In this study, a game-based rehabilitation system is designed to be adopted at home for a range of neurological disabilities, including stroke, traumatic brain injury, spinal cord injury, and cerebral palsy diseases. Six novel video games are designed in VR settings to enhance patient motivation. Moreover, a mobile application, namely “Recover Me,” was developed to interact with a patient and to allow remote monitoring by the physiotherapist. Each game was designed to address a specific disease and improve the physical and/or cognitive abilities of an impaired individual. Therefore, a novel rehabilitative game-based program was designed and implemented based on a VR environment. The plan incorporates a smart adaptive framework combining practicing activities and an alert system to adjust the challenges to each patient’s unique abilities. A general layout of the proposed system is depicted in Fig. 1. To adjust the physiotherapy needs for each disease, several meetings were held with a physiotherapist to indicate the actual requirements for restoring physical activities and cognitive abilities. According to these requirements, game procedures were assigned.

Fig. 1

An overview layout of the proposed rehabilitative game-based system for various neurological disorders

Game engine architecture

According to the addressed impairments, six novel games were designed to adapt the required physiotherapy activities for each disorder. The goal was to act and assess through the developed system. Each game was designed according to the gathered feedback from the physiotherapist to fulfil the rehabilitation protocol of the patient. Based on ongoing suggestions, each game was tested until the physiotherapist approved it. To promote game functionality, repetitions should be done, and according to the scores gained, the physiotherapist can assess its effectiveness. Therefore, the games enable the patients to engage in physical therapy with enjoyment.

Unity is one of the most widely used game engines, renowned for its versatility and accessibility. It is a cross-platform engine compatible with various systems, including iOS, Android, desktop platforms, virtual reality (VR), and augmented reality (AR), making it ideal for developing applications across multiple environments. Unity supports both 2D and 3D game development, offering significant flexibility in design. Game objects in Unity are created using modular components, with each component controlling a specific aspect of the game, allowing for highly customizable and intuitive development. The engine utilizes the C# programming language, which facilitates efficient scripting. Scripting defines the behavior of game objects, and Unity provides a comprehensive library of variables and functions to support the creation of complex and interactive features (Barczak and Woźniak 2019; Vohera et al. 2021; Singh and Kaur 2022).

Virtual reality tool

Virtual reality technology has widely emerged in game-based rehabilitation systems. Among a variety of utilized VR headsets, the Oculus Quest 2 is one of the best VR technologies that was developed and launched by Meta Platforms, Inc. The Oculus Quest 2 is a commonly used VR headset, valued for its affordability and ability to create an immersive experience (Raymer et al. 2023). For this research, the Oculus Quest 2 was chosen as the VR device. It includes a headset with integrated tracking sensors and two handheld controllers for user interaction. The device has a memory capacity of 6 GB RAM and storage options up to 256 GB. It features an LCD display with a resolution of 1832 × 1920 pixels per eye, providing clear visuals. Additionally, the headset supports audio functionality, enhancing its usability for gaming and other VR applications (Oculus Quest 2 Specifications, 2020).

Design of games

The Unity engine was used for scripting the designed games with the Firebase backend. Each game was developed based on the desired target of the rehabilitation program. In this study, Unity 2021.3.6f1 version was used to design all games. Six games were created to engage with a VR setting for full immersion in gaming for rehabilitation purposes. The games are “Piano Game,” “Connect Game,” “Drag Drop Game,” “Little Intelligence Game,” “Memory (2D) Game,” and “Hack & Slash Game.”. It is hypothesized that the functional exercises within each game contribute to the recovery of both physical and cognitive abilities. The details and specific characteristics of each game are explored in the following sections.

In “Paino Game (G1),” starting of the gaming scene shows a group of piano tiles. When a patient presses on a tile, a score is recorded. If s/he fails to press a tile, the game is terminated, and a total score of the game appears on the display. The aim of the game was to improve the motor skills of a traumatic brain injury patient and stroke by increasing the strength of the muscles. Moreover, it improves mental skills, such as the memory and recall ability, because the patient can play melodies and favourite songs. The game composed of three scenes as shown in Fig. 2: the start, the gaming, the ending and the submission. The start scene presents three functions; start, quit, and registration. The gaming scene depicts descending piano keys for pressing. In game design, we added the assets, such as scripts and models, based on the playing protocol of the game. Additionally, each scene was created by dragging and dropping appropriate objects from the “Assets” menu.

Fig. 2

The “Piano Game” showing the scenes of the game; “A” is the start scene, “B” is the gaming scene, and “C” is the ending scene

The “Connect Game (G2),” contains several levels of playing connection according to the difficulty grade. Seven levels were created, including novice A, novice B, regular A, regular B, advanced, expert, and master, as shown in Fig. 3. When a patient plays at a particular level and wins, the next level appears for continuity. For example, in the level” novice B”, the patient needs to connect the dots of one colour without the intersection of another colour path. The aim of the game was to enhance problem-solving skills, eye-hand coordination, and mental concentration. This game is appropriate for patients with cerebral palsy and stroke. The game starts with pressing the “play” button, then a patient selects the required level based on its difficulty. Each level is composed of 1–50 boxes to play with. By ending the level, the word “win” or “fail” appears, allowing moving to another level or terminating the game.

Fig. 3

Gaming scene of the “Connect Game,”

The “Drag & Drop Game (G3),” simulates the process of hand grip strengthening. In this game, the patient opens a group of particular shapes, and s/he is required to drag and drop any shape related to the group. Each time s/he succeeds in doing that, a score is reported. This game improves motor skills, memory, and attention as well. It is helpful to those who are poststroke. Figure 4 depicts a sample of “drag & drop game”. The example shown in Fig. 4 presents dragging and dropping of a dog. The game is composed of two scenes: starting scene and playing scene. The time of dragging and dropping was set for 1 min for each trail.

Fig. 4

Gaming scene of the “Drag & Drop Game,”

The “Little Intelligent (G4),” shows how each player can identify a particular object related to a specific class. In this way, five classes were presented in this game, namely fruits, vegetables, geometric shapes, animals, and colours. For instance, when a player opens the vegetable class, as shown in Fig. 5, and the question is, where is the carrot? A countdown timer counts until the correct or incorrect answer is indicated. If the answer is correct, a score is given. This game targets who those with poststroke, cerebral palsy injuries, and traumatic brain injuries. Three scenes were developed for the game, including the start scene, the playing scene with five levels, and the ending scene. In the playing scene, the game opens initially for level 1; otherwise for the previous level of the player.

Fig. 5

Gaming scene of one level of “Little Intelligent Game”

The “Memory Game (G5),” entails activities that improve memory and cognitive abilities. In addition, it can enhance attention and reduce stress because it focuses on improving the mood. In this game, the patient is asked to seek out information about something that relates to an object. For example, naming the shapes and their colours, and giving real examples of an underlined shape. The game allows many attempts to increase the score obtained. Figure 6 shows an example of the “Memory Game”. Alzheimer, cerebral palsy, and stroke patients are the targets of this game. Similarly to the previous games, starting, playing, and ending scenes were created. In the playing scene, the scores and the attempts appeared with the option of restarting the game.

Fig. 6

The gaming scene of “Memory Game”

The sixth game “Hack & Slash (G6),” is mainly designed to regain motor abilities. For physical rehabilitation, this game might be a helpful resource for recovery. Users can conduct exercises and activities that are challenging or unattainable to complete in the actual world by moving and controlling their avatars in a virtual setting. They will have fun participating in an engaging activity that can help them gain strength and mobility. This game necessitates repetitive exercises for effective therapy. For instance, a patient with limited mobility might be encouraged to play a game where s/he must swing a sword or use a bow and arrow to increase the mobility range. Moreover, practicing the “Hack & Slash” allows the player to explore a virtual world that impacts his/her cognitive skills. It develops problem solving and critical thinking skills. A virtual scene of the game is presented in Fig. 7. Considering that the game was designed based on three objects: the environment, the enemy, and the hero.

Fig. 7

The gaming scene of the “Hack & Slash” game

Mobile application

To communicate with patients who go under the developed rehabilitation system, a mobile application was created. The application permits each patient to select one game out of the six to play and interact with. A score index is given to each game; therefore, it can be assessed by the physiotherapist. Based on the given score, the physiotherapist can estimate the level of improvement for each participant. An Android-based application has called “Recover Me” tailors all unique needs of both the patient and the physiotherapist. The architecture of the mobile application is depicted in Fig. 8, which presents interactions between client-side (Flutter) and server-side (Firebase), showing the flows of data among key components.

Fig. 8

A block diagram of “Recover Me” application showing the flow of data

The application is open for logging in for both users (patient and physiotherapist). First, for patient log-in, each patient should register, indicating his or her disease. Therefore, the patient selects a game from the list of games to engage with a VR environment. The games are chosen based on the patient’s preference. For each trail, a score is recorded, and then a summary of the total number of trials with a score index is given through a graph. Notably, before playing, a description of each game is associated. For the physiotherapist, a generated score for each patient indicates how s/he interacts with the game, showing an indication of therapy progress level. In a risky situation, which means failing to record scores in almost all games, the physiotherapist can interface with the patient, giving instructions to interact with. It is worth mentioning that the profile of each physiotherapist includes a photo, the specialization, and a WhatsApp connection.

Artificial neural network model

The second stage of the rehabilitation system involved developing an artificial neural network (ANN) model to remotely predict the progress level of patients without the direct involvement of a physiotherapist. ANNs are computational models inspired by the structure and function of biological neural systems, consisting of numerous interconnected neurons (Saleh and Salaheldin 2022; Ye et al. 2024). The ANN was designed as a feedforward network with an input layer corresponding to game-derived features, two hidden layers (64 and 32 neurons, respectively) using ReLU activation, and a single output neuron with a linear activation function for regression. The model was trained using Mean Squared Error (MSE) as the loss function and the Adam optimizer with an initial learning rate of 0.001, a batch size of 32, and 100 epochs, incorporating early stopping to prevent overfitting. Hyperparameter tuning via grid search optimized the number of layers, neurons, activation functions, and learning rate. Regularization techniques, including dropout (0.2) and L2 regularization (0.01), were applied to enhance generalization. They are widely utilized in various applications, such as disease detection and diagnosis, due to their ability to map relationships between inputs and outputs through weighted connections, which represent the strength of these relationships (Cao et al. 2023). In this study, the ANN model was trained using the generated scores from each game, with regression techniques applied to account for the continuous nature of the training data. This approach enables ongoing monitoring and prediction of patient performance and progress levels. Consequently, the system allows patients to use the application independently without the need for real-time supervision by a physiotherapist.

Experimental setup

The developed rehabilitation system utilizes VR settings to support the patient’s recovery process. A comfortable environment with adequate space for movement is essential for optimal engagement. To evaluate the system’s effectiveness, a four-week study was conducted involving 50 patients with various disorders. The participants included 28 females (56%) and 22 males (44%), with an average age of 41.2 years. Detailed participant information, including gender, age, diagnosis, initial assessment, and injury duration, is provided in Appendix 1. Initially, each patient completed six games, scoring out of 10, to assess their baseline skill level. The order of games was determined by the patient’s preference. Following this assessment, patients participated in a structured training program lasting four weeks, designed according to a physiotherapist’s guidance. Their performance was recorded weekly, and average scores from all training sessions were computed to derive a comprehensive overall score for each patient. Patients were allowed multiple attempts to achieve satisfactory outcomes for each game.

Several challenges were encountered by patients while engaging with the designed rehabilitation system. These included:

1.

Coordinating movements Patients needed to align their physical movements with the system to effectively engage with the selected games.

2.

Adjusting eye movements Synchronizing eye movements with the VR headset posed initial difficulties for some patients.

3.

Communication with physiotherapists Ensuring effective communication to monitor performance and address concerns was essential during the sessions.