Brain-computer interface (BCI) is believed to be the translator of brain signals into actions based on the model, built on the machine learning (ML) algorithms, incorporated in it. This study reports on the performance of various ML algorithms in evaluating efficacy of neurofeedback applied for treatment of central neuropathic pain (CNP). In the first phase of this study, we applied different ML algorithms for classification of electroencephalography (EEG) patterns, associated with CNP, obtained from three groups of participants, during imagined movement of their limbs, named as able-bodied (AB), paraplegic patients with (PWP) and without (PNP) neuropathic pain. In the second phase, we tested the accuracy of BCI-classifier by applying new EEG data obtained from PWP participants who have completed neurofeedback training provided for the management of pain. Support vector Machine (SVM) algorithm gained higher accuracy, with all groups, than the other classifiers. However, the highest classification accuracy of 99 ± 0.49% was obtained with the right hand motor imagery of (AB vs PWP) group and 61 electrodes. In Conclusion, SVM based BCI-classifier achieved high accuracy in evaluating efficacy of neurofeedback applied for treatment of CNP. Results of this study show that the accuracy of BCI changes with ML algorithm, electrodes combinations, and training data set.

Depressive disorders (including both unipolar and bipolar depression) continue to present treatment challenges, with many patients failing to achieve adequate symptom relief. Non-invasive neuromodulation techniques (NINTs) have emerged as promising alternative interventions, particularly valuable in resource-limited settings. Our systematic review, drawing from PubMed and Scopus databases (2016-January 2024), evaluates current evidence on NINTs protocols for depression treatment. Key findings indicate that while most clinical applications still employ open-loop systems, closed-loop approaches utilizing EEG biomarkers (notably alpha peak frequency and frontal-midline theta power) demonstrate enhanced treatment precision. However, significant challenges remain in establishing reliable biomarker protocols for closed-loop implementation. NINTs offer particular advantages for lower middle-income countries (LMICs) through their cost-effectiveness, scalability, and minimal infrastructure requirements, though initial costs and regulatory barriers limit widespread adoption. This review highlights the urgent need for standardized protocols and identifies promising research directions to optimize NINTs' therapeutic potential. The transition from open-loop to biomarker-guided closed-loop systems represents a critical frontier in depression treatment innovation.