Rehabilitation-exercise-induced cortical reorganization mainly depends on the principle of motor learning. Motor learning has benefits for dendritic sprouting, formation of new synapses, modulation of existing synapses, and production of neurochemicals [129]. Sports learning produces remarkable results when carrying out meaningful, repetitive, and intensive rehabilitation exercises [130]. Functional recovery through motor learning can be classified into two types: substantive and compensatory motor recovery. Substantive motor recovery occurs in an uninjured or alternative pathway and transmits nerve impulses to the same muscles used before injury, while compensatory motor recovery uses other muscles different from the original muscles to achieve the same goal [131]. For patients with SCI, substantive motor recovery is the goal of functional recovery; however, it is often difficult to achieve due to damage to conduction pathways and the lack of spontaneous regeneration. Compensatory motor recovery can be achieved by means of residual neural pathways and brain central reorganization, and it is the most important type of functional recovery currently [132].

Many different rehabilitation exercises can be performed after SCI. In animal research, rehabilitation exercises can be divided into voluntary and forced exercises [133]. The most typical voluntary exercise is voluntary wheel running, where animals living alone voluntarily use running wheels for exercise. Forced exercise includes running on a treadmill and swimming in a closed swimming pool. To some extent, these two models are similar to the human rehabilitation exercise model, and they are both beneficial for regulating metabolic protein [134] and BDNF [135] and improving the degree of nervous system injury [133]. Although forced exercise can effectively quantify and reduce isolation stress, some reports suggest that voluntary exercise is beneficial for SCI rats [133] and mice [136].

For human beings, rehabilitation exercises can be classified into anaerobic and aerobic exercise, both of which can promote the change and recovery of brain structure and function after SCI. Chisholm et al. [119] found that anaerobic exercise such as resistance training can improve somatosensory sensation and corticospinal excitability in patients with motor incomplete SCI and regulate brain functional connections at rest. After resistance training, somatosensory excitability and corticospinal excitability were enhanced, and motor-evoked potentials increased [119]. The analysis of fMRI in the resting state of the patients showed that the functional connection of the motor cortex was more obvious in the less affected side after exercise. Aerobic exercise mainly refers to dynamic exercise that relies on aerobic metabolism to promote muscle contraction and the exercise of large muscle groups, including jogging, cycling, and swimming. Aerobic exercise can improve walking and balance function in patients with SCI [137]. It can also improve cerebral blood flow and neurovascular coupling to promote changes in brain structure and function [138]. In addition to SCI, aerobic and anaerobic exercise play an important role in the treatment of brain injury caused by other brain diseases. In individuals who experienced stroke, compared with the impedance exercise group, fixed bicycle aerobic training can improve the speed of information processing [139]. Although both aerobic and anaerobic exercise can independently improve the structure and function of the brain, the effects of the two forms of exercise are distinct. Combining the two modes of exercise is beneficial for improving cognitive function in the brain [139]. During rehabilitation exercises, individuals with severe motor dysfunction generally require one-to-one guidance from rehabilitation therapists to move their paralyzed limbs. This assistance promotes the recovery of muscle strength and the rebuilding of motor nerves [140]. At the same time, various instruments are necessary to support patients during rehabilitation exercises, such as weight-support treadmill training for individuals with SCI. This method reduces the biomechanical and balance constraints of walking by providing weight support through harness systems, allowing for a more normal walking pattern [141].

It is worth noting that exercise may have a negative impact. Chapman et al. [142] discovered that vigorous exercise was connected to a higher risk of amyotrophic lateral sclerosis (ALS). A 2021 study looked at roughly 10% of patients with ALS and found that in these individuals, higher rates of exercise were associated with an earlier onset of the disease [143]. At the same time, some researchers believe that the positive effects of exercise on brain function can be described by an inverted U-shaped curve, suggesting that the appropriate amount of rehabilitation exercise may result in the best possible outcome for the human body [144,145,146]. However, further evidence is needed to explore the potential mechanism and effect of this theory.

For patients with severe SCI and those who are unable to exercise autonomously, additional auxiliary means are often necessary to strengthen limb rehabilitation training, and functional electrical stimulation is proposed. Functional electrical stimulation activates muscle contraction by stimulating currents on human muscles, allowing patients to complete the corresponding movements on their own [147]. Epidural electrical stimulation (EES), which is developed from functional electrical stimulation, is a kind of rehabilitation stimulation after SCI with clinical potential. Its mode of action involves the implantation of an electrode into the epidural space of the spinal canal to make electrical stimulation act on the spinal cord tissue. EES transmits a sequence pulse current through the electrode to induce the depolarization of nerve cells in the stimulated part of the spinal cord, causing the dominant muscle to contract to drive joint movement, thereby improving the motor ability of the limbs of patients with SCI [148,149].

The use of epidural stimulation in combination with rehabilitation exercise after SCI can significantly enhance the functional recovery of patients. The first research on the use of EES technology to assist motor function recovery in patients with SCI began in the early 20th century, in which the mode of open-loop stimulation was applied [150]. When performing open-loop stimulation, the stimulator uses a constant or cyclic stimulation mode regardless of the real-time position of the limb and feedback from the brain or peripheral nervous system. This study proved that the combination of EES with exercise training can improve the motor function of patients with incomplete SCI. Since then, researchers have been exploring the closed-loop stimulation model of EES. In the closed-loop mode, EES stimulation can be triggered based on brain activity [151,152,153] or movement at different stages of the gait cycle [154,155]. Closed-loop stimulation aims to activate the target muscle while maximizing sensory inputs [156]. Wagner et al. [157] showed that after a period of closed-loop EES stimulation combined with lower limb rehabilitation training, two patients with SCI were able to autonomously control the paralyzed legs and restore certain functional activities without active stimulation. EES stimulation alone can also promote the recovery of motor and autonomic nervous function in patients with SCI [158]. In another clinical trial, seven patients were treated with EES for 5–9 months without intensive exercise, and more than half of them were able to resume autonomous exercise continuously without active stimulation after EES treatment [159]. The above studies show that epidural electrical stimulation can induce limb movement and proprioceptive sensory nerve input, which can promote the recovery of motor sensory function below the injury level and strengthen the autonomic nervous function of patients.

Electrical stimulation at the site of SCI can also regulate the pathway above the injury and activate the motor cortex [160]. Some studies have confirmed that in the mouse SCI model, EES combined with photogenetic stimulation of the motor cortex can immediately restore weight-bearing movement in mice. When the motor cortical stimulation stops, motor function is also blocked, indicating that EES induces cortical activity to control movement [161]. EES can also increase the level of BDNF in patients [162] and inhibit nerve inflammation at the injured site [163]. Although current neuroimaging studies on the effects of EES on brain structure and function in patients with SCI are not perfect, the combination of EES with repeated rehabilitation exercises plays an important role in the recovery of motor sensation after SCI and activates the ascending pathway of the spinal cord and motor cortex.

During the rehabilitation training process, clinicians need to choose the appropriate EES intensity and mode based on the patient’s condition. The implanted electrode of EES not only has risks related to surgical damage but also needs to maintain biocompatibility for a long time. Hence, the clinical application of this rehabilitation technique requires further research.

An exoskeleton rehabilitation robot is a mechanical device that simulates the structural characteristics of the human body and attaches to the exterior of the patient’s body to assist or replace the patient’s muscle strength and skeletal joint function, thereby helping patients recover or improve the function of their limbs [164]. Mechanical exoskeletons provide rehabilitation training for patients with CNS injury through their mechanical structure and control system [165], and this technique has the characteristics of high efficiency, quantification, and real-time interaction [166].

Currently, the main training modes of wearable upper limb exoskeleton rehabilitation robots include the passive, cooperative, and active modes. In the passive mode, the patient is in a completely relaxed state, and the exoskeleton drives all parts of the upper limb to the intended target position. Most mechanical exoskeletons are equipped with this mode. The cooperative mode improves upon the traditional passive mode, and the patient can drive the exoskeleton according to the trajectory of movement. When the collaboration mode is used, the exoskeleton can judge whether the exoskeleton is helpful or not based on the motion state of the patient. Patients actively participate to some extent, which plays a better role in promoting the reconstruction of the nervous system and the recovery of motor function [167]. Through rehabilitation-robot-assisted training, patients’ walking ability [168], walking speed [169,170], leg muscle strength [171], stride length [172], and gait symmetry [173] have been improved.

Rehabilitation-robot-assisted training is widely used in artificial therapy after SCI, thereby providing early, intensive, task-specific motor sensory training [174], and it can also promote the adaptive plasticity of brain and spinal cord sensorimotor networks [175]. Chintan and Simon [176] compared rats that received only running training with those that received active stepping robot-assisted running training after SCI. The results showed that the motor characteristics of the cortex of rats with robot-assisted rehabilitation training were abundant, and the motor cortex was effectively activated. Effective robot rehabilitation training can induce the reorganization of the motor cortex and partially reverse some plasticity changes. Chintan and Simon [176] used functional near-infrared spectroscopy technology, which utilizes the scattering properties of blood’s main components to near-infrared light to obtain the changes in oxygenated hemoglobin and deoxyhemoglobin during brain activity, to explore the role of robot-assisted rehabilitation in the rehabilitation of patients with SCI. The results showed that the activity of the motor cortex increased significantly during treatment. In addition to research on patients with SCI, the effect of rehabilitation robot-assisted training on cerebral cortex reorganization has also been examined. In a clinical study of rehabilitation robot-assisted gait training for 40 patients who experienced stroke, the researchers observed an improvement in the effective connection between the frontal and parietal cortex after eight weeks of Ekso^TM rehabilitation robot training compared with regular training [177]. Wagner et al. [178] studied the spectral patterns of active and passive robot-assisted walking through EEG signals. The results indicate that both active and passive robot-assisted training can stimulate the cerebral cortex. Furthermore, during active robot-assisted training, changes in cerebral cortex activation are associated with the gait cycle stage.

The positive role of exoskeleton rehabilitation robots in assisting rehabilitation training has been preliminarily confirmed. In the future, the rehabilitation mechanical exoskeleton will need to further effectively exchange dynamic information with patients to achieve man–machine fusion. This type of rehabilitation training can effectively improve the active participation of patients and significantly improve the rehabilitation effect.

In addition to rehabilitation methods that indirectly stimulate the CNS of the brain by passively acting on the distal limbs, new rehabilitation methods have been developed to directly reshape the cerebral cortex by making full use of the subjective intention of patients in order to improve the efficiency of rehabilitation. Among these methods, motor imagination (also known as psychological imagination) refers to the execution of specific actions or tasks without actual signal output [179]. Motor imagination (MI) can activate areas in the primary motor cortex, cerebellum, and basal ganglion circuits [180]. It can also induce functional redistribution and regulation of neural circuits, and then reshape the brain neural network and improve the relearning ability of motor function [181]. In the early stages of SCI, most of the structures and functions of the brain are preserved, providing a matrix for motor imagination [182]. MI therapy does not rely on the residual function of patients, which can give full play to the subjective initiative of patients with SCI and can run through the whole process of rehabilitation of patients with SCI, making it important for the functional recovery of patients with SCI [183].

Sabbah et al. [184] conducted MI training in patients with complete SCI and healthy individuals. Based on the randomized controlled trial, the activated brain regions of the two groups were generally similar during motor imagination, and the primary motor cortex, supplementary motor area, and premotor cortex were all activated, demonstrating that patients with SCI could carry out motor imagination training. Chen et al. [185] performed the right ankle dorsiflexion metatarsal flexion, motor imagination task, and motor execution task in 17 patients with incomplete SCI with partial motor dysfunction of the right ankle. The results showed that the activated brain areas were generally similar, including the bilateral supplementary motor areas and inferior frontal gyrus. However, compared with the motor execution task, the activation degree of some areas in the motor imagination task was lower. Rienzo et al. [186] found that after adding MI training to the rehabilitation program of patients with SCI, the activation of compensatory brain areas decreased, and the cortical activity was closer to the normal state, suggesting that MI training partially reversed the compensatory neural activation in patients with SCI, which was beneficial to the integration of normal neural networks. The latest study by Wang et al. [187] compared the effect of MI in children with SCI with that in healthy subjects by fMRI. The results showed that the activation degrees of bilateral paracentral lobules, auxiliary motor area, putamen, and cerebellar lobules in patients with SCI were higher than those in healthy ones.

The above studies indicate that motor imagination enhances the motor rehabilitation of patients with SCI. MI can also be combined with brain–computer interface technology, which can restore part of the patient’s motor function without relying on the normal nerve conduction pathway of the brain [188]. Motor imagination–brain–computer interface will be an important future direction for the recovery of motor function after SCI. Current research on the effects of MI mainly shows that motor imagination can improve the CNS function of patients with brain injury in the short term, but its long-term effects on brain structure and function reorganization require further investigation.

Different rehabilitation training methods have varied characteristics and effects (Table 1), and their effects on brain structure and function also differ. In clinics, the most suitable rehabilitation program is typically chosen based on the patient’s status in order to achieve the best possible outcome [189].