A 2025 multicenter pilot study evaluated a novel “reward system EEG–fMRI-pattern” neurofeedback approach (Prism for Depression™) for adults with major depressive disorder (MDD) and clinically significant anhedonia. The logic is straightforward but powerful: if anhedonia reflects a reward circuit that is under-responding to positive experience, then training the brain to reliably activate that circuit could be a mechanism-driven way to reduce symptoms.

Anhedonia is more than “feeling down.” It’s the dulling of motivation and pleasure—the coffee that tastes like hot water, the music that lands like static, the social invitation that feels like homework. Neurobiologically, converging evidence links anhedonic depression to reduced activation in the ventral striatum and broader meso–cortico–limbic reward circuitry during reward anticipation and consumption. This study sits right in that lane: it attempts to strengthen reward-related activation through repeated, structured practice.

Neurofeedback and biofeedback are forms of training that provide real-time information about physiological activity (brain activity for neurofeedback; bodily signals such as heart rate variability or skin conductance for biofeedback), helping a person learn self-regulation through reinforcement. In this trial, the feedback is not a generic “more alpha” or “less theta” target. Instead, it is anchored to a biomarker designed to reflect reward-system activation—an approach meant to be more anatomically and functionally specific to anhedonia.

If the reward system is the brain’s “yes button,” anhedonia can feel like the button is stuck. This study asks: can repeated, guided practice help unstick it?

Methods

Design and participants

The study used a prospective, single-arm, open-label design across multiple medical centers in Israel. Adults aged 22–50 with DSM-5 MDD and clinically significant anhedonia were screened, with anhedonia defined by a clinician-administered Snaith–Hamilton Pleasure Scale score (SHAPS-C) at or above a threshold level. Diagnosis was confirmed using a structured interview (MINI for DSM-5). Concomitant psychotropic medications were allowed if stable for at least four weeks before training.

A total of 49 individuals were screened and 44 enrolled; 34 completed the full training protocol and were included in the efficacy analyses.

The neurofeedback target: a reward-system EEG–fMRI “fingerprint”

Prism uses an fMRI-informed EEG model—Reward System EEG–fMRI-Pattern (RS-EFP)—developed to estimate ventral striatum-related activation from EEG time–frequency features. In other words, EEG features are weighted by model coefficients that were derived from simultaneous EEG–fMRI recordings during reward processing, producing a continuous, second-by-second signal intended to reflect reward system engagement.

Operationally, the RS-EFP is computed from six EEG electrode sites: Cz, Pz, C3, C4, P3, and P4 (referenced to FCz). The system performs time–frequency processing and applies an AI-based model to generate the RS-EFP signal in real time.

Training protocol and feedback interface

Participants completed ten neurofeedback sessions, scheduled twice weekly on nonconsecutive days. Each session used an interactive audio–visual scenario with avatars designed to probe sub-domains of reward processing—especially anticipation and consumption.

Each session consisted of five cycles (about three minutes each). Each cycle included three repeated blocks with rest periods. Within a block, the interface progressed through phases: an anticipation period (about 25 seconds), a reward consumption period (about 10 seconds), and a reward holding/maintenance period (about 20 seconds). Participants were encouraged to discover mental strategies that elicited positive affect (for example, a memory, music, or sensory imagery), with successful up-modulation changing the scenario and delivering reinforcing audio–visual rewards.

Outcomes

Primary outcome: clinician-rated depression severity (HDRS-17) change from baseline to the end of training (six weeks). Secondary outcomes included clinician-rated anhedonia (SHAPS-C) and additional depression/anxiety measures (QIDS-SR-16, PHQ-9, GAD-7), plus clinical global improvement and safety monitoring.

Results

Retention and safety

Of 44 enrolled participants, 34 completed all ten sessions (about 77% completion among those who started training). Adverse events occurred in a minority of participants and were mostly mild; importantly, only one event was considered related to the neurofeedback software (a headache) and it resolved without treatment. One serious adverse event occurred (kidney stone removal) but was unrelated to the device or training.

Depression outcomes

Clinician-rated depression severity decreased substantially by the end of treatment. The adjusted mean change on the HDRS-17 was an eight-point reduction on average, meeting a clinically substantial improvement range by the study’s prespecified interpretation. A large proportion achieved clinically meaningful symptom reduction, and a subset reached remission by end of training.

Depression-related improvements also showed up in self-report measures, with statistically significant reductions in symptom scales of depression and anxiety.

Anhedonia outcomes

Clinician-rated anhedonia (SHAPS-C) improved meaningfully, with an average reduction of roughly six points by end of training. A high proportion of completers showed large relative improvements on the anhedonia measure.

Time course and potential dose-response

Symptom improvement appeared progressive from baseline to mid-treatment to end-of-treatment, suggesting cumulative learning rather than a one-session “wow effect.” The authors noted the trajectory did not clearly plateau after ten sessions, raising the possibility that additional sessions or booster training might extend gains—an important point for future dosing studies.

Effect sizes and acceptability

Effect sizes ranged from moderate to large across outcomes, with the largest values observed on clinician-rated depression and anhedonia measures. Patient satisfaction was generally high, with most participants rating the intervention at least moderately satisfactory and indicating they would recommend it.

Discussion

This pilot study offers a clinically intriguing proposition: train a person’s reward system using an EEG signal that has been calibrated to an fMRI-defined target, and you may reduce both global depressive symptoms and the stubborn “flatness” of anhedonia. The safety profile was favorable and the completion rate was strong for a behavioral neuromodulation intervention. Clinically, the magnitude of change in clinician-rated depression and anhedonia—paired with a gradual improvement over time—fits a learning-based model of self-regulation.

What makes this approach conceptually distinct is the target. Traditional EEG neurofeedback for depression often focuses on broad cortical rhythms or asymmetries, which may be relevant but indirect. Here, the feedback signal is intended to represent reward-circuit engagement, including ventral striatum-related activity, derived from EEG features trained against simultaneous EEG–fMRI data. It’s like using a “shadow” on the wall (EEG) that has been carefully learned to track the movement of an object you can’t see directly (ventral striatum activity on fMRI). The hope is that the shadow becomes informative enough to train the object.

From a clinical angle, anhedonia is often the symptom that lingers after mood improves—especially when treatments primarily act on negative affect or anxiety. A reward-focused training protocol may offer a complementary pathway: practicing the generation and sustainment of positive affect, motivation, and approach behavior. The training blocks themselves are structured around reward anticipation, consumption, and maintenance. That matters because “wanting” and “liking” can be dissociated, and different patients may struggle more with initiation than enjoyment, or vice versa.

There are also practical advantages: the intervention is non-invasive, appears well tolerated, and builds a sense of agency. For many people, feeling that they can influence their internal state—however incrementally—can be therapeutic in itself. The interactive scenario and reinforcement schedule may help translate an internal shift (a rise in the RS-EFP signal) into a learnable skill, especially when the mental strategies are personalized.

At the same time, the study design demands humility. Without a sham control or blinded comparison group, we cannot separate the specific effects of RS-EFP training from expectancy, attention, symptom fluctuation, regression to the mean, or the motivating impact of participating in a structured program. Medication stability helps, but most participants were on antidepressants, and interactions are possible. The sample was also relatively homogeneous and not treatment-resistant by definition, limiting generalizability.

Clinically, the most compelling next step is not simply “does it work?” but “for whom, and what dose?” If improvements continue past ten sessions, booster sessions may matter. If certain profiles respond better—baseline reward responsiveness, cognitive style, comorbid insomnia, medication status—then personalization could turn a promising tool into a reliably effective one.

The deeper theme is a process-based shift: targeting specific brain–behavior systems rather than diagnosis labels. If anhedonia is a positive-valence system problem, then training the positive-valence system is a reasonable hypothesis. This study provides early, safety-forward, clinically meaningful data that supports testing that hypothesis in larger controlled trials.

Brendan’s perspective

If there’s one thing I keep coming back to in clinic, it’s that “anhedonia” is not a single knob you turn up or down. It’s more like a stereo with at least two sliders: one for wanting (anticipation, drive, initiation) and one for liking (in-the-moment pleasure, savoring, satisfaction). Some people arrive with plenty of desire on paper—goals, plans, lists—but their nervous system doesn’t deliver the spark to start. Others can initiate just fine but report that the experience lands flat, like watching fireworks through thick glass. This study’s reward-circuit framing is valuable because it encourages us to stop treating anhedonia as a moral failing (“try harder”) and start treating it as a trainable state problem (“let’s build the signal”).

Individualizing neurofeedback: mapping wanting versus liking

In practice, protocol individualization starts with assessment and listening for the phenotype.

Wanting-dominant deficits often sound like: “I don’t feel pulled toward anything,” “I can’t get going,” “everything feels effortful,” and “I procrastinate even on things I care about.” Behaviorally, these clients tend to struggle with initiation, approach, and persistence. Their self-report often highlights apathy and low motivation more than sadness.

Liking-dominant deficits tend to sound like: “I do the thing, but it doesn’t register,” “I can’t feel enjoyment,” “I’m numb,” and “good moments don’t stick.” These clients may be behaviorally active yet emotionally unrewarded; they often describe reduced savoring, reduced warmth, and rapid fading of positive emotion.

This is where I appreciate the logic of a reward-focused biomarker (like the RS-EFP in this paper) even if we don’t have it in routine EEG clinics. The real win is the conceptual target: building reliable access to a positive-valence state and learning to sustain it. We can approximate this in EEG neurofeedback by matching protocols to the bottleneck.

For wanting profiles, I often think in terms of improving initiation, cognitive energy, and approach. Depending on the person’s qEEG and symptom profile, that might mean:

• Supporting stable alertness and behavioral inhibition with SMR training (typically 12–15 Hz) at central sites (Cz, C3, C4). When initiation problems are paired with distractibility or emotional impulsivity, a calmer, more stable sensorimotor rhythm can reduce the “noise” that makes starting feel impossible.

• For clients who are under-aroused or cognitively sluggish, carefully increasing low-beta (often ~15–18 Hz) at fronto-central sites (for example Fz or FCz) can be helpful—provided anxiety and rumination are not the dominant issues. The aim here is not to create a wired state, but a usable level of activation.

• If frontal slowing is prominent (excess theta or alpha), a gentle reduction of excessive slow activity at frontal sites (Fz, F3/F4) may support executive initiation. This is always a “less is more” move; pushing too hard can backfire into irritability or sleep disruption.

For liking profiles, the bottleneck is often access to and maintenance of positive affect—being able to “let good land,” and then hold it long enough to become meaningful. Protocols that support downshifting threat and enabling parasympathetic recovery can create the conditions for pleasure to register:

• Alpha enhancement (often ~9–12 Hz) in posterior regions (POz) can support calm attentional breadth and sensory openness—states that make savoring possible.

• For clients whose nervous system is chronically guarded, training coherent midline stability (for example, smoothing excess high-beta at frontal sites like Fz/F3/F4, if present) can reduce the “background alarm” that blocks enjoyment.

• When ruminative self-focus crushes positive emotion, I sometimes prioritize strategies and protocols that reduce perseveration—often by emphasizing regulation and flexibility over simple “activation.”

A useful clinical trick is to set a micro-goal per session that matches the phenotype. For wanting, the goal might be “initiate within 10 seconds and re-initiate after distractions.” For liking, it might be “amplify and sustain a positive state for 20–30 seconds, then recover it when it fades.” Those two skills look similar on a calendar (“did neurofeedback today”), but they feel totally different inside the nervous system.

Strategy matters: training the mind that drives the brain

One of the smartest design elements in the study is that training is not “stare at a bar and will it upward.” It’s embedded in reward anticipation and consumption, nudging the participant to discover strategies that reliably increase the target signal. That maps onto what we see clinically: the brain learns faster when the strategy is concrete, personally meaningful, and emotionally believable.

For wanting deficits, strategies that leverage approach and agency tend to work: imagining a first step, rehearsing a “tiny start,” recalling a time of competence, or using action-focused imagery (movement, getting out the door, completing a small task). For liking deficits, strategies that leverage sensory detail and warmth tend to work: savoring a memory with texture and smell, imagining sunlight on skin, recalling an emotionally safe connection, or pairing imagery with music that genuinely moves the person.

Crucially, I don’t treat strategy as a pep talk. I treat it as a hypothesis test: we try a strategy, we watch what the signal does, we keep what works, and we discard what doesn’t. Over time, the client builds a personalized “menu” of reliable state-shifts.

Combining reward-focused neurofeedback with psychotherapy without turning sessions into talk therapy

A common pitfall is trying to do neurofeedback and full psychotherapy at the same time. The result is usually diluted training: too much talking, too little reinforcement, and the nervous system never gets enough repetitions to learn. The goal is integration, not simultaneity.

Here’s how I like to structure it:

• A brief, focused check-in (2–5 minutes): identify the target skill for the day (initiation vs savoring), select the strategy, and set the micro-goal.

• Training blocks as the centerpiece: keep the reinforcement loop intact. Talking is minimal and functional—coaching, not processing.

• A short debrief (5–10 minutes): what worked, what didn’t, and how to translate it into daily life.

Behavioral activation fits beautifully here because it turns reward training into action. For wanting deficits, BA becomes “schedule tiny starts,” and the neurofeedback session becomes practice for the nervous system state required to begin. ACT also integrates cleanly: the strategy isn’t “feel better,” it’s “practice a workable state while moving toward values.” CBT contributes by identifying prediction errors (“this won’t help,” “I won’t feel anything”), then using the session as an experiment that updates those predictions.

A practical way to avoid turning sessions into talk therapy is to externalize the plan. Before training, write one sentence: “Today we train initiation,” or “Today we train savoring.” After training, write one sentence: “This week I will practice the same state for 2 minutes during one real-life activity.” That’s psychotherapy in the service of learning, not a competing agenda.

A constructive critique: clinic reality versus study reality

Pilot studies like this are often cleaner than real clinics. Participants are screened, motivated, monitored, and supported. That doesn’t invalidate the findings—it just means that translation requires craftsmanship. In routine practice, comorbid insomnia, trauma histories, chronic stress physiology, and medication changes can all distort reward learning. That’s why individualization matters so much. Some clients may need a few sessions of stabilizing arousal (sleep, autonomic regulation, basic attentional control) before reward-focused training becomes possible. Others may thrive immediately with reward work.

The deeper promise here is not a single proprietary biomarker; it’s a direction: neurofeedback that is mechanistically anchored, psychologically meaningful, and clinically adaptable. A future where we can track more precise targets—reward, threat, interoception, cognitive control—without losing the human art of tailoring strategies to the person in front of us.

If we can keep that balance—rigor without rigidity, personalization without mystique—this is what I hope the future of neurofeedback looks like.

Conclusion

This multicenter pilot study suggests that reward-system–targeted neurofeedback, guided by an fMRI-informed EEG biomarker, may be a safe and promising approach for MDD with prominent anhedonia. Ten sessions delivered over roughly six weeks were associated with meaningful reductions in clinician-rated depression and anhedonia, alongside improvements in self-reported symptoms and generally high satisfaction.

The clinical story here is compelling: when anhedonia is framed as diminished reward-circuit engagement, a structured intervention that repeatedly practices reward anticipation, enjoyment, and maintenance becomes more than symptom management—it becomes training. Still, the open-label, single-arm design means conclusions must remain provisional until controlled trials clarify specific efficacy and durability.

If future randomized studies confirm these findings, reward-focused neurofeedback could help fill a stubborn gap in depression care: restoring the brain’s capacity to respond to positive experience, one trained repetition at a time.

References

Amital, D., Gross, R., Goldental, N., Fruchter, E., Yaron-Wachtel, H., Tendler, A., Stern, Y., Deutsch, L., Voigt, J. D., Hendler, T., Harmelech, T., Singer, N., & Sharon, H. (2025). Reward system EEG–fMRI-pattern neurofeedback for major depressive disorder with anhedonia: A multicenter pilot study. Brain Sciences, 15(476). https://doi.org/10.3390/brainsci15050476