SAPS-P/SAPS-D	No	95% antidepressant, 25% anxiolytic, 40% antipsychotic, 5% Lithium	–
Bingham et al. 202312	80 (48)	–	–	54.9	–	–	5.6	–	–	HAMD-17	Yes	100% medicated (antidepressant + antipsychotic/placebo)	–
Cohen et al. 201313	20 (9)	19 (13)	22 (11)	36.4	37.3	32.2	30.0/31	23.7/15.3	0.6/0.3	HAMD-21/BPRS	No	19/23 medicated (15 antidepressant, 16 antipsychotic, 4 mood stabilizer)	–
Colombo et al. 201214	20 (15)	–	94 (41)	29.9	–	30.2	–	–	–	–	Yes	84/122 antipsychotic, no PD subgroup	–
Coryell et al. 20057	10 (4)	–	10 (4)	21.9	–	22.1	–	–	–	–	NA	NA	–
de Azevedo-Marques Périco et al. 201115	20 (15)	–	94 (41)	29.9	–	30.2	19.6/9.2	–	–	HAMD/PANSS-positive	Yes	18/20 medicated (antidepressant 55%, antipsychotic 75%, mood stabilizer 20%, benzodiazepine 25%)	–
Keller et al. 200816	23 (11)	19 (12)	22 (11)	36.5	36.6	32.2	30.5/48.9	23.7/33.2	0.6/18.3	HAMD-21/BPRS	No	19/23 medicated (15 antidepressant, 16 antipsychotic, 4 mood stabilizer)	–
Kim et al. 199917	19 (7)	26 (9)	–	64.8	65.1	–	37.4/26.1	32.5/27.2	–	MADRS/MMSE	No	Naïve	–
Kim et al. 202418	126 (78)	–	–	55.3	–	–	–	–	–	–	Yes	100% medicated (antidepressant + antipsychotic/placebo)	–
Lesser et al. 199119	14 (10)	–	72 (44)	56.9	–	61.5	27.1/28.1	–	1.4/29.5	HAMD/MMSE	Yes	10/14 neuroleptic, 10/14 antidepressants	–
Neufeld et al. 202020	86 (52)	–	159 (87)	54.5	–	44.7	5.8/28	–	-/29.3	HAMD-17/MMSE	Yes	100% medicated (antidepressant + antipsychotic/placebo)	–
O'Brien et al. 199721	22 (14)	39 (29)	–	72.9	70.3	–	27.7/25.3	23.2/24.9	–	HAMD/MMSE	No	NA	ECT
Oudega et al. 201422	24 (20)	31 (16)	23 (12)	74.3	70.8	70.3	33.7a	–	MADRS	No	7/24 antipsychotic	ECT
Rabl et al. 202523	26 (14)	27 (16)	32 (18)	34.5	42.1	37.2	30.1/11.9	23.9/4.1	0.5/4	HAMD-21/BPRS positive symptom subscale	Yes	89% medicated (antidepressant, antipsychotic, anxiolytic, mood stabilizer)	–
Salokangas et al. 200224	20 (12)	17 (9)	19 (7)	34	38.4	30.5	–	–	–	–	–	NA	–
Simpson et al. 19996	18 (11)	81 (57)	–	75.2	74.3	–	38.2/25.4	31.1/26.3	–	MADRS/MMSE	Yes	100% medicated (neuroleptic + antidepressant)	ECT
Takamiya et al. 202225	56 (35)	52 (32)	–	71.7	70.2	–	30.5a	24.6a	–	HAMD-17	No	6/8 sites medicated (54% antidepressant, 39% antipsychotic, 2% lithium, 18% benzodiazepine)	ECT
Vassilopoulou et al. 201326	17 (12)	22 (17)	18 (13)	52.9	52.1	46.6	23.4/28.7	16.3/29.2	–	HAMD-21/MMSE	Yes	100% medicated (17 antidepressant, 15 antipsychotic, 9 benzodiazepine)	–
Voineskos et al. 2020-S + O27	38 (21)	–	–	54.4	–	–	5.1/28.2	–	–	HAMD-17/MMSE	Yes	100% medicated (antidepressant + antipsychotic/placebo)	–
Voineskos et al. 2020-S + P27	34 (19)	–	–	56.1	–	–	6.4/27.8	–	–	HAMD-17/MMSE
Yamasaki et al. 202028	7 (3)	7 (3)	–	71.3	49.4	–	30.7a/20.7a	22.4a/24.4a	–	HAMD-17/MMSE	No	13/14 medicated (13 antidepressant + antipsychotic)	ECT
Functional studies
Chen et al. 202429	42 (25)	–	–	37.6	–	–	25.1	–	–	HAMD-24	No	Naive	–
Garrett et al. 201130	16 (7)	15 (10)	19 (6)	34.1	39.8	34.9	27.9/28.8	24.2/15.7	0.6/0.4	HAMD-21/BPRS	Yes	14/16 medicated (4 anxiolytic, 8 antidepressant, 10 antipsychotic, 1 mood stabilizer)	–
Neufeld et al. 2018—sample 131	28 (16)	–	39 (22)	56.2	–	55.1	28.5/28.3	–	-/29.4	HAMD-17/MMSE	Yes	100% medicated (antidepressant + antipsychotic)	–
Neufeld et al. 2018—sample 231	29 (14)	–	36 (22)	56.1	–	48.3	27.8/27.6	–	-/29.1	HAMD-17/MMSE
Neufeld et al. 2023—S + O32	28 (16)	–	–	53.4	–	–	4.8/28.1	–	–	HAMD-17/MMSE	Yes	100% medicated (antidepressant + antipsychotic/placebo)	–
Neufeld et al. 2023—S + P32	30 (16)	–	–	55	–	–	6.2/27.8	–	–	HAMD-17/MMSE
Oudega et al. 2019–site 133	14	9	–	68.7	–		32.9/26.5b	–	–	MADRS/MMSE	No	5/14 (site 1: 1 antipsychotic, 3 antidepressant, 1 mood stabilizer), 1/9 (site 2) medicated	–
	23 (16)
Oudega et al. 2019—site 233	9	17	–	72.0	–		33.8/26.0b	–	–	MADRS/MMSE
	26 (17)
Sommer et al. 200734	14 (6)	–	14 (6)	46	–	30	14/9	–	–	HAMD/PANSS-positive	–	NA	–
Sudheimer et al. 201535	22 (13)	39 (28)	39 (24)	39.7	42.2	34.13	29.9/47.5	23.3/32.8	0.5/18.5	HAMD/BPRS	Yes	19/22 medicated (antidepressant, antipsychotic, anxiolytic)	–
Tani et al. 2024—S + O36	15 (10)	–	46 (20)	53.6	–	42.2	3.8	–	–	HAMD-17	No	100% medicated (antidepressant + antipsychotic/placebo)	–
Tani et al. 2024—S + P36	18 (11)			54.3			5.6			HAMD-17

Note: BPRS, the Brief Psychiatric Rating Scale; ECT, electroconvulsive therapy; HAMD, the Hamilton Depression Rating Scale; HC, healthy controls; MADRS, the Montgomery–Asberg Depression Rating Scale; MMSE, Mini-Mental State Examination; NA, not available; No., number; NPD, non-psychotic depression; PANSS, the Positive and Negative Syndrome Scale; PD, psychotic depression; S + O, sertraline and olanzapine; S + P, sertraline and placebo; SANS/SAPS, the Scales for the Assessment of Negative and Positive Symptoms; SAPS-D, SAPS–Disorganization; SAPS-P, SAPS–Psychoticism.

^aWhere there are multiple measurement results, only the score at baseline is displayed.

^bMedian.

3.2. Bias assessment

Following the independent dual-assessment process, the inter-rater reliability for observational studies was high (Cohen's Kappa = 0.797), and the absolute agreement for RCTs was 100%. Based on the final consensus, all 19 case-control studies and all 6 cohort studies were judged to have a low risk of bias. All three randomized controlled trials were judged to have a high risk of bias, mainly because not all participants completed the full study protocol, resulting in missing data that could potentially influence the results. However, attrition-related bias is relatively common in longitudinal and interventional studies, and its impact on the overall conclusions of this review was considered limited. Details of the final consensus bias assessment scores and questionnaires are given in the Supplementary Materials (Figs. S1–S6).

3.3. Sample characteristics

3.3.1. Demographic variables

The 28 reviewed studies included 965 PD patients, 420 NPD patients, and 778 HC. Their mean age was 52.5 (SD 17.6) years, 59.0 (17.1) years, and 41.4 (17.0) years, respectively, and the proportion of female participants was 59%, 64%, and 52%, respectively (these figures exclude a study of 49 participants, which did not report separate demographic information for PD and NPD³³). Based on the methodological review, primary neuroimaging studies generally included age as a covariate across statistical frameworks (e.g., general linear models including Analysis of covariance (ANCOVA),^13,16 mixed-model regression,^27,32 and Cox proportional hazards regression models¹²) to adjust for age as a well-established determinant of brain structure, thereby minimizing the confounding influence of age.

3.3.2. Medication status and comorbidities

Most participants (in 22 studies) received pharmacological treatment during the study period. Based on the available medication profiles, the predominant pharmacological strategy was a combination of antidepressants and antipsychotics. Specifically, participants in seven studies were enrolled in the Study of the Pharmacotherapy of Psychotic Depression (STOP-PD) II randomized controlled trial,^{12,18,20,27,31,32,36} which used a standard combination protocol (antidepressants plus antipsychotics or placebo). In the remaining medicated cohorts, although individual medication regimens varied, the majority of patients were similarly prescribed either antidepressants, antipsychotics, or a concurrent combination of both. Adjunctive treatments, including anxiolytics (e.g., benzodiazepines) and mood stabilizers (e.g., lithium), were also frequently reported. In only two studies were participants medication-free,^17,29 while four studies did not report details of medication.^7,21,24,34 Participants in five studies received non-pharmacological treatment, all electroconvulsive therapy (ECT), during the study period.^{6,21,22,25,28}

Half of the studies involved participants with psychiatric or medical comorbidities;^{6,12,14,15,18–20,23,26,27,30–32,35} the remainder either excluded comorbid conditions or did not explicitly report comorbidity status.

3.3.3. Symptom measurements

Depressive symptoms were assessed most commonly using the Hamilton Depression Rating Scale (HAMD) (71% of studies), followed by the Montgomery–Åsberg Depression Rating Scale (MADRS) (18%), the Core Endogenomorphic Scale (11%), and the Carroll Rating Scale for Depression (4%). Four studies did not assess depressive symptoms.^7,14,18,24 Psychotic symptoms were assessed most commonly using the MMSE (39% of studies), followed by the Brief Psychiatric Rating Scale (BPRS) (18%), the Positive and Negative Syndrome Scale (PANSS) (7%), and the SANS/SAPS (4%). Nine studies did not assess psychotic symptoms.^{7,12,14,18,22,24,25,29,36}

Of the 11 studies that directly compared PD and NPD patients,^{6,13,16,17,21,23,25,26,28,30,35} most (10 studies) reported that PD had more severe depressive symptoms than NPD, and about half (5 studies) reported more severe psychotic symptoms in PD vs. NPD.

3.3.4. Analytic methods

Eleven studies conducted whole-brain analyses of structural or functional alterations,^{6,14,20,22,25,28–32,35} most using voxel-based morphometry (VBM) for structural analysis. Of the remaining 17 studies,^{7,11–13,15–19,21,23,24,26,27,33,34,36} most adopted a ROI-based approach to investigate MRI changes in specific brain regions. Scanning methods and statistical analysis parameters are summarized in Table 2.

Table 2.

Scanning methods and statistical analysis parameters of the studies included in the review.

	Scanning methods			Whole brain analysis				Statistical threshold
Study	Scanner	Tesla	Thickness		Analysis software	Coordinate type	Covariates	(correction)
Bijanki et al. 201411	GE Signa/Siemens Avanto	1.5	1.5–1.7 mm (T1); 3 mm (T2)	No	BRAINS2	–	–	p < 0.05 (FDR)
Bingham et al. 202312	GE Discovery MR750/Philips Achieva/Siemens TIM Trio	3	0.9/1 mm (T1); 2.9/2.5 mm (DTI)	No	FreeSurfer 6.0	O'Donnell Research Group white matter atlas	Age	p unknown (FDR-Benjamini-Hochberg)
Chen et al. 202429	GE Signa	3	4 mm (fMRI); 1 mm (sMRI)	Yes	VBM	MNI	Age, education factors	p unknown (GRF)
Cohen et al. 201313	GE Signa	3	1.5 mm	No	BrainImage	–	Total brain volume, age	p < 0.05
Colombo et al. 201214	GE Signa	1.5	1.5 mm	Yes	VBM	Talairach and Tournoux Atlas	Any unmatched demographic variables,	p < 0.05 (FWE)
Coryell et al. 20057	NA	1.5	1 mm	No	BRAINS2	–	Age, total frontal lobe volume	Unknown
de Azevedo-Marques Périco et al. 201115	GE Signa	1.5	1.5 mm	No	VBM	AAL;
Talairach and Tournoux atlas	Total GM volume, gender, scanner type, any unmatched demographic variables	p < 0.05 (FWE)
Garrett et al. 201130	GE Signa	3	4 mm	Yes	VBM	MNI; Talairach	Gender, response time, HAMD score	p = 0.01 (dual-significance corrected threshold)
Keller et al. 200816	GE Signa	3	1.5 mm	No	BrainImage	Talairach	Total brain volume, age	p < 0.05
Kim et al. 199917	GE Signa	1.5	1.5 mm	No	An Allegro digital image workstation	–	Age, sex, education	p < 0.05 for PFC and STG; p < 0.01 for other regions
Kim et al. 202418	GE Discovery MR750/Philips Achieva/Siemens TIM Trio	3	0.9/1 mm (T1); 2.9/2.5 mm (DTI)	No	FreeSurfer 6.0 longitudinal	Desikan Killiany atlas, Johns Hopkins University white matter atlas	3 stratification variables: daily dosage and steady state concentrations, time from baseline to final scan, intracranial volume	p < 0.05
Lesser et al. 199119	Picker MRI	1.5	10 mm	No	Viewing console	Atlas of the Human Brain in Section	–	–
Neufeld et al. 201831	GE Discovery MR750/Siemens	3	5/2.8/4 mm (fMRI); 0.9/1 (sMRI)	Yes	VBM	MNI	Age, sex, years of education, mean frame displacement, site	p < 0.05 (FWE, Bonferroni)
Neufeld et al. 202020	GE Discovery MR750/Philips Achieva/Siemens TIM Trio	3	0.9/1 mm (T1)	Yes	FreeSurfer 6.0	fsaverage5 template	Age, sex, site, education, total CIRSG scores, SCID-IV item, sertraline and olanzapine dosages	p < 0.05 (FDR)
Neufeld et al. 202332	NA	3	5/2.8/4 mm (fMRI); 0.9/1 mm (sMRI)	Yes	fMRIPrep 1.2.1	Multimodal Parcellation 1.0 atlas and Melbourne Subcortex Atlas	Sex, age	p < 0.15 (FDR)
O'Brien et al. 199721	Fonar Melville	0.3	5.1/5 mm	No	–	The Human Central Nervous System, 3rd ed.	–	p < 0.05
Oudega et al. 201422	Siemens	1	1.5 mm	Yes	VBM	MNI	Total GM volume, initial MADRS score	A priori defined brain: p < 0.001; not defined a priori: p < 0.05 (FWE)
Oudega et al. 201933	GE Signa/Philips Intera	3	3/4 mm (fMRI); 1/1.2 mm (sMRI)	No	VBM	MNI	–	p < 0.05(cluster corrected, the null distribution of max voxelwise test statistic)
Rabl et al. 202523	GE Signa	3	1.5 mm	No	FreeSurfer 5.3	–	Age, sex, modeling Wechsler Memory Scale 4th edition, antipsychotic	p < 0.05 (Bonferroni-Holm)
Salokangas et al. 200224	Siemens	1.5	5.4 mm	No	New Regional Image Analysis package	–	Overall intra-cranial volume, age, gender	p < 0.05
Simpson et al. 19996	GE Vectra	0.5	NA	Yes	CNS software package ANALYZE	Multispectral analysis	Intracranial skull volume, age	–
Sommer et al. 200734	Philips ACS-NT	1.5	3.51 mm (fMRI); 1.2 mm (sMRI)	No	VBM	MNI	–	–
Sudheimer et al. 201535	GE Signa	3	2 mm	Yes	VBM	Anatomical Automatic Labeling Atlas, Human Brain Atlas, Human Brain Anatomy in Computerized Images, Atlas of the Human Brain	Age, total symptom scores	p < 0.05 (SVC FWE)
Takamiya et al. 202225	NA	3	1.5 mm	Yes	VBM	MNI	Age, sex, total intracranial volume, site, HAMD total scores, medications	p < 0.05 (FWE)
Tani et al. 202436	Philips Achieva/GE Discovery MR750	3	1/0.9 mm (sMRI)	No	LCModel 6.3–0E	–	Age, sex, years of education	p < 0.05 (Bonferroni)
Vassilopoulou et al. 201326	Philips Intera	1.5	3 mm	No	Philips Medical Systems Release 4.1	–	–	p unknown (Bonferroni)
Voineskos et al. 202027	GE Discovery MR750/Philips Achieva/Siemens TIM Trio	3	0.9/1 mm (T1); 2.9/2.5 mm (DTI)	No	FreeSurfer 6.0 longitudinal	Desikan Killiany atlas, Johns Hopkins University white matter atlas	Sex, age, total brain volume	p < 0.0035 (primary outcome) (Bonferroni, FDR), p < 0.0125 (secondary outcome) (Bonferroni, FDR)
Yamasaki et al. 202028	Siemens	3	1 mm	Yes	VBM	MNI	Age, sex, total intracranial volume, psychotic symptoms	p < 0.05 (FWE)

Note: NA, not available; MRI, magnetic resonance imaging; GE, General Electrics; fMRI, functional magnetic resonance imaging; sMRI, structural magnetic resonance imaging; DTI, diffusion tensor imaging; VBM, voxel-based morphometry; BRAINS2, Brain Research: Analysis of Images, Networks, and Systems; MNI, Montreal Neurological Institute; AAL, Automatic Anatomical Labeling; GM, gray matter; MADRS, Montgomery-Asberg Depression Rating Scale; HAMD, Hamilton Depression Rating Scale; CIRSG, Clinical Illness Rating Scale-Geriatrics; SCID-IV, Structured Clinical Interview for DSM-Ⅳ; DSM, Diagnostic and Statistical Manual of Mental Disorders; PFC, prefrontal cortex; STG, superior temporal gyrus; FWE, family-wise error; FDR, false discovery rate; SVC, small volume corrections; GRF, Gaussian random field theory correction.

3.4. General findings

The studies investigated structural and functional differences between PD patients and NPD patients or HC, and also the relationships between brain alterations, clinical features, and treatment outcomes (pharmacotherapy or ECT).

3.4.1. Structural imaging: gray matter alterations

Eleven studies examined gray matter volume (GMV) differences among PD patients, NPD patients, and HC. The reported findings were heterogeneous across studies, likely reflecting differences in regions of interest, analytic approaches and sample characteristics. Several studies reported lower GMV in frontal regions in PD, although the specific subregions varied considerably and cross-study spatial overlap was limited. Some studies also reported increases in regional GMV following pharmacological or electroconvulsive treatment. Figure 2 summarizes these findings, and additional details are given in the Supplementary Methods and Table S2.

Fig. 2.

Anatomical distribution of GMV alterations in patients with PD. These brain maps summarize the regions with reported GMV alteration in PD, synthesized from both region-of-interest (ROI)-based and whole-brain morphometry studies. Panel A compares PD and healthy controls (HC), and Panel B compares PD and non-psychotic depression (NPD), showing each in anterior and superior views. Node size and a cyan-to-magenta color scale code the cross-study reporting frequency of each region; larger nodes and higher color intensity indicate more frequent reporting across studies. Note that the maximum overlap ranges from 1 to 2 studies, reflecting the current heterogeneity in the literature. Anatomical coordinates and detailed source information are provided in the Supplementary Materials. Abbreviations: AMYG, amygdala; DLPFC, dorsolateral prefrontal cortex; GMV, gray matter volume; HC, healthy controls; HIP, hippocampus; IFG, inferior frontal gyrus; L, left; MFG, middle frontal gyrus; MTG, middle temporal gyrus; NPD, non-psychotic depression; PCG, posterior cingulate gyrus; PD, psychotic depression; PFC, prefrontal cortex; R, right; ROI, region of interest; scACC, subcallosal anterior cingulate cortex; SFG, superior frontal gyrus; sgPFC, subgenual prefrontal cortex; STG, superior temporal gyrus.

Seven ROI-based structural MRI studies reported GMV alterations within the fronto-limbic system in adult PD patients,^{7,11,13,15,16,23,26} particularly the prefrontal cortex (PFC). Compared with HC, PD patients had lower GMV in the subgenual^7,26 and dorsolateral PFC¹⁵ and the posterior insula.¹³ These alterations were associated with clinically relevant outcomes, and a longitudinal treatment study specifically identified the subgenual PFC as a dynamic biomarker of functional prognosis and treatment response: lower baseline cortical depth was associated with poorer function, and subsequent volume recovery paralleled clinical improvement.⁷ Furthermore, structural alterations within fronto-limbic regions were associated with symptom severity and underlying vulnerability, as reflected by the negative correlation between left anterior insula GMV and psychotic symptoms,¹³ and the lower GMV in left subcallosal anterior cingulate cortex (ACC) in patients with a positive family history of affective disorders.¹¹

In contrast to these aforementioned cortical findings, evidence for subcortical volumetric alterations, particularly in the hippocampus and amygdala, remained heterogeneous. For hippocampus, most studies reported no volume differences,^15,16,23,26 only one study finding lower bilateral hippocampal volume in PD vs. HC.¹¹ For the amygdala,^11,15,16,26 results were inconsistent, studies reporting both lower¹⁶ and higher volumes²⁶ in PD vs. HC. Nevertheless, subcortical structures show distinct clinical associations. One study found a diagnosis-specific association with hypothalamic–pituitary–adrenal axis measures: in NPD the correlation between hippocampal volume and cortisol levels was associated with depressive severity, whereas in PD cortisol levels were directly associated with psychotic symptom severity.²³ Similarly, subcortical volumetric variance is associated with age of onset,^16,26 and at least in PD, with a significant group-by-age interaction:¹⁶ Keller et al. found a significant group-by-age interaction for amygdala volume that was specific to PD (younger patients showing smaller volumes), with no such age-dependent difference in NPD.¹⁶ This dynamic pattern suggests that standard linear age covariates cannot fully resolve demographic mismatch, underscoring the need to account for age and age-of-onset distributions when synthesizing subcortical volumetric findings.

Studies in elderly PD populations^17,22,25,28 corroborate prefrontal structural deficits^17,22 and highlight the prognostic relevance of baseline GMV, as well as the neuroplastic effects associated with ECT.^22,25,28 Elderly PD patients have lower PFC vs. both HC and NPD,^17,22 and lower left inferior frontal gyrus (IFG) GMV vs. NPD,²² and the cerebral-to-PFC volume ratio is negatively related to cognitive performance (Wisconsin Card Sorting Test).¹⁷ Baseline morphometry predicts ECT response: using MADRS-based response metrics, lower right IFG and higher right premotor cortex volumes predicted faster symptom improvement, whereas lower GMV in right lateral temporal cortex and right ventrolateral PFC predicted better overall clinical improvement.²² Longitudinally, ECT may drive widespread, stage-specific neuroplastic recovery. Post-ECT structural volume recovery in bilateral hippocampus, left insula, and subcallosal cingulate cortex was strongly associated with clinical trajectories: early volume increases in left superior and middle temporal gyri were associated with improved psychotic symptoms, while later volume increases in left hippocampus were associated with improved depressive symptoms.^25,28

3.4.2. Structural imaging: white matter alterations

Seven studies examined white matter alterations in PD, including white matter volume (WMV), white matter hyperintensities (WMH) (defined as regions with prolonged T2 relaxation times), and white matter microstructural integrity.^{6,12,14,19,21,24,27} In short, WMH and microstructural integrity are critical markers of cognitive decline, symptom severity, and treatment trajectory, but macroscopic WMV alterations in PD are heterogeneous.

One study reported regionally reduced frontal WMV in PD vs. NPD,²⁴ while another found no significant WMV differences in PD vs. HC.¹⁴ However, a robust pattern emerges in elderly PD populations, characterized by widespread vascular and structural burden. Three studies focused on WMH and WMV alterations in elderly PD patients:^6,19,21 compared to HC or NPD, elderly PD showed higher prevalence of deep WMH,^19,21 strongly correlated with poorer neuropsychological performance in verbal fluency, memory, and visuospatial functions;¹⁹ also, regional primary brain atrophy involving the diencephalon, brainstem, and left frontotemporal regions, as well as lesions within the reticular activating system, were strong predictors of psychotic features.⁶

Beyond macroscopic lesions, white matter microstructural integrity assessed by DTI has prognostic value for pharmacological effects and disease recurrence.^12,27 In the STOP-PD II trials, compared with those receiving placebo plus sertraline, patients receiving olanzapine plus sertraline who were in remission, relapse, or treatment discontinuation showed bigger increases in 36-week mean diffusivity (MD) across the white matter skeleton particularly in frontotemporal pathways, indicating disrupted microstructural integrity, although these effects did not survive multiple-comparison correction.²⁷ Additionally, psychomotor disturbance (measured by CORE scores) and elevated MD in the corpus callosum, left striato-frontal, and bilateral thalamo-frontal tracts significantly predicted PD relapse following remission, although there was no evidence that MD mediated the relationship between psychomotor disturbance and recurrence.¹²

3.4.3. Structural imaging: cortical thickness

Longitudinal evidence from the STOP-PD II trials revealed cortical thickness in PD as a dynamic biomarker, rather than a static index of disease burden, with close links to treatment trajectory and clinical prognosis.^18,20,27 At baseline, even remitted PD patients had widespread cortical thinning vs. HC, in the insular-limbic, occipito-temporal, temporal, parahippocampal-limbic, and inferior fronto-temporal networks; the degree of thinning within the insular-limbic network predicted symptom severity at follow-up.²⁰ The trajectory of cortical thinning is influenced by both treatment type and disease state: olanzapine plus sertraline treatment was associated with greater bilateral cortical thinning in remitted PD patients, while in recurrent PD patients cortical thinning was greater in the placebo group.²⁷ Furthermore, greater longitudinal cortical thinning predicted a lower risk of disease recurrence, independent of olanzapine dosage.¹⁸ Thus cortical thinning in PD may not simply represent neurodegeneration, but rather complex neuroplastic adaptations shaped by the interactions between pharmacological intervention and disease processes.

3.4.4. Functional imaging: task fMRI

Task-based fMRI studies in PD generally find altered hemispheric asymmetry, with right hemisphere hyperactivation across multiple cognitive domains. Rather than being task-specific, this rightward functional shift appears to reflect core pathophysiological mechanisms, potentially encompassing both pathological dysfunction and compensatory neural responses.³⁰ During language processing, PD patients showed higher right-hemisphere activation vs. HC, reflecting a degree of loss of language lateralization.³⁴ During the n-back working memory task, this aberrant right-hemisphere recruitment persisted. Specifically, during the 1-back condition, PD patients showed increased activation in right parahippocampal gyrus vs. HC. During the 2-back condition, PD patients exhibited hyperactivation in right temporoparietal junction (TPJ) vs. both NPD and HC, and higher activation in right dorsolateral PFC vs. NPD (which exhibited hypoactivation in this region). PD patients also exhibited increased functional connectivity between right TPJ and left dorsolateral PFC.³⁰

3.4.5. Functional imaging: resting-state fMRI

Resting-state fMRI studies have provided preliminary evidence for distinct patterns of functional dysconnectivity in PD, which may involve disruptions in stress-response systems and large-scale cognitive networks, as well as dynamic alterations associated with treatment.^31–33,35  Figure 3 provides a visual summary of the inter-network functional connectivity (FC) alterations reported in three of these included studies (details are provided in the Supplementary Methods and Table S3). In the stress-response system, both PD and NPD vs. HC showed lower FC between hypothalamus and subgenual PFC, accompanied by higher cortisol secretion in the circadian nadir. Lower FC between the hypothalamus and subgenual PFC differentiates depressed patients (both PD and NPD) from HC, while high circadian cortisol levels are more closely associated with symptom variability in the depressed individuals.³⁵

Fig. 3.

Reported inter-network resting-state FC findings in PD. A chord diagram illustrates significant inter-network FC findings synthesized from three included studies. Brain networks are defined according to the CAB-NP framework, comprising 12 cortical networks and one subcortical category. Link colors denote different FC comparisons: blue lines indicate lower FC in PD vs. HC, and brown lines indicate stable FC in patients receiving antipsychotic treatment (sertraline plus olanzapine) compared with placebo (sertraline plus placebo). Line thickness reflects the number of studies reporting altered FC, but does not vary as there was only one study for each connection. Given that these inter-network connectivity profiles are synthesized from only three studies, they should be interpreted as preliminary findings rather than definitive neurobiological conclusions. Abbreviations: AUD, auditory; CAB-NP, Cole-Anticevic brain-wide network partition; CON, cingulo-opercular network; DAN, dorsal attention network; DMN, default mode network; FC, functional connectivity; FPN, frontoparietal network; HC, healthy controls; LAN, language; ORA, orbito-affective; PD, psychotic depression; PMM, posterior multimodal; SMN, somatomotor network; SUB, subcortical regions; VIS1, primary visual; VIS2, secondary visual, VMM, ventral multimodal.

At the network level, preliminary findings suggest that PD may involve disrupted coordination between internal cognitive processing and external sensory input. Emerging evidence in PD suggests lower FC between the default mode network (DMN) and crucial interoceptive/exteroceptive regions, including bilateral insula, somatosensory/motor cortex, and auditory cortex, even during remission.³¹ In contrast, reports on FC within the right frontoparietal network are inconsistent.³³ The trajectory of these large-scale networks appears to be modifiable by pharmacological intervention, suggesting potential dynamic network reorganization. In a longitudinal study, combined treatment with sertraline plus olanzapine seemed to exert a protective effect on FC; in contrast, patients receiving placebo plus sertraline showed decreases in FC, particularly between the secondary visual network and several networks including the orbito-affective, cingulo-opercular, subcortical, and auditory networks.³²

Finally, measures of intrinsic brain activity may have potential value for the differential diagnosis of PD. Compared with patients with schizophrenia, PD patients have relatively preserved cognitive function, particularly in visuospatial/constructional abilities and delayed memory. This is biologically reflected in lower regional homogeneity (ReHo) in the left precuneus; the strong negative correlation between left precuneus ReHo and cognitive performance establishes it as a potential neurobiological signature for PD,²⁹ though further independent validation is required.

3.4.6. Functional imaging: magnetic resonance spectroscopy

Magnetic resonance spectroscopy studies have provided evidence for metabolic abnormalities in the dorsal ACC and potential drug-related neuroprotective effects in the STOP-PD II trials, where remitted PD patients had higher levels of myo-inositol in the dorsal ACC following pharmacological treatment.³⁶ Furthermore, treatment type influenced metabolic stability: patients receiving sertraline plus placebo showed lower creatine levels in dorsal ACC, an effect not seen in those treated with sertraline plus olanzapine, suggesting that antipsychotic medication may exert a stabilizing neurochemical effect.³⁶

4. Discussion

This review summarizes evidence from structural and functional MRI studies on neurobiological alterations in psychotic depression. The key findings can be briefly summarized. First, because of their psychotic symptoms and more severe depressive symptoms, PD patients, compared with NPD, generally show more pronounced abnormalities in gray matter, white matter, cortical thickness, and functional brain activity. Second, structural and functional alterations in PD appear to overlap in several brain regions including PFC, insula, and ACC. Structural abnormalities in these regions are often accompanied by decreased functional activity at resting state and compensatory hyperactivation during task performance, presumably to maintain behavioral performance. Structural MRI studies have implicate the fronto-limbic system as a key site of structural abnormalities in PD. Functional MRI studies in PD reveal lower FC between frontal regions, including within the right frontoparietal network and between the hypothalamus and the subgenual PFC. Third, correlations with clinical scales suggest that several neuroimaging metrics have potential prognostic value, including gray matter volume in the PFC, insula, and inferior frontal gyrus, cortical thickness within the insular-limbic network, and regional homogeneity in the precuneus. Moreover, structural and functional abnormalities may show partial reversibility following ECT or pharmacological treatment, manifested as restoration of regional brain volume or stabilization of functional connectivity patterns. Finally, although the therapeutic benefits of pharmacological treatment should not be overlooked, attention should also be paid to the potential relationship between medication exposure and disease recurrence in PD patients. We discuss some of the points in more detail below.

4.1. Disorder-specific alterations and structural–functional overlap in PD

Comparisons of PD and NPD allow identification of neurobiological markers specific to psychotic features. Compared with NPD, PD patients have more pronounced structural alterations, especially in PFC^{6,17,22,24,25} and temporal lobe.^6,28 These PFC abnormalities may provide a structural basis for the functional alterations observed in both resting-state and task-based paradigms. Task-based fMRI studies reported higher activation of right dorsolateral PFC during a 2-back working memory task in PD relative to NPD.³⁰ PD also showed hyperactivation in TPJ, which was accompanied by increased FC with left dorsolateral PFC, perhaps reflecting a compensatory top-down mechanism to maintain task performance despite sensory gating deficits due to temporoparietal junction dysfunction.³⁰ Representing an excessive expenditure of neural resources for filtering, this unique TPJ hyperactivation successfully distinguished PD from NPD independently of depression severity, thereby offering a specific functional signature for the psychotic state.³⁰ While the aforementioned working memory task revealed PD-specific functional signatures, using symptom-specific paradigms could further optimize the disentangling of depressive and psychotic dimensions. In this context, task-based fMRI studies employing the von Zerssen paranoia-depression scale provide a valuable methodological framework.^37–39 Recent cross-diagnostic studies and their independent replication³⁹ demonstrated that while psychosis (schizophrenia) and major depressive disorder recruit overlapping large-scale cognitive-emotional networks (e.g., DMN, central executive network, and salience network), they exhibit distinct modulation patterns.³⁸ For instance, processing paranoia-relevant content in psychosis uniquely engages the motor–premotor–cingulate network³⁹ and reveals aberrant effective connectivity, such as inhibitory effects from the dorsolateral PFC to the anterior insula.³⁷ Although these studies did not directly investigate PD, they provide a crucial context for exploring potential shared mechanisms of psychotic symptoms. Applying such paranoia-depression tasks to directly compare PD and NPD in future research could help determine whether the psychotic features in PD share the paranoia-related neural mechanisms of schizophrenia, or if they rely exclusively on distinct PD-specific functional abnormalities (such as the aforementioned TPJ hyperactivation). Beyond these task-evoked functional alterations, a resting-state study showed intrinsic network-level deficits, characterized by lower FC within the right frontoparietal network.³³ Furthermore, MRS evidence of increased myo-inositol levels in dorsal ACC following pharmacological treatment³⁶ suggests treatment-related metabolic alterations in this region.

Compared with HC, patients with PD show widespread abnormalities. There is lower GMV in key cognitive and affective regions, including PFC^7,15,22,26, insula,¹³ and ACC.¹¹ These structural alterations are accompanied by aberrant brain activity. During task performance, patients with PD show altered right-hemispheric activation, indicative of disrupted language lateralization,³⁴ as well as hyperactivation in the right parahippocampal gyrus.³⁰ Resting-state fMRI studies reveal pronounced FC disruptions across large-scale networks. Notably, lower FC between the subgenual PFC and hypothalamus in PD vs. HC³⁵ suggests disruption of stress-related regulatory circuits. There is a similar low FC between the insula and the DMN, particularly within interoceptive and exteroceptive processing regions.³¹

The parallel findings of structural deficits and functional dysconnectivity within overlapping regions (e.g., PFC and insula) suggest that localized structural abnormalities may disrupt network-level activity, thereby contributing to the emergence of severe affective symptoms in PD. However, these findings come from independent studies, and no study to date has simultaneously examined both structural and functional alterations within the same PD cohort. This would clearly be useful in the future.

4.2. Inconsistent reports of subcortical volume changes

The hippocampus or amygdala have been the focus of eight studies of PD.^{11,15–17,21,23,25,26} Most studies have reported no significant volume differences in either region in PD vs. either NPD or HC.^15,17,21,23 Two studies have reported lower hippocampal volume in PD.^11,25 Interestingly, hippocampal volume has been shown to increase following ECT, which may reflect treatment-related structural plasticity associated with symptom improvement.^25,28 Findings on amygdala volume are inconsistent: two studies reported opposite findings, one reporting lower¹⁶ and another reporting higher amygdala volume in PD vs HC.²⁶ Differences in sample characteristics and medication status may partly explain these discrepancies: the study reporting lower amygdala volume included patients with earlier age of onset, a more balanced sex distribution, and a small proportion of unmedicated participants;¹⁶ the study reporting higher amygdala volume included patients with later onset, a higher proportion of female participants, a greater proportion of patients in active depressive episodes, and universal pharmacological treatment.²⁶ The reliability of these reported group differences is fundamentally challenged by the notable age gap between HC and patient groups across the pooled literature. Because age interacts with the disease state (e.g., younger age correlate with smaller amygdala volumes specifically in PD¹⁶), the observed structural discrepancies may partially reflect this uncaptured age-by-disease interaction rather than uniform pathological traits. Methodological differences may also contribute to inconsistent findings, particularly the use of automated segmentation techniques compared with manually delineated ROI methods.¹¹ This highlights the need for standardized study designs and analytic methods in future studies of subcortical volume abnormalities in PD.

4.3. Symptom predictability based on correlational analyses

Correlations between MRI alterations and clinical symptoms suggest that some brain regions showing abnormalities in PD may be potential symptom-related biomarkers. For example, GMV in the left anterior insula was negatively correlated with psychotic symptom severity, with lower insular volumes being associated with more severe psychotic symptoms.¹³ In addition, longitudinal and post-ECT studies have reported associations between volume restoration in affected brain regions and clinical improvement. For example, changes in the volume of the left posterior subgenual PFC were positively correlated with GAS score, bigger increases in PFC volume being associated with better clinical outcomes.⁷ These findings suggest that structural recovery in previously affected regions may have prognostic value for treatment outcomes in PD patients.

Furthermore, distinct neuroimaging changes may occur at different stages of ECT treatment, reflecting stage-specific therapeutic effects: increases in GMV within the left superior and middle temporal gyri during the early phase of treatment were associated with improvements in psychotic symptoms, whereas later increases in left hippocampal GMV were associated with improvements in depressive symptoms.²⁸ However, despite these promising associations, many studies have failed to identify significant correlations between neuroimaging alterations and clinical symptoms, even when examining the same brain regions.^15,16,26 Replication across independent cohorts will be necessary to establish the reliability of potential imaging biomarkers which may ultimately enable individualized treatment based on objective neurobiological characteristics.

4.4. Therapeutic effects of pharmacological treatment

Evidence on white matter alterations is limited and somewhat inconsistent. Abnormal white matter changes in frontotemporal regions are reported in PD,^6,24,27 most often lower WMV in frontotemporal regions.^6,24 Following pharmacological treatment, a reported rise in MD in these regions may reflect microstructural alterations within white matter pathways.²⁷ However, higher MD in some frontal regions has also been associated with symptom recurrence in PD.¹² Further research is therefore needed to determine whether frontotemporal white matter alterations can reliably predict clinical outcomes in PD. Heterogeneity in sample size, imaging methodology, and age-related vascular burden across studies underscores the need for larger, standardized, and age-stratified investigations. Although most studies adjusted for age as a covariate, the substantial age differences between their cohorts (mean 34–75 years) should be kept in mind when comparing these white-matter findings across studies.

Seven studies from the STOP-PD II randomized controlled trials have examined both neuroimaging alterations and the effects of pharmacological treatment in PD.^{12,18,20,27,31,32,36} Structural MRI findings regarding drug-related changes and their relationship to symptom improvement or recurrence remain controversial. For example, while remitted PD patients showed increased MD in frontotemporal regions following pharmacological treatment, similar increases were also observed in patients who subsequently experienced recurrence.²⁷ Subsequent analyses reported associations between increased MD and recurrence risk.¹² A similar pattern was observed in cortical thickness measures, where symptom improvement following treatment was accompanied by reductions in cortical thickness, a finding that appears counterintuitive from a structural perspective.^20,27 Interestingly, patients who experienced recurrence without pharmacological treatment showed more pronounced structural alterations than those who relapsed despite ongoing treatment, suggesting that medication may exert a partial protective effect on brain structure.

Although subsequent studies did not find a direct association between olanzapine use and recurrence risk,¹⁸ the potential influence of pharmacological treatment on brain structure and disease progression should not be overlooked. Functional MRI findings suggest that pharmacological treatment may help stabilize functional brain networks, preventing further deterioration following medication withdrawal. For example, even in clinically remitted PD patients, FC remained abnormal, characterized by reduced connectivity between the DMN and interoceptive/exteroceptive regions compared with HC.³¹ However, longitudinal studies showed that patients receiving olanzapine maintained relatively stable FC, whereas patients receiving placebo exhibited further decrease in FC, particularly between the secondary visual network and several cognitive and affective networks.³² Together, these findings suggest that antipsychotic medication may help maintain functional network stability in PD, although its long-term structural effects require further investigation.

4.5. Limitations and future directions

Although research on PD is gathering pace, several limitations remain. First, as evidenced by the low cross-study spatial overlap, the literature exhibits a high degree of heterogeneity. Many studies include relatively small samples, which may contribute to these inconsistent findings due to variability in factors such as sex distribution, age, medication status, and disease severity. Importantly, the notable age gap between the healthy control group and patient groups (averaging about a decade) is a major potential confounder; although original studies typically adjusted for age internally, this discrepancy limits the reliability of the synthesized structural and connectivity differences between groups. Future studies should recruit larger, strictly age-matched cohorts, and more fully record baseline characteristics. Second, studies across long time periods and multiple research centers suffer from heterogeneity in imaging acquisition parameters and image quality; future research should implement standardized acquisition protocols to improve comparability. Third, most studies have focused on either structural or functional MRI measures alone, which limits full characterization of the neurobiological mechanisms underlying PD; multimodal MRI approaches integrating structural, functional, and metabolic imaging may provide a more comprehensive understanding. Additionally, although several studies have reported associations between neuroimaging alterations and symptom improvement, further research is needed to evaluate the diagnostic and prognostic value of these imaging markers for predicting treatment response. Finally, the present study was conducted as a systematic review rather than a meta-analysis, because the number of studies using comparable analytic methods was insufficient for inclusion in standard coordinate-based meta-analytic frameworks; future meta-analyses may become feasible as more studies employing similar methodologies are published, allowing quantitative integration of findings.

5. Conclusions

In conclusion, structural and functional neuroimaging studies of PD have highlighted variable regional alterations, particularly involving the prefrontal cortex, insula, and anterior cingulate cortex. Future research should prioritize standardized investigation of these regions to better understand the neurobiological mechanisms underlying PD. Potential associations between neuroimaging alterations in regions such as the PFC, insula, and inferior frontal gyrus and clinical symptoms may provide valuable information for predicting treatment response and prognosis in PD patients, though these warrant further validation. Finally, given the potential dual effects of pharmacological treatment on both symptom improvement and brain structure, future studies should aim to optimize therapeutic strategies that maximize clinical benefit while minimizing adverse neurobiological effects.

Data availability

No new data were generated or analyzed in this study.

Author contributions

Conceptualization: YZ

Data curation: YM, QZ

Formal analysis: YM, QZ

Investigation: YM, QZ

Methodology: QZ, YZ

Validation: AZ, XZ, QL, HZ

Visualization: YM, YZ

Writing – original draft: YM, QZ

Writing – review & editing: GJK, YZ

Funding information

This work was supported by the Post-Doctor Research Project of West China Hospital of Sichuan University (2021HXBH025), National Natural Science Foundation (82001795), and Sichuan Science and Technology Program (2022YFS0069, 2024YFFK0364).

Supplementary material

Supplementary data are available with the article at https://doi.org/10.1139/jpn-2026-0096.