The most advanced BCI systems use “closed-loop” strategies in which the implanted devices have in-built read–response mechanisms and embedded algorithms that automatically adjust simulation-related factors to match the patient's needs (Lee et al., 2015). These devices sense the composition of the patient and stimulate signals only when required, thereby reducing the side-effects during treatment. This can help conserve power and thus minimize battery replacement surgeries. These systems profoundly impact the clinical pathways of patients with complex nervous system diseases. This strategy has been used for mobility assistance—for example, for wheelchair control or rehabilitation purposes, such as for controlling an electrical stimulator.

A closed-loop BCI system usually involves a control paradigm, measurement, processing, prediction, and feedback from the application (Ahn et al., 2014) as shown in Figure 1. These functions help understand or modulate the user's mental condition or intention. The system uses these data to execute some predefined functions to interact with the environment by seeing, hearing, or sensing the action. An external device provides communication and neurofeedback, which enables the user to determine how well they can control the device. Closed-loop BCI can signal to the brain (e.g., through feedback via electrical/magnetic stimulation, and optogenetic and sonogenetic techniques) to correct the action or obtain additional information about it.

The control paradigm is the input provided by the BCI user to generate brain signals related to their intent. The user generates this input via mental tasks, including the kinematical/visual imagining of the physical movement of a body part, or by concentrating on a specific object to generate a P300 wave. Some BCI systems (e.g., affective BCI) do not require the users' intentions. Still, they function by identifying the emotional and mental statuses of the user, and can be active, reactive, or passive (Gürkök et al., 2012). In any case, these generated signals must be measured via invasive or noninvasive techniques. Invasive techniques, including ECoG, microelectrode arrays, and single microelectrodes, can detect signals from the brain's surface and produce high-quality signals. However, these techniques require risky implantation surgeries.

Therefore, BCI research commonly uses noninvasive methods to detect signals (e.g., magnetoencephalography and functional-magnetic resonance imaging). EEG is the most popular and preferred technique in this regard (Nicolas-Alonso and Gomez-Gil, 2012) as it is economical and portable, and can even be measured by using wireless devices. The most popular closed-loop BCI is the one based on motor imagery, where the disabled can receive a piece of additional sensory information from the BCI device. This BCI-based control system of closed-loop brain stimulation can help sense the effect of a stimulus and adjust this stimulation in response to the observed effect. The closed-loop BCI can be used to create synaptic plasticity through spike-triggered stimulation.

BCIs allow users to communicate with external or control any external devices, such as robotic arms, with their thoughts by monitoring brain activity from the motor cortex and decoding movement intentions using machine learning techniques which leads to significantly improve the quality of life of healthy and unhealthy people such as patients with spinal cord injuries or other paralysis to restore some motor functions (Gao Q. et al., 2017; Belkacem et al., 2018, 2020; Al-Nuaimi et al., 2020; Belkacem, 2020; Shao et al., 2020; Chen et al., 2021a,b; Jamil et al., 2021). Closed-loop brain stimulation can also assist motor learning and enhance the recovery of motor function in stroke patients and those with motor deficits (Xu et al., 2013).

BCIs and closed-loop brain stimulation can be used to treat a variety of neurological and psychiatric conditions, such as Parkinson's disease, epilepsy, depression, and obsessive-compulsive disorder (OCD) (Liang et al., 2010; Widge and Moritz, 2016; Arlotti et al., 2021; Sani et al., 2021). For instance, in Parkinson's disease, closed-loop brain stimulation can administer electrical stimulation to the brain in response to aberrant neural activity, alleviating symptoms such as tremors and stiffness. By detecting seizure activity and applying focused brain stimulation to avoid seizures, BCIs can also treat epilepsy.

Moreover, closed-loop BCI can improve cognitive performance in healthy persons. BCIs can increase attention and working memory, for instance, by monitoring brain activity linked with these cognitive processes and giving the user feedback. By giving tailored brain stimulation during certain cognitive activities, closed-loop brain stimulation can also be utilized to improve cognitive performance (Jamil et al., 2022).

Insomnia and sleep apnea are two examples of sleep problems that can be treated using BCIs and closed-loop brain stimulation. By administering tailored brain stimulation to encourage breathing or alertness, closed-loop brain stimulation may be used to detect and respond to sleep-related events such as apnea episodes. This type of brain stimulation can be used to detect and respond to sleep-related events. BCIs may also improve sleep quality by identifying sleep-related neural activity and giving tailored brain stimulation to enhance the sleep state. This can be accomplished by monitoring the neural activity that occurs during sleep (Choi et al., 2020).

The measured brain signals are then processed for the maximum signal-to-noise ratio. The selection is performed via certain algorithms (Bashashati et al., 2007; Roman-Gonzalez, 2012) including those for spatial and spectral filtering, to derive information from these signals, which are then used as inputs for the classification modules. The steps of prediction involve making decisions based on the user's intention or by quantifying their emotional and mental statuses. Machine learning algorithms and artificial neural networks are usually applied for prediction (Bashashati et al., 2007; Al-Ani et al., 2010). Once the user's intention has been determined, this output is used to change the environment. This change is then provided as feedback to the user. Figure 2 shows the control and communication involved in closed-loop brain feedback or stimulation: its methods, applications, and effects.

User experience and neural feedback are very important in closed-loop BCIs as they provide information regarding the success of the effort. Furthermore, neurofeedback allows users to improve BCI control through self-regulation and learning. In neurorehabilitation, this feedback is linked to neuroplastic stimulation that can be used to modify and modulate the user's neural activity. Neurofeedback training is the conscious alteration of brain signals by the user (Boyd et al., 2017). The BCI user learns to control their brain activities based on measurements of the brain activity and feedback signals. For example, in case of diseases like stroke, the rhythm of the brain slows down and there is a substantial decrease in cortical activity, resulting in motor and cognitive impairments (Boyd et al., 2017; Kim and Winstein, 2017). Neurofeedback training aims to restore these brain signals to facilitate a more standard state, possibly leading to functional recovery (Remsik et al., 2016). Neurofeedback can be visual, auditory, and tactile. It can be used for various ailments, including depression, anxiety, stress, pain, attention deficit hyperactivity disorder, cognitive impairments, insomnia, schizophrenia, and motor recovery.

Neurofeedback, is a form of biofeedback, a noninvasive therapeutic approach that detects a patient's brain activity and delivers real-time feedback regarding how the brain works. Neurofeedback treatment is used to train patients to control their brain processes by showing them how the brain reacts to certain stimuli. It involves analyzing brain activity and offering instant feedback, frequently via visual or audio signals.

Attention deficit hyperactivity disorder (ADHD), anxiety, depression, and post-traumatic stress disorder (PTSD) have all been treated with neurofeedback, a therapy method that uses real-time brain activity displays to teach people how to regulate their brain activity. Several studies have shown that neurofeedback can assist people with these diseases manage their symptoms and improve their quality of life. For instance, a meta-analysis of trials looking at neurofeedback for ADHD discovered that it significantly reduced hyperactivity/impulsivity and improved attention, which effects persisted over time (Arns et al., 2009). It is important to note, though, that not all studies have found neurofeedback to be effective, and some have found its effects to be either temporary or restricted (Sonuga-Barke et al., 2013).

Auditory feedback (AF) is any audible output that helps the user interact better with the system. It is used during speech learning in infants. AF can be used in BCI training by manipulating the audio input to achieve the target outcome. Two commonly used manipulations are delayed auditory feedback (DAF) and altered auditory feedback (AAF). In DAF, there is no change in the auditory signal, but it is sent to the listener after a short delay. In AAF, some sections of the signal, such as its pitch or structure, are manipulated and then sent back to the listener without delay. DAF has been shown to improve fluency among people who stutter (Yates, 1963; Ryan and Van Kirk, 1974). With regard to AAF, shifts in the pitch and other parameters in the direction opposite to that of vocal compensation occur automatically, with no conscious awareness, in neurologically healthy speakers (Houde and Jordan, 1998; Liu et al., 2011).

Visual and auditory feedback has been used for many BCI modalities (P300 speller, SSVEP, and motor imagery). It can easily cause brain fatigue, dizziness, nausea, and other adverse reactions. Most visual BCIs are based on flickering stimuli, and continuous flickering can cause visual fatigue and reduce the user's comfort. Auditory BCI is not widely used due to its susceptibility to environmental interference and relatively low accuracy. In addition, it cannot protect the user's privacy from surrounding systems. Tactile feedback is an alternative with many advantages over visual/auditory feedback, such as the ability to generate ideal target signals without repeated training. For instance, BCI-based MI (MI-BCI) allows users to communicate via the imaginary movements of their extremities by using a computer.

Although MI-BCI represents a promising strategy for control, it uses visual feedback to teach the user about the system's decisions. This makes it challenging to use with visually interactive tasks. Indeed, MI-BCIs have rarely been used outside the laboratory (Jeunet et al., 2015) dowing to their flawed classification algorithms and the difficulty of sight-based learning. These mechanisms simulate the sensation of tapping as a response to touch. The response is realized via vibrations. However, device operation is not interrupted by tactile feedback (Lukoyanov et al., 2018).

Several studies have explored tactile (vibration) stimulation in MI-BCI (Cincotti et al., 2007; Leeb et al., 2013; Gwak et al., 2014). The replacement of visual feedback by vibrotactile feedback does not inhibit EEG measurements in MI-BCI (Leeb et al., 2013) and thus does not negatively affect the classification accuracy (Cincotti et al., 2007; Leeb et al., 2013). However, this replacement reduces the visuomotor load during the tracking of multiple objects (Gwak et al., 2014). Thus, MI-BCI provides tactile feedback, and pays more attention to the problem and less to the feedback (Cincotti et al., 2007), to ensure a high accuracy of classification.

In the event of an injury or a disease that compromises a particular area of the brain, neural prosthesis (neuroprosthetics) can be used to restore function in the affected area, whether it is motor, sensory, or cognitive. Swann et al. (2018) showed that fully implanted neural prostheses can be used to produce adaptive deep brain stimulation for patients of Parkinson's disease. The nanocognitive device called an “endomyccorhizae-like interface” (the future of neuroprosthetics) was created to improve features of the neural network in people with neurodegenerative diseases such as Alzheimer, Parkinson's, or dementia (Saniotis et al., 2018).

Cochlear implants are the most widely used among neuroprosthetics. Individuals with moderate-to-profound and severe sensorineural hearing loss (SNHL) benefit more from cochlear implants than hearing aids. Cochlear implants directly stimulate the auditory nerve, avoid injured cochlear hair cells, and offer salient coded information for better speech perception (Buchman et al., 2020). When the retina is damaged, it can cause visual problems and lead to total blindness in extreme situations. The retina is the part of the eye linked to the brain. It contains photoreceptors that generate electrical signals from light, which are then transmitted to the brain via the optic nerve. Many other bionic eye systems have been developed in recent years, including artificial silicon retina that uses a silicon chip containing solar-cell microphotodiodes. These photodiodes convert light energy into electrical impulses sent to the brain via the optic nerve (Buchman et al., 2020; Suresh, 2020).

Neurostimulation is the intentional modulation of the activity of the nervous system by using invasive methods, such as microelectrode implantation, or noninvasive techniques, such as transcranial magnetic stimulation (TMS). Neurons in the brain collaborate in vast networks to regulate and coordinate activities of the body, such as seeing, listening, sensing, and feeling to perform actions, and control respiration and pulse. An electrical signal is produced by the neuron whenever it is stimulated, and can be changed through TMS by applying an electromagnetic coil to the scalp. The electromagnet provides a magnetic pulse that activates nerve cells in the mood regulation- and depression-related areas of the brain in a harmless manner. It is thought to stimulate brain areas that have diminished activity in case of depression.

Deep brain stimulation (DBS) is an invasive neurostimulation technique that requires surgery to implant a neurostimulator device that sends electrical signals to specific parts of the brain responsible for body movements. Electrodes are placed on the right and left sides of the brain, and are connected via long wires. They are then are placed under the skin, traveling down the neck, and are connected to a battery-powered stimulator placed under the skin near the chest (Artusi et al., 2018). The patient can use a handheld controller to control the DBS system. The stimulation settings can be adjusted per the patient's condition. DBS can be used to treat both movement-related and psychiatric disorders. Further, it has shown therapeutic success for otherwise treatment-resistant activity-related and affective disorders, such as tremors, dystonia, Parkinson's disease, chronic pain, and psychiatric disorders (such as depression, bipolar disease, obsessive–compulsive disorder, and Tourette's syndrome; Lozano et al., 2019). DBS has recently been considered to regulate action in memory circuits, where this indicates its potential for therapeutic use to treat dementia and Alzheimer's disease (Freund et al., 2009; Kuhn et al., 2015; Mirzadeh et al., 2016). Different DBS targets have been used to treat patients with Alzheimer's disease and yielded promising outcomes, including slower cognitive decline and improved functional brain connectivity (Laxton et al., 2010; Lozano and Lipsman, 2013). As the biological history of neurodegeneration cannot be reversed in humans, DBS may serve as supplementary treatment by controlling memory circuits (Lv et al., 2018). DBS transmits electrical impulses to the area of the brain responsible for movement-related symptoms caused by Parkinson's disease. Electric impulses disrupt these symptoms, resulting in abnormal activity in the brain's circuitry. In people with Parkinson's disease, three brain regions responsible for motion control are targeted by using DBS: the subthalamic nucleus, the globus pallidus internus, and the thalamic ventral intermediate nucleus. Targeting a specific brain area in this case depends on the treatable symptoms. Transcranial direct-current stimulation (tDCS) is a non-invasive technique for brain stimulation widely used in clinical trials for neurological and psychiatric disorders. It can mitigate depression by stimulating nerves of the left dorsolateral prefrontal cortex (Shiozawa et al., 2014).

Optogenetics is a new approach that combines optics and genetics to control the activity of certain neurons. The key element of optogenetics is the use of light. Many studies have implemented optogenetics in models of diseases, such as epilepsy, Alzheimer, Parkinson's, sensory system degeneration, and depression.

Epilepsy is a prevalent neurological illness marked by seizures. Tønnesen et al. (2009) demonstrated that light-induced halorhodopsin activity can suppress epileptiform activity while hyperpolarizing the primary neurons of the hippocampus. The optogenetic activation of hippocampal neurons at 10 or 20 Hz induces seizure-like after-discharges in rats given doses of ketamine and xylazine as anesthesia (Osawa et al., 2013). Furthermore, Paz et al. (2013) made closed-loop devices that can stop seizures by stimulating the brain with light in real time.

With regard to Alzheimer's disease, Wang et al. (2019)demonstrated that optogenetics can be utilized to modulate the neuronal-glial network to improve memory in mice with Alzheimer's disease. In addition, optogenetics has been used to analyze the activation of a particular neural circuit in transgenic mice with the amyloid precursor protein (APP) to determine the causal relationship between synaptic activity and β-amyloid peptide (Aβ) disease (Yamamoto et al., 2015).

Optogenetics was used to examine the graft function and graft–host connection for Parkinson's disease (Steinbeck et al., 2015). Moreover, Magno et al. (2019) revealed dopamine-depleted male mice benefit from optogenetic stimulation of the secondary (M2) motor cortex in case of Parkinson's. Fougère et al. (2021) claimed that the optogenetic stimulation of glutamatergic neurons in the cuneiform nucleus improves locomotion, regulates speed, and elicits limb motions comparable to those seen in intact animals during spontaneous locomotion.

Optogenetics, in conjunction with behavioral paradigms, has frequently been utilized in rats to understand the significance of various types of neurons and their pathways in people with depression (Biselli et al., 2021). An immediate effect can be mediated in the model of depression by laser or LED light to activate the expressed channelrhodopsin to obtain the specific target of dopamine neurons in the ventral tegmental area (VTA) in mice by applying the optogenetic technique (Chaudhury et al., 2013; Tye et al., 2013). Therefore, optogenetics can be used to regulate neurons to cure neurodegenerative and psychiatric diseases.

Recent years have witnessed a rise in interest in targeted ultrasound techniques for neurodegenerative and psychiatric illnesses. Sonogenetics, which is noninvasive and has a high spatial resolution, involves the genetic manipulation of ultrasound-sensitive neurons and their unique responses to it through the development of mechanosensitive receptors. Fan et al. (2021) demonstrated the proof of concept for the therapeutic application of sonogenetics to slow neurodegeneration in animal models of Parkinson's disease. Leinenga and Götz (2015) claimed improvements in Alzheimer's memory tests by using repeated scanning ultrasound treatments on the mouse brain to eliminate Aβ, without requiring any extra therapeutic agent.

Sonogentics technology that combines deep penetration with a regionally focused ultrasound has evolved, with major therapeutic applications to seizure, depression, and Parkinson's disease (Leinenga et al., 2016). Duque et al. (2021) experimentally stimulated neurons in a mammalian brain by using ultrasound and observed the changes in behavior. Ultrasound is one of the effective tools in physiotherapy, surgery, chemotherapy, medication administration, and sonography (Mason, 2011). Neurostimulation improves the quality of life of patients with severe paralysis, sensory loss, and chronic pain. It plays a vital role in neuroprosthetics, such artificial organs as bionic eyes and limbs, and cochlear implants. Further, it can alter disease symptoms in cases where medications cannot be used owing to their severe side-effects. It is also an option for many movement disorders, with relatively minimal side-effects.

The research questions (RQs) considered in this review, along with the rationales for them, are shown in Table 1.

This review adhered to the guideline for Preferred Reporting Items for Systematic Reviews and Meta-Analyses (Liberati et al., 2009). The following digital databases were searched: IEEEXplore, Scopus, and PubMed. The search began between early and mid-September 2022, and was constrained to the title of each paper, its abstract, and keywords to lower the number of results. The following was the primary search string used to find the relevant literature: (“closed loop BCI” OR “brain–computer interface” OR “brain–machine interface” OR BCI) AND (sonogenetic* OR optogenetic* OR “deep brain stimulation” OR “transcranial magnetic stimulation” OR “transcranial direct current stimulation” OR TMS OR DBS OR tDCS) AND (vision OR hearing OR motor OR sensory OR dementia OR Alzheimer OR Parkinson OR depression OR anxiety OR “psychiatric disease” OR neurodegenerat*). The search string was customized for each database. We eliminated duplicate publications after checking the titles of the papers. The titles and abstracts of all publications were evaluated to ensure their relevance. To establish the legitimacy of each article, we obtained and screened the complete text of all pertinent papers by using the criteria for inclusion. Figure 3 shows the process of retrieving articles for this review.

Articles were included in this review if they met the following inclusion criteria (IC): (IC1) a primary study that used and focused on brain stimulation (TMS, DBS, tDCS, optogenetic, sonogenetic); (IC2) articles that focused on closed-loop models; (IC3) articles that concluded an improvement, enhancement, or the restoration of the brain function of the patients; and (IC4) experiments involving healthy or unhealthy participants or animals.

The exclusion criteria (EC) for this review were as follows: (EC1) articles that were non-peer-reviewed, abstracts, survey paper, and review papers; (EC2) non-English articles; (EC3) articles that focused only on EEG, MEG, ECoG, or fMRI; (EC4) articles that focused on improving or comparing machine learning or some algorithm; (EC5) articles related to ethics or organization; and (EC6) articles that could not be retrieved in full.

A full-text article was obtained for each study that satisfied the inclusion criteria with the assistance of a librarian. We retrieved the features of the articles, such as the following:

The number of papers published annually in the area from 2000 to 2022 is shown in Figure 4. From 2000 to 2009 (excluding 2000 and 2007), there was no publication in the area. However, since 2010, the number of publications rose gradually until 2018. During 2019, the number of publications dramatically dropped to two, is likely multifactorial and complex possibilities such as changes in legislation or constraints on research may have contributed to a decline in 2019 publication totals. Regulatory bodies likely implemented tougher restrictions for closed loop brain stimulation research, resulting in a temporary decline in publications. Additionally, the advent of new study topics or a shift in research goals may have diverted resources away from closed loop brain stimulation research, resulting in a temporary decline in the number of publications. However, there was a substantial increase again starting in 2020, and the number of published papers in 2022 can increase further as the year has not ended. The largest number of published articles were from journals, with 11 papers in 2018. By contrast, the largest number of papers published in conference proceedings in any given year was two.

Figure 5 provides information on the ratio of published papers based on the type of brain stimulation studied. The overall trend of the data shows a steep rise in the ratio of TMS in the published papers, coupled with a decline in papers on tDCS. The most highest was that 27 journals had published these articles in TMS, which is a sharp contrast with only three papers in conferences. We found no publication on sonogenetics for closed-loop BCI.

The numbers of published papers on different techniques of brain stimulation were different in conferences from those in journals. The largest number of conference papers were related to tDCS (five), followed by TMS (three). No conference paper related to tACS, optogenetics, and sonogenetics had been published. The most popular techniques of brain stimulation for closed loop BCI models from 2000 to 2022 were TMS, tDCS, and DBS.

Figure 6 depicts the distribution of articles with regard to the techniques of brain stimulation used and the types of diseases considered. They latter have been divided into the two clusters listed below:

Overall, Figure 6 shows the potential of how the closed-loop BCI based on brain stimulation systems improves the quality of life of patients. Many researchers did experiments based on the type of disease to prove that brain stimulation is one method that can restore, replace, or repair impaired brain functioning and alleviate symptoms in individuals suffering from various neurological disorders. Results have demonstrated that closed-loop DBS is superior to open-loop DBS in symptom management and in reducing adverse effects. For instance, DBS decreased the intensity of tremors in individuals who suffered from essential tremors while simultaneously decreasing the stimulation-induced adverse effects. For example, Liang et al. (2010) demonstrated that closed-loop DBS could detect and suppress epileptic seizures in real-time. Based on Figure 6 also, the selected articles show the significant clinical experiments as the practice to demonstrate the improvement in individuals with a variety of illnesses and conditions. As research into this field develops, it is anticipated that more effective and individualized treatments will become available to patients in the hope that they will have a better quality of life.

Figure 7 was constructed using the metadata from the retrieved articles. The most frequently occurring terms among the 832 keywords are human (56 times) and humans (44 times). These words must be extensively used since they constitute the foundation of the subject (patient) for experiments. This can conclude that many of researchers aims to improve the life for human who have neurodegenerative and psychiatric diseases.

Table 2 presents the mapping between the technique of stimulation used and the participants in the selected studies. Only such animals as mice or rats were used in experiments on optogenetic techniques. Healthy human participants were the most commonly considered, in 41 articles, even though they were recruited in experiments for only TMS and tDCS. By contrast, non-animal was used in TMS and tDCS brain stimulation techniques. Participants with stroke or Parkinson's diseases were the most participants in the experiments for all stimulation techniques except optogenetic.

Table 3 shows examples of studies that focused on different techniques of brain stimulation to improve the quality of life of patients with neurodegenerative and psychiatric diseases based on close-loop BCI. We selected only one example publication for each technique of brain stimulation based on the type of publication (the top journal with the highest impact factor) and the participants involved.