Overview of data- and theory-driven MDD treatment response prediction studies. Relevant model performance metrics (BAC and R2) that were not reported by the studies but were possible to calculate from the reported values are included in parentheses. MDD - major depressive disorder, TRD - treatment-resistant depression, LLD - late-life depression, SSRI - selective serotonin reuptake inhibitor, rTMS - repetitive transcranial magnetic stimulation, SNRI - serotonin-norepinephrine reuptake inhibitor, NDRI - norepinephrine-dopamine reuptake inhibitor, CBT - cognitive behavioral therapy, ECT - electroconvulsive therapy, BDI - Beck Depression Inventory, HAMD - Hamilton Rating Scale for Depression, MADRS - Montgomery-Asberg Depression Rating Scale, QIDS-SR - Quick Inventory of Depressive Symptomatology—Self Report, SOFAS - Social and Occupational Functioning Assessment Scale, CIDI - Composite International Diagnostic Interview, LCI - Life Chart Interview, STFT - short-time Fourier transform, EMD - empirical mode decompositions, ACC - anterior cingulate cortex, FC - functional connectivity, ROI - region of interest, ICA - independent component analysis, BOLD - blood oxygen level dependent, dlPFC - dorsolateral prefrontal cortex, amPFC - anterior medial prefrontal cortex, mPFC - medial prefrontal cortex, DMN - default mode network, SN - salience network, PCC - posterior cingulate cortex, AN - affective network, VIS - visual, SVM - support vector machine, ROC - receiver operating characteristic, CV - cross-validation, LOOCV - leave-one-out cross-validation, Acc - accuracy, SE - sensitivity, SP - specificity, BAC - balanced accuracy, AUC - area under curve, RMSE - root mean square error.
| Reference . | Subjects . | Treatment . | Modality . | Outcome definition . | Features . | Model . | Validation . | Performance . |
|---|---|---|---|---|---|---|---|---|
| Mumtaz et al. (2017) | 34 MDD | SSRIs (type not specified) | tbEEG: 3-stimulus visual oddball task | Response: ≥50% ↓ in BDI-II | Combination of wavelet coefficients, STFT, & EMD features | Logistic regression | 10-fold CV | Acc 92% |
| SE 90% | ||||||||
| SP 90% | ||||||||
| N. Bailey et al. (2018) | 39 TRD | rTMS | tbEEG; Working memory task | Response: ≥50% ↓ in HAMD-17 | Baseline and Week 1 MADRS scores, task accuracy & reaction time, alpha, theta, gamma power & connectivity, theta gamma coupling | SVM | 5-fold CV | BAC 91% |
| SE 90% | ||||||||
| SP 92% | ||||||||
| Miller et al. (2013) | 17 MDD | SSRI (escitalopram) | tbfMRI; emotional words task | Δ in HAMD-24 | Clusters whose active during negative word processing was associated with treatment outcome | Linear regression | 10-fold CV | (R2 = 0.23) r = 0.48, p < 0.05 |
| Godlewska et al. (2018) | 32 MDD | SSRI (escitalopram) | tbfMRI: emotional faces task | Response: ≥50% ↓ in HAMD-17 | Mean cluster activity within ACC for sad vs. happy faces | Single-feature ROC | LOOCV | BAC 72% |
| Fu et al. (2008) | 19 MDD | SSRI (fluoxetine) | tbfMRI; implicit sad facial affect recognition task | Remission: HAMD-17 ≤ 8 | Whole-brain FC for each facial expression intensity (low, medium, high) | SVM | LOOCV | Low intensity of sad facial expressions: |
| SE 75% | ||||||||
| SP 62% | ||||||||
| p value = 0.11 (n.s.) | ||||||||
| Fonzo et al. (2019) | 115 MDD (sertraline), 122 MDD (placebo) | SSRI (sertraline) | tbfMRI; Emotional conflict task | Δ in HAMD-17 | incongruent trials–congruent trials brain activation in several ROIs | Relevance vector machine | 10 × 10 stratified CV | (R2 = 0.24) r = 0.49, p < 0.001 |
| Karim et al. (2018) | 49 LLD | SNRI (venlafaxine) | rsfMRI and tbfMRI; emotional regulation & emotional reactivity task | Remission: MADRS ≤ 10 | Active regions during emotional reactivity, emotion regulation, or centrality at baseline & after single dose | Logistic regression | 10-fold CV | (BAC 70%) |
| AUC 77% | ||||||||
| SE 72% | ||||||||
| SP 68% | ||||||||
| Williams et al. (2015) | 80 MDD | SSRI (escitalopram, sertraline) or SNRI (venlafaxine) | tbfMRI; Supraliminal & subliminal facial emotion task | Response: ≥50% ↓ in HAMD-17 | Emotion vs. neutral amygdala activation | Discriminant analysis | LOOCV | All medication types: Acc 75% |
| SNRI only: Acc 77% | ||||||||
| Goldstein-Piekarski et al. (2016) | 70 MDD | SSRI (escitalopram, sertraline) or SNRI (venlafaxine) | tbfMRI; emotional faces task | Functional remission: HAMD-17 ≤ 7 and QIDS-SR ≤5 and ≥10 improvement to achieve ≥61 on SOFAS | Early life stress & anygdala reactivity | Logistic regression | LOOCV | (BAC 77%) |
| SE 84% | ||||||||
| SP 69% | ||||||||
| AUC 81% | ||||||||
| Crane et al. (2017) | 29 MDD | SSRI (escitalopram) or SNRI (duloxetine) | tbfMRI; Go/No-Go task | Remission: HAMD score < 8 | No-Go accuracy, two ICA component beta weights, & within-component clusters | Logistic regression, Random forest | LOOCV | Logistic regression: |
| Acc 90% | ||||||||
| (BAC 90%) | ||||||||
| SE 90% | ||||||||
| SP 89% | ||||||||
| Random forest: | ||||||||
| Acc 84% | ||||||||
| BAC (82%) | ||||||||
| SE 84% | ||||||||
| SP 80% | ||||||||
| Tozzi et al. (2020) | 124 MDD | SSRI (escitalopram, sertraline) or SNRI (venlafaxine) | tbfMRI; Go/No-Go task | Response: >50% ↓ in QIDS-SR-16 | ROIs with BOLD response in No-Go > Go condition in MDD patients & healthy controls | Logistic regression | LOOCV | Venlafaxine response: |
| BAC 79% | ||||||||
| SE 67% | ||||||||
| SP 89%; | ||||||||
| Sertraline response: | ||||||||
| BAC 84% | ||||||||
| SE 95% | ||||||||
| SP 74% | ||||||||
| Marquand, Mourão-Miranda, Brammer, Cleare and Fu (2008) | 20 MDD | SSRI (fluoxetine) | tbfMRI; n-back task | Response: ≥50% ↓ in HAMD-17 | Principal components based on whole-brain activity for each task condition | SVM | LOOCV | BAC 69% |
| SE 85% | ||||||||
| SP 52% | ||||||||
| Meyer et al. (2019) | 22 MDD | SSRI (escitalopram) | tbfMRI; n-back task | Remission: MADRS ≤ 5 Nonremission: MADRS ≥ 10 | dlPFC, amPFC, & parietal lobe | Single-feature ROC | LOOCV | (BAC 87%) |
| AUC 85% | ||||||||
| SE 82% | ||||||||
| SP 91% | ||||||||
| Nguyen et al. (2019) | 37 MDD | NDRI (bupropion) | tbfMRI; reward processing task | Δ in HAMD-17; Remission: HAMD-17 score < 7 | ROIs from anticipation contrast maps & reward expectation contrast maps | Dense feedforward neural networks | 3 × 3 nested CV | R2 = 0.26 |
| RMSE = 4.71 | ||||||||
| AUC 71% | ||||||||
| Brandt et al. (2021) | 90 MDD | SSRI (escitalopram); an optional switch to SNRI (duloxetine) from Week 4 | tbfMRI; reward processing task | HAMD-6 Remission: >50% ↓ at week 4 and <5 at Week 8; Nonresponse: <25% ↓ a Week 4 and <50% ↓ at Week 8 | Age, sex, baseline HAMD-6 and number of omissions in the task, & reward-related brain responses in striatum, anterior insula, & mPFC | Logistic regression | 5-fold CV | AUC 56% |
| Costafreda, Khanna, Mourao-Miranda, and Fu (2009) | 16 MDD | CBT | tbfMRI; sad facial expression task | Remission: HAMD-17 ≤ 7 | Principal components based on whole-brain activity to viewing sad faces | SVM | LOOCV | BAC 79% |
| SE 71% | ||||||||
| SP 86% | ||||||||
| Cook et al. (2020) | 129 MDD | SSRI (escitalopram) or NDRI (bupropion) | rsEEG | Remission: HAMD-17 ≤ 7 | Theta and alpha power values from FT7-FPz and FT8-FPz channels at baseline and Week 1 (QEEG) | Single-feature ROC analysis | Jack-knife CV | Escitalopram remission: |
| Acc 64% | ||||||||
| (BAC 65%) | ||||||||
| SE 74% | ||||||||
| SP 55% | ||||||||
| Erguzel et al. (2015) | 55 TRD | rTMS | rsEEG | Response: ≥50% ↓ in HAMD-17 | Frontal cordance calculated from delta & theta bands (QEEG) | Artificial neural network | k-fold CV | Acc 85–89% |
| (BAC 86–89%) | ||||||||
| SE 87–94% | ||||||||
| SP 84% | ||||||||
| AUC 87–91% | ||||||||
| Patel et al. (2015) | 33 LLD | SSRI (escilatopram) or SNRIs (duloxetine or venlafaxine) | rsfMRI; sMRI | Response: HAMD score < 10 | Demographics, cognitive ability scores, functional connectivity index of dorsal DMN and anterior SN, & structural imaging variables | Alternating decision tree (ADTree) | Nested LOOCV | Acc 89% |
| SE 89% | ||||||||
| SP 90% | ||||||||
| Pei et al. (2020) | 98 MDD | SSRIs (mainly escitalopram) & SNRIs (mainly venlafaxine) | rsfMRI | Response: ≥50% ↓ in HAMD-6 | 14 priori brain regions of interest based on previous literature | SVM | LOOCV | BAC 81% |
| SE 78% | ||||||||
| SP 84% | ||||||||
| Sikora et al. (2016) | 29 MDD | 1-week placebo & 10-week open-label antidepressant (SSRI, SNRI, NDRI, atypical) | rsfMRI | Δ in QIDS | Baseline SN rsFC | Multivariate relevance vector regression | LOOCV | Placebo response: (R2 = 0.17) R = 0.41, p value = 0.018; Antidepressant response: r = 0.03, p value = 0.340 |
| Braund et al. (2022) | 229 MDD | SSRI (escitalopram, sertraline) or SNRI (venlafaxine) | rsfMRI | Response: ≥50% ↓ in HAMD-17 or QIDS-SR | Whole-brain network intrinsic FC associated with neuroticism | SVM | LOOCV | Acc 75% |
| (BAC 74%) | ||||||||
| SE 63% | ||||||||
| SP 85% | ||||||||
| AUC 76% | ||||||||
| Goldstein-Piekarski et al. (2018) | 75 MDD | SSRI (escitalopram, sertraline) or SNRI (venlafaxine) | rsfMRI | Remission: HAMD-17 ≤ 7 | Seed-based posterior cingulate cortex | Logistic regression | LOOCV | PCC–ACC/mPFC: |
| (BAC 78%) | ||||||||
| SE 73% | ||||||||
| SP 82% | ||||||||
| AUC 77% | ||||||||
| C. T. Wu et al. (2018) | 22 TRD | Sham-rTMS | rsfMRI | Δ in HAMD-17 | Two feature sets: 1. Global brain activity; 2. Rostral ACC seed-based | Elastic-net regression | LOOCV | Global brain activity: (R2 = 0.24) r = 0.49; p = 0.023; rostral ACC FC: (R2 = 0.25) r = 0.50; p = 0.018 |
| Cash et al. (2019) | 47 MDD | rTMS | rsfMRI | Response: ≥25% ↓ in MADRS; Δ in MADRS | rsFC within the DMN & AN | SVM | LOOCV, k-fold CV | Acc 93% |
| SE 95% | ||||||||
| SP 92 % | ||||||||
| (R2 = 0.46) | ||||||||
| r = 0.68 | ||||||||
| p value < 0.001 | ||||||||
| Hopman et al. (2021) | 70 TRD | rTMS | rsfMRI | Response: ≥50% ↓ in MADRS | Seed-based analysis with left dlPFC and subgenual ACC | SVM | Nested CV | Acc 89% |
| Moreno-Ortega et al. (2019) | 18 TRD | ECT | rsfMRI | Remission: 24-HAMD score ≤ 7 | dlPFC, DMN, & VIS networks | Logistic regression | LOOCV | Acc 89% |
| Frässle et al. (2020) | 85 MDD | SSRIs or no treatment | tbfMRI; emotional face perception task | Long-term outcomes: remitted vs. improved vs. chronic, based on a latent class growth analysis using CIDI and LCI | Effective connectivity pattern in the network mediating emotional face perception | SVM | Nested LOOCV | Chronic vs. remitted: |
| AUC 87% | ||||||||
| BAC 79% | ||||||||
| SE 97% | ||||||||
| SP 60% | ||||||||
| Improved vs. remitted: | ||||||||
| AUC 63% | ||||||||
| BAC 61% | ||||||||
| SE 77% | ||||||||
| SP 45% | ||||||||
| Queirazza et al. (2019) | 25 MDD | Computerized CBT | model-based fMRI; RL task | Response: >50% ↓ in BDI-II | Neural activity encoding weighted reward prediction errors | SVM | LOOCV | AUC 82% |
| BAC 72% | ||||||||
| SE 63% | ||||||||
| SP 83% |
| Reference . | Subjects . | Treatment . | Modality . | Outcome definition . | Features . | Model . | Validation . | Performance . |
|---|---|---|---|---|---|---|---|---|
| Mumtaz et al. (2017) | 34 MDD | SSRIs (type not specified) | tbEEG: 3-stimulus visual oddball task | Response: ≥50% ↓ in BDI-II | Combination of wavelet coefficients, STFT, & EMD features | Logistic regression | 10-fold CV | Acc 92% |
| SE 90% | ||||||||
| SP 90% | ||||||||
| N. Bailey et al. (2018) | 39 TRD | rTMS | tbEEG; Working memory task | Response: ≥50% ↓ in HAMD-17 | Baseline and Week 1 MADRS scores, task accuracy & reaction time, alpha, theta, gamma power & connectivity, theta gamma coupling | SVM | 5-fold CV | BAC 91% |
| SE 90% | ||||||||
| SP 92% | ||||||||
| Miller et al. (2013) | 17 MDD | SSRI (escitalopram) | tbfMRI; emotional words task | Δ in HAMD-24 | Clusters whose active during negative word processing was associated with treatment outcome | Linear regression | 10-fold CV | (R2 = 0.23) r = 0.48, p < 0.05 |
| Godlewska et al. (2018) | 32 MDD | SSRI (escitalopram) | tbfMRI: emotional faces task | Response: ≥50% ↓ in HAMD-17 | Mean cluster activity within ACC for sad vs. happy faces | Single-feature ROC | LOOCV | BAC 72% |
| Fu et al. (2008) | 19 MDD | SSRI (fluoxetine) | tbfMRI; implicit sad facial affect recognition task | Remission: HAMD-17 ≤ 8 | Whole-brain FC for each facial expression intensity (low, medium, high) | SVM | LOOCV | Low intensity of sad facial expressions: |
| SE 75% | ||||||||
| SP 62% | ||||||||
| p value = 0.11 (n.s.) | ||||||||
| Fonzo et al. (2019) | 115 MDD (sertraline), 122 MDD (placebo) | SSRI (sertraline) | tbfMRI; Emotional conflict task | Δ in HAMD-17 | incongruent trials–congruent trials brain activation in several ROIs | Relevance vector machine | 10 × 10 stratified CV | (R2 = 0.24) r = 0.49, p < 0.001 |
| Karim et al. (2018) | 49 LLD | SNRI (venlafaxine) | rsfMRI and tbfMRI; emotional regulation & emotional reactivity task | Remission: MADRS ≤ 10 | Active regions during emotional reactivity, emotion regulation, or centrality at baseline & after single dose | Logistic regression | 10-fold CV | (BAC 70%) |
| AUC 77% | ||||||||
| SE 72% | ||||||||
| SP 68% | ||||||||
| Williams et al. (2015) | 80 MDD | SSRI (escitalopram, sertraline) or SNRI (venlafaxine) | tbfMRI; Supraliminal & subliminal facial emotion task | Response: ≥50% ↓ in HAMD-17 | Emotion vs. neutral amygdala activation | Discriminant analysis | LOOCV | All medication types: Acc 75% |
| SNRI only: Acc 77% | ||||||||
| Goldstein-Piekarski et al. (2016) | 70 MDD | SSRI (escitalopram, sertraline) or SNRI (venlafaxine) | tbfMRI; emotional faces task | Functional remission: HAMD-17 ≤ 7 and QIDS-SR ≤5 and ≥10 improvement to achieve ≥61 on SOFAS | Early life stress & anygdala reactivity | Logistic regression | LOOCV | (BAC 77%) |
| SE 84% | ||||||||
| SP 69% | ||||||||
| AUC 81% | ||||||||
| Crane et al. (2017) | 29 MDD | SSRI (escitalopram) or SNRI (duloxetine) | tbfMRI; Go/No-Go task | Remission: HAMD score < 8 | No-Go accuracy, two ICA component beta weights, & within-component clusters | Logistic regression, Random forest | LOOCV | Logistic regression: |
| Acc 90% | ||||||||
| (BAC 90%) | ||||||||
| SE 90% | ||||||||
| SP 89% | ||||||||
| Random forest: | ||||||||
| Acc 84% | ||||||||
| BAC (82%) | ||||||||
| SE 84% | ||||||||
| SP 80% | ||||||||
| Tozzi et al. (2020) | 124 MDD | SSRI (escitalopram, sertraline) or SNRI (venlafaxine) | tbfMRI; Go/No-Go task | Response: >50% ↓ in QIDS-SR-16 | ROIs with BOLD response in No-Go > Go condition in MDD patients & healthy controls | Logistic regression | LOOCV | Venlafaxine response: |
| BAC 79% | ||||||||
| SE 67% | ||||||||
| SP 89%; | ||||||||
| Sertraline response: | ||||||||
| BAC 84% | ||||||||
| SE 95% | ||||||||
| SP 74% | ||||||||
| Marquand, Mourão-Miranda, Brammer, Cleare and Fu (2008) | 20 MDD | SSRI (fluoxetine) | tbfMRI; n-back task | Response: ≥50% ↓ in HAMD-17 | Principal components based on whole-brain activity for each task condition | SVM | LOOCV | BAC 69% |
| SE 85% | ||||||||
| SP 52% | ||||||||
| Meyer et al. (2019) | 22 MDD | SSRI (escitalopram) | tbfMRI; n-back task | Remission: MADRS ≤ 5 Nonremission: MADRS ≥ 10 | dlPFC, amPFC, & parietal lobe | Single-feature ROC | LOOCV | (BAC 87%) |
| AUC 85% | ||||||||
| SE 82% | ||||||||
| SP 91% | ||||||||
| Nguyen et al. (2019) | 37 MDD | NDRI (bupropion) | tbfMRI; reward processing task | Δ in HAMD-17; Remission: HAMD-17 score < 7 | ROIs from anticipation contrast maps & reward expectation contrast maps | Dense feedforward neural networks | 3 × 3 nested CV | R2 = 0.26 |
| RMSE = 4.71 | ||||||||
| AUC 71% | ||||||||
| Brandt et al. (2021) | 90 MDD | SSRI (escitalopram); an optional switch to SNRI (duloxetine) from Week 4 | tbfMRI; reward processing task | HAMD-6 Remission: >50% ↓ at week 4 and <5 at Week 8; Nonresponse: <25% ↓ a Week 4 and <50% ↓ at Week 8 | Age, sex, baseline HAMD-6 and number of omissions in the task, & reward-related brain responses in striatum, anterior insula, & mPFC | Logistic regression | 5-fold CV | AUC 56% |
| Costafreda, Khanna, Mourao-Miranda, and Fu (2009) | 16 MDD | CBT | tbfMRI; sad facial expression task | Remission: HAMD-17 ≤ 7 | Principal components based on whole-brain activity to viewing sad faces | SVM | LOOCV | BAC 79% |
| SE 71% | ||||||||
| SP 86% | ||||||||
| Cook et al. (2020) | 129 MDD | SSRI (escitalopram) or NDRI (bupropion) | rsEEG | Remission: HAMD-17 ≤ 7 | Theta and alpha power values from FT7-FPz and FT8-FPz channels at baseline and Week 1 (QEEG) | Single-feature ROC analysis | Jack-knife CV | Escitalopram remission: |
| Acc 64% | ||||||||
| (BAC 65%) | ||||||||
| SE 74% | ||||||||
| SP 55% | ||||||||
| Erguzel et al. (2015) | 55 TRD | rTMS | rsEEG | Response: ≥50% ↓ in HAMD-17 | Frontal cordance calculated from delta & theta bands (QEEG) | Artificial neural network | k-fold CV | Acc 85–89% |
| (BAC 86–89%) | ||||||||
| SE 87–94% | ||||||||
| SP 84% | ||||||||
| AUC 87–91% | ||||||||
| Patel et al. (2015) | 33 LLD | SSRI (escilatopram) or SNRIs (duloxetine or venlafaxine) | rsfMRI; sMRI | Response: HAMD score < 10 | Demographics, cognitive ability scores, functional connectivity index of dorsal DMN and anterior SN, & structural imaging variables | Alternating decision tree (ADTree) | Nested LOOCV | Acc 89% |
| SE 89% | ||||||||
| SP 90% | ||||||||
| Pei et al. (2020) | 98 MDD | SSRIs (mainly escitalopram) & SNRIs (mainly venlafaxine) | rsfMRI | Response: ≥50% ↓ in HAMD-6 | 14 priori brain regions of interest based on previous literature | SVM | LOOCV | BAC 81% |
| SE 78% | ||||||||
| SP 84% | ||||||||
| Sikora et al. (2016) | 29 MDD | 1-week placebo & 10-week open-label antidepressant (SSRI, SNRI, NDRI, atypical) | rsfMRI | Δ in QIDS | Baseline SN rsFC | Multivariate relevance vector regression | LOOCV | Placebo response: (R2 = 0.17) R = 0.41, p value = 0.018; Antidepressant response: r = 0.03, p value = 0.340 |
| Braund et al. (2022) | 229 MDD | SSRI (escitalopram, sertraline) or SNRI (venlafaxine) | rsfMRI | Response: ≥50% ↓ in HAMD-17 or QIDS-SR | Whole-brain network intrinsic FC associated with neuroticism | SVM | LOOCV | Acc 75% |
| (BAC 74%) | ||||||||
| SE 63% | ||||||||
| SP 85% | ||||||||
| AUC 76% | ||||||||
| Goldstein-Piekarski et al. (2018) | 75 MDD | SSRI (escitalopram, sertraline) or SNRI (venlafaxine) | rsfMRI | Remission: HAMD-17 ≤ 7 | Seed-based posterior cingulate cortex | Logistic regression | LOOCV | PCC–ACC/mPFC: |
| (BAC 78%) | ||||||||
| SE 73% | ||||||||
| SP 82% | ||||||||
| AUC 77% | ||||||||
| C. T. Wu et al. (2018) | 22 TRD | Sham-rTMS | rsfMRI | Δ in HAMD-17 | Two feature sets: 1. Global brain activity; 2. Rostral ACC seed-based | Elastic-net regression | LOOCV | Global brain activity: (R2 = 0.24) r = 0.49; p = 0.023; rostral ACC FC: (R2 = 0.25) r = 0.50; p = 0.018 |
| Cash et al. (2019) | 47 MDD | rTMS | rsfMRI | Response: ≥25% ↓ in MADRS; Δ in MADRS | rsFC within the DMN & AN | SVM | LOOCV, k-fold CV | Acc 93% |
| SE 95% | ||||||||
| SP 92 % | ||||||||
| (R2 = 0.46) | ||||||||
| r = 0.68 | ||||||||
| p value < 0.001 | ||||||||
| Hopman et al. (2021) | 70 TRD | rTMS | rsfMRI | Response: ≥50% ↓ in MADRS | Seed-based analysis with left dlPFC and subgenual ACC | SVM | Nested CV | Acc 89% |
| Moreno-Ortega et al. (2019) | 18 TRD | ECT | rsfMRI | Remission: 24-HAMD score ≤ 7 | dlPFC, DMN, & VIS networks | Logistic regression | LOOCV | Acc 89% |
| Frässle et al. (2020) | 85 MDD | SSRIs or no treatment | tbfMRI; emotional face perception task | Long-term outcomes: remitted vs. improved vs. chronic, based on a latent class growth analysis using CIDI and LCI | Effective connectivity pattern in the network mediating emotional face perception | SVM | Nested LOOCV | Chronic vs. remitted: |
| AUC 87% | ||||||||
| BAC 79% | ||||||||
| SE 97% | ||||||||
| SP 60% | ||||||||
| Improved vs. remitted: | ||||||||
| AUC 63% | ||||||||
| BAC 61% | ||||||||
| SE 77% | ||||||||
| SP 45% | ||||||||
| Queirazza et al. (2019) | 25 MDD | Computerized CBT | model-based fMRI; RL task | Response: >50% ↓ in BDI-II | Neural activity encoding weighted reward prediction errors | SVM | LOOCV | AUC 82% |
| BAC 72% | ||||||||
| SE 63% | ||||||||
| SP 83% |