An Overview of Medical Imaging-Based Brain Tumor Detection

An Overview of Medical Imaging-Based Brain Tumor Detection

Vaibhav Kumar Shrivastava

M.Tech Scholar, Department of Computer Science & Engineering, Bansal Institute Of Engineering and Technology, Lucknow, (U.P.) India vaibhavacpu@gmail.com

Abstract – Medical imaging has made significant progress by incorporating deep learning approaches to accurately segment and classify brain tumors in Magnetic Resonance Imaging. This innovative method utilizes the potential of deep learning, a particular subfield Machine Learning, to increase the detection and examination of brain tumors in MRI. The procedure consists of two essential stages accurate tumor segmentation, which entails delineating the borders of the tumor within the images, and subsequent classification based on the obtained information. This technique seeks to increase the accuracy and effectiveness of tumor detection by incorporating advanced Deep Learning (DL) methods in the processing of brain MRI. This discovery has the potential to greatly influence neurosurgeons by providing more precise diagnoses and improved treatment approaches for neurological diseases.

Right now, the topic of automatically detecting brain tumors is one that really interests people. To ascertain whether a brain tumor is present in the patient, we look at the information that the patient has given us, such as MRI images. In this particular scenario, it is our responsibility to ascertain whether the patient is suffering from a brain tumor. Giving patients a good life expectancy depends on detecting malignancies in their early stages. There is a wealth of research on identifying different kinds of brain tumors and improving the precision of such detections. In this survey, we quantify the severity of a brain tumor using the Convolutional Neural Network algorithm, which yields reliable results.

Index Terms – Brain tumor, Medical Images, Deep Learning, Tumor Segmentation, Segmented Image Classification, etc.

1. INTRODUCTION

   Brain tumors are inexplicable growths of abnormal tissue that grow out of control inside the complicated human brain. This has led researchers to try to figure out what causes them. If these growths are not stopped, they may have an immense impact on how the brain works and can even spread to other important parts, which is very dangerous. Although these tumors only make up 2% of all cancers in humans, they can be very dangerous and have a wide range of deadly complications. Because of this, life expectancy is very low, usually not more than 14 months after diagnosis [1].

   There are many ways to group these tumors, including telling the difference between main as well as secondary tumors. Beginnings of primary tumors in the cerebrum, while beginnings of secondary tumors somewhere else in body and after which it progressed to the brain, usually with a cancerous nature. In this case, we are only interested in primary cancers.

   Dr Chhote Lal Prasad Gupta

   Professor, Department of Computer Science & Engineering, Bansal Institute Of Engineering and Technology, Lucknow, (U.P.) India

   Clpgupta@gmail.com

   Gliomas, which make up 30% of adult head tumors, are the most dangerous of these and have a very bad outlook for people with serious cases. These are grouped into four grades (G1G4) based on how cancerous they are, which is determined by histopathology standards and molecular biomarkers. Grade 4 is the most aggressive.

   The other way around is with non-invasive methods, like an MRI which is safer but has trouble showing all the different features of a tumor because of the way tissues are made.

   In light of this problem, we support the creation of entirely automated Computer aided diagnostics (CAD) to make the work of doctors and radiologists easier and to make sure that choices are made quickly and correctly. These systems have done very well. They usually include dividing lesions into groups, using statistical or mathematical parameters to describe tumors, and classifying them using ML. On the other hand, traditional ML methods that depend on manually extracting features have been shown to be time-consuming and require significant experience [2].

   In order to generate detailed images of internal body components, MRI is an extremely sophisticated medical imaging method that makes use of powerful magnetic fields and radio waves. The MRI system has several essential components that operate on the principles of nuclear MRI. Presented below is a thorough and detailed explanation of the MRI.

   MRI System: The MRI system is centered on a powerful magnet that produces a robust and consistent magnetic field, which is measured in Tesla (T). Higher-field MRI systems, such as 1.5T and 3T, are frequently used in clinical settings. This magnet induces the alignment of the nuclear spins of hydrogen atoms inside the body.

   RF Coils: Around the patient's body, radio frequency (RF) coils are positioned in a strategic manner on the patient to emit radio frequency pulses that disrupt the aligned hydrogen nuclei, causing them to release signals. Different coil kinds are designed to target specific body areas, enhancing the reception of signals and improving the quality of images.

   Coils used for generating a gradient in a magnetic field: Gradient coils inside the MRI system generate varying magnetic fields along particular spatial directions. The use of gradient coils allows for precise manipulation of magnetic field intensity, which in turn allows for the spatial encoding of information. This enables the generation of highly detailed cross-sectional images.

   Device for generating radio frequency pulses: The pulse generator is responsible for creating RF pulses using the RF coils. It creates regulated pulses with specific frequency, amplitude, and duration. This manipulation affects the arrangement of hydrogen nuclei, allowing for the production of precise imaging signals.

   Computing system and console: The computer system and console serve as the central hub of the MRI system, enabling radiographers and physicians to establish imaging parameters, manage the scanning process, and evaluate obtained pictures. The computer does data processing to transform unprocessed data into images with high resolution.

   Protective and security characteristics: MRI systems are housed in shielded enclosures to mitigate external magnetic field interference. The safety elements, such as emergency stop buttons and monitoring systems, give priority to the well- being of patients and workers during imaging procedures.

   Gradient and shimming systems: Gradient systems generate essential gradient fields for spatial encoding, while shimming systems rectify magnetic field is homogeneous, guaranteeing uniformity and clarity in the resulting images.

   Software for console and image reconstruction: The console is equipped with user friendly interfaces that allow operators to input scan parameters. Image reconstruction software analyzes unprocessed data, converting it into intricate cross-sectional images of the anatomical structure being studied.

   1. Anatomy of Brain Tumours

      Malignant development consist of brain tumors, originating in brain or migrating from other body regions to brain. These tumors may originate from various categories of brain cells, including neurons, glial cells, and supportive tissue. The brain's anatomical complexity makes brain tumor research difficult for medical professionals. Understanding brain tumor anatomy is indispensable for accurate diagnosis, treatment planning, as well as patient management. The brain is a highly complex organ comprising numerous regions and structures, each performing a particular function. When the brain tumor develops, tumors locaton, size, as well as form can significantly impact the individual's symptoms and prospective treatment options[3]. To fathom the anatomy of brain tumors, it is necessary to become familiar with the various brain regions. There are four major regions of the brain: cerebrum, cerebellum, medulla, and diencephalon. The greatest portion of brain, cerebrum, is responsible of higher order cognitive processes like conceiving, memory, and voluntary movement. Cerebellum is essential for motor control as well as coordination. The area of the brain that joins the brain and spinal cord is called the brainstem and is accountable for controlling essential activities like breathing and heart rate. The diencephalon contains sensory processing and hormone regulation structures, such as the thalamus and hypothalamus.

      Brain tumors can develop in any of these regions, and the tumors location often determines its symptoms and potential complications. For instance, tumors in the cerebrum can cause personality changes, cognitive impairment, seizures, and

      motor deficits. In contrast, tumors of the cerebellum can affect coordination, balance, and fine motor abilities. Brainstem tumors can cause respiration, speech, and facial movement difficulties, whereas diencephalic tumors can interfere with endocrine functions and hormone regulation. Brain tumors may also be classified according to the origin, behaviour, and histology. Primary brain tumors are further classified as gliomas, meningiomas, pituitary adenomas, and others. However, the development of secondary brain tumors can be traced back to cancerous cells that have proliferated from other bodily areas. Neurosurgeons, neurologists, oncologists, radiologists, and other healthcare professionals are frequently required to diagnose and treat brain tumors. Advanced imaging approaches, like the MRI and the PET examinations, are vital in visualizing and characterizing brain tumors, facilitating precise diagnosis and development of therapeutic strategies.

      Fig 1 Common Brain Tumor

   2. Detection and Classification of Brain Tumor Utilizing Image Processing

      Brain tumor detection as well as classification are pivotal in medical diagnostics and treatment. With complex and potentially life-threatening brain tumors, their early detection and accurate classification are of utmost significance. These processes aid in timely intervention and personalized treatment planning and contribute to ongoing research and innovations in that field. Understanding the relevance of detecting brain tumors and classification lies in their ability to improve patient outcomes, enable informed decision-making, and pave the way for innovative therapeutic approaches. In this article, we will explore the various aspects highlighting the significance of brain tumor detection and classification, emphasizing their impact on diagnosis, treatment, monitoring, research, and patient empowerment. By understanding their significance, we can appreciate the profound implications these processes have on the lives of individuals affected by brain tumors and the medical community [4].

      The importance of early treatment, therapy preparation, and overall care of brain tumors may be attributed to the detection and categorization of brain cancers. Whether harmless (non-

      cancerous) or malignancy (cancerous), tumors on the brain are strange extensions comprising neurons. For the following reasons, it is crucial to recognize as well as classify the brain tumors efficiently:

      Early Diagnosis: The prognosis of patients with brain tumors must be improved by early identification. Early tumor detection enables quick intervention and treatment, increasing the likelihood of positive results and perhaps saving lives. Malignant brain tumors, in particular, may grow quickly and pressure nearby brain tissue, resulting in neurological effects and problems. Quickly identifying and diagnosing brain tumors helps medical practitioners to start the proper treatment plans without interruption.

      Treatment Planning: Brain tumor recognition and classification provide crucial data for therapy management. Various brain tumors need different techniques for therapy. For many brain tumors, surgical excision is a frequent therapeutic; however, the scope of the operation and requirement for further treatments like radiation or chemotherapy are contingent upon the kind and grade of the tumor. Medical experts may customize therapy regimens to the unique features of the tumor thanks to precise diagnosis and classification, which ensures the best results and reduces unnecessary treatments.

      Prognosis and Survival Prediction: Correctly classifying brain tumors is important for patient prognosis and survival prediction. Brain tumors are divided into distinct grades by the World Health Organization based on their histological traits, genomic traits, and severity. Higher-grade tumors often have a worse prognosis and a lower likelihood of survival. By correctly categorizing brain tumors, medical professionals may determine the tumors possible aggression, forecast patient outcomes, and direct conversations about possible treatments, diagnosis, and therapeutic care.

      Monitoring Disease Progression: Brain tumors are identified and classified according to their nature are essential for tracking the course of the illness. Medical personnel may evaluate how well it responded to therapy, identify recurring or advancement, and adapt their treatment plan as needed via contrasting successive scans and examining alterations in tumor size, features, and associated brain cell engagement. Assessing brain tumors often aids in enhancing treatment plans and enhancing customer service.

      Research and Development: Identifying and categorizing brain tumors aid in the continuous study and advancement of neuro-oncology. Investigators may examine tumor features, comprehend underlying biological pathways, and investigate prospective therapy targets by correctly identifying and categorizing brain tumors. The creation of innovative approaches to therapy, which includes targeting treatments, immunotherapies, also and personalized medicine strategies, is made easier by this information. The research efforts are further fueled by improvements in neurological tumor recognition and categorization methods, which open up more effective diagnostic and therapeutic alternatives.

   3. Real-world Applications

      Brain tumour detection and classification using advanced imaging technologies like MRI has pivotal real-world applications in medical diagnostics and treatment planning. These applications have a profound impact on patient care and prognosis:

      1. Early Diagnosis and Treatment Planning: Rapid and accurate identification of brain tumors, especially their grade classification, plays a critical role in devising effective treatment strategies. An automated and precise categorization of tumor grades helps healthcare professionals determine suitable treatment plans promptly, potentially extending the life expectancy of patients.

      2. Enhanced Medical Imaging Techniques: Innovations in imaging technologies, such as MRI, have revolutionized the visualization and understanding of brain tumor characteristics. These advancements enable clinicians to differentiate between various tissue types, providing a comprehensive view of the tumours extent and aiding in targeted treatment approaches.

      3. Reduced Diagnostic Risks: Automated computer-aided diagnosis systems mitigate the risks associated with invasive diagnostic procedures like histopathology. By utilizing non-invasive techniques like MRI, these systems offer a safer alternative for accurate tumor detection without subjecting patients to invasive surgeries or biopsies.

      4. Improved Medical Decision-Making: Computer-aided systems, powered by ML algorithms and DL techniques like CNNs, assist healthcare professionals and radiologists in making faster and more accurate diagnostic decisions. These systems enable the effcient extraction of tumor features and classification, facilitating better- informed medical interventions.

      5. Research and Development: The use of extensive datasets like Br35H Mask RCNN, Kaggles, BRATS supports research endeavors aimed at refining and developing more effective diagnostic tools. By employing advanced CNN models like U-net and ResNet, researchers can explore the potential of these models in classifying glioma grades across different MRI modalities, contributing to ongoing advancements in brain tumor diagnosis.

      In summary, the real-world applications of advanced imaging techniques and automated diagnostic systems in brain tumor detection and classification significantly impact patient care by enabling early, accurate diagnoses and facilitating personalized treatment strategies. These innovations pave the way for improved outcomes and better prognoses for individuals affected by brain tumors.

   4. Challenges of Study

      Detecting and categorizing brain tumors using imaging techniques used in medicine, like MRI is difficult and complex. On a number of different MRI sequences, with the T1-Weighted(T1), the T2-Weighted(T2), and the FLAIR, individual forms of brain tumors might present themselves with a variety of characteristics.

      T1-Weighted (T1) MRI: T1-Weighted images deliver superior anatomical information and are frequently utilized for brain imaging. Depending on their properties, tumors may appear hyperintense (bright) or hypointense (dark) relative to adjacent tissues on T1 images. T1 images are effective for pinpointing the position of a tumor and its intimacy to surrounding brain structures.

      T2-Weighted (T2) MRI: T2-Weighted images emphasize that fluid-filled tissues and brain tumours frequently exhibit increased signal intensity on T2-weighted images. T2 images aid in detecting oedema (swelling) surrounding the tumor, which can provide crucial diagnostic information.

      Fluid Attenuated Inversion Recovery (FLAIR) MRI: The FLAIR sequence is a sequence that inhibits the signal from the cerebrospinal fluid (CSF), thereby enhancing the visibility of lesions close to CSF spaces. FLAIR images are especially valuable for tumor detection in regions where T1 and T2 images may be less informative.

   5. Challenges in Detection and Classification of Brain Tumor

   Tumor Size and Location: Brain tumors differ in size, location, and shape, which makes their detection difficult. Some tumors may be extremely small or located in regions with complex anatomical structures, making them challenging to identify.

   Tumor Heterogeneity: Brain tumors can contain various components, such as necrosis, oedema, and active tumor regions, each exhibiting distinctive MRI sequence characteristics. To accurately classify these regions, distinctions must be made between them.

   Noise and Artifacts: Noise and artefacts in MRI scans can obscure or imitate tumor features, leading to false-positive or false-negative results.

   Interpatient Variability: Brain anatomy and tumor characteristics can vary substantially between patients, necessitating adaptable and individualized detection and classification strategies.

   Expertise and Time Constraints: The interpretation of brain MRI images requires the expertise of seasoned radiologists or neurologists. The process can be lengthy, and prompt diagnosis is essential for effective treatment.

   Data Imbalance: Obtaining a diverse dataset with a proportionate representation of various tumor types can be difficult, negatively affecting the efficacy of ML algorithms. Utilizing advanced imaging methods, creating machine learning algorithms and integrating medical information for increases in both precision and effectiveness in brain tumor detection and classification are frequently required to overcome these obstacles. Ongoing research and cooperation between medical experts and AI researchers are necessary for further development in this field.

2. LITERATURE REVIEW

   1. Survey of Existing Works

      Classifying and segregating brain tumors to aid in diagnosis has been the focus of technological advancements and research. Regardless of the advancements in DL algorithm

      development, there are still limitations and shortcomings when it comes to evaluating their efficiency. Studies have shown limitations and weaknesses; thus, these restraints persist despite their dominating influence. These gaps must be filled and the precision of brain tumor classification must be enhanced by further research. The review by Badza and Barjaktarovi´c (2020) [5], offers a detailed categorization of brain tumors utilizing a variety of approaches, encompassing both conventional ML and DL. However, its contribution is limited as it only covers material published before 2019, but subsequent studies have made diverse contributions. Furthermore, it is important to note that it has a restriction in terms of its restricted scope in covering the literature on classification and segmentation (Magadza and Viriri, 2021) [6]. In 2021, a survey was conducted to investigate brain tumor segmentation algorithms that utilize DL principles. A comprehensive analysis of the methods was performed. Nevertheless, their evaluation still has a number of shortcomings. Due to its exclusive investigation using the BRATs dataset, the restricted investigation into segmentation algorithms is a significant constraint (Rao and Karunakara, 2021) [7] identified several methods for segmenting brain tumors, including unsupervised and traditional ML and DL methods for edge-delineation, area of growth in terms of pixel intensity and non-linear feature momentum. Similar to other contributions, this study has a few limitations. With the exception of two papers, their survey focused on examining research published before 2020 in a chronological manner. This approach may be insufficient and obsolete given the current advancements. A survey was conducted in 2021 to compare unsupervised segmentation approaches like supervised tumor segmentation techniques and otsu thresholding (Pei et al., 2020) [9]. A major shortcoming was the lack of evaluation of the evaluated techniques performance and the restricted discussion of DL methodologies and biometric technologies (Almahfud et al., 2018) [9], especially for handwritten signature verification, enhance identification processes using DL neural networks for image processing. The study proposes a fine-tuned Inception V3 transfer learning model, tested against six pre-trained models, achieving superior accuracy (88%) and outperforming in precision, sensitivity, and F1-score.

   2. Brain Tumour Detection Methods

      The study utilized a Depth Wise Convolution Network (DWN) to identify change in spatial representation in MRI images. Several experiments conducted on benchmark datasets available on Kaggle (Morgello, 2020) [10]. The study employed a convolutional neural network as well as transfer learning to classify tumour from brain MRI images, enhancing performance and accuracy through various CNN frameworks and image preprocessing techniques (Bathe et al., 2021) [11]. The image undergoes a high-pass filter to improve inhomogeneities of tumour in MRI slices, which are then combined with the input MRI slices to The CNN model detects brain tumours in images using a Kaggle dataset of around 1000 images (More and Bhisikar, 2021). The image

      undergoes a high-pass filter to improve in homogeneities of tumour in MRI slices, which are then combined with the input MRI slices to The CNN model detects brain tumors in identify and segment tumor images, reducing computation time (Sarah et al., 2023). A neural network models to accurately detect brain malignancies like meningitis, glioma, and pituitary tumors, an automated method using pre-trained convolution network architectures by Ambeshwar and Ramachandran (2022).

   3. Brain Tumour Segmentation Methods

      New imaging artifacts have led to low-quality, hard-to- interpret images due to interconnected pixels wth homogeneous texture or intensity. Growing-based segmentation effectively targets distinct regions but faces challenges like resemblance criterion and noise response. Selecting suitable seeds is challenging supervised machine learning can improve brain tumour segmentation images with scattered regions (Bhandari et al.,2020). BT segmentation is a method that utilizes supervised ML algorithms, with a less important contribution from Mano and Anand (2023) and the authors developed a model for automatic MRI image segmentation for brain tumor images using DCNN known for their capability in modelling intricate biological occurrences. The authors propose a multi-context-aware deep CNN for robust segmentation and survival anticipation in brain tumor classification using structural multimodal MRI prediction accuracy (Pei et al., 2020). The classification of tumors was accomplished using 3D CNN as well as the machine learning techniques to forecast survival, making use of the data set contained within the Multimodal Brain Tumor Segmentation Challenge 2019. The DenseNet121 U-Net model has been utilized for robust segmentation and classification, achieving second place in classification results (Kumar et al.,2020). In comparison to popular deep networks such as ResNet and U- Net, the studys glioma segmentation obtained a higher Dice Score Coefficient (DSC) of 0.959. In order to identify ischemic stroke lesions, the study created a classifier- segmented network. The study developed a classifier- segmented network for identifying ischemic stroke lesions, reducing computing burden and achieving higher-quality outcomes compared to current methods. The study introduces a fully convolutional network for segmenting gliomas, improving data flow, low-level traits, and gradient flow, with an inception module for additional characteristics learning (Liu et al., 2019).

   4. Brain Tumour Classification Methods

   Unsupervised methods for high-quality image generation have gained interest in neural tumor classification due to their minimal error margins. These methodologies vary based on factors like data augmentation, pre-processing, data set use, model design, and Region Of interest (ROI) segmentation. Publicly accessible T1-weighted contrast enhanced cerebral tumor MRI is available (Cheng, 2017) [12]. The presence of such ineptitude is attributed to the elevated density. The DL surpasses traditional ML due to its ability to acquire

   hierarchical characteristics through data representation, in spite of the complex structures that make up the brain's anatomy ((Khan et al., 2020); (Kang et al., 2021)) [13-14]. DL models quickly categorize brain tumors using MRI scans, providing valuable insights into different malignancies, replacing manual feature-driven classification in brain tumor classification (D´az-Pernas et al., 2021) [15]. Structural multimodal MRIs improved brain tumor classification accuracy, demonstrating strong segmentation and ranking second in classification. Another study, the BrainMRNet deep CNN is used for brain tumour classification, with attention modules underpinning architecture and hyper-column techniques incorporating residual networks. CNN layers sequentially transfer to MRI, with a hyperactive column displaying the BrainMRNet model, developed using a crucial convolution layer approach, and each last layer containing brain layer features (Togac¸ar et al., 2020) [16] and co- evolution network was utilized to build a hybrid model for classifying brain tumors as malignant or benign. The study utilized a maximum fuzzy-sure entropy approach to segment MRI data, and used K-Nearest Neighbour (KNN) as well as SVM classifiers for classification and segmentation of brain images.

   Table 1: Overview of Literature Survey

   Reference	Objective	Methods Used	Outcome
   (Kaplan et al., 2020) [17]	Classify tumor types	Local binary patterns, and classification of types based on the extraction of features type.	The utilization of optimization in order to achieve optimal weight. Improved segmentation, and accuracy.
   (Ramzan et al., 2020) [18]	Separate tumors using Linknet.	Linknet network, CNN, and most common kinds of brain tumors.	Automatic tumor separation, CNNbased approach.
   (Javed et al., 2020) [19]	Detect brain tumor malignancy.	PCA, RST, dB-wavelets, CNN, lateral filters, and HE methods that are utilized.	Elimination of background noise, extraction of features, and CNN classification.
   (Nikam et al., 2021) [20]	Determine the ROI and categorize brain tumors.	ANN, edge detection, and adaptive thresholding are all examples.	Edge detection accuracy, tumor classification.
   (Ye and	Enhance	Texture-	Texture-based

   Yang, 2021) [21]	texture- focused tumor segmentation.	focused features, The mean DSC LOO, as well as longitudinal MRI.	segmentation, performance measurements.
   (Saleem et al., 2021) [22]	Use probabilistic NN for segmentation	PCA, classification algorithms for neural networks.	Utilization of high- dimensional data suppression and performance assessment.
   (Guan et al., 2021) [23]	Improve visual quality, tumour location proposals.	Efficient-net, image- preprocessing.	Segmentation of tumors using images from multiple modalities.

   dataset is 2D and is extensively employed for classification tasks.

3. DATASETS AVAILABILITY

   The newly developed CNN approach requires a standardized dataset that is publicly available to evaluate and quantify its efficacy in comparison to existing methods. The brain imaging community utilizes publicly available brain imaging datasets, including BRATS, Ischemic Stroke Lesion Segmentation

4. CONCLUSION

This study began by providing background knowledge on medical data, tumor recognition, tumour region delineation, and form-segmented region sequences, and highlighting the significant challenges associated with medical images in MRI. It also explored the potential applications of MRI and how they can be manifested using existing solutions. The detection as well as classification of brain Tumours are indispensable for early diagnosis and efficient treatment of these abnormal growths of brain tissue. Medical imaging, specifically MRI is most widely used non-intrusive method for assessing brain Tumours. The accuracy and efficacy of Brain Tumor detection and classification have been substantially enhanced through advances in medical imaging as well as machine learning.

In the future, developments are expected to concentrate on two key areas: enhancing accuracy through hybrid methods and leveraging the benefits of 3D volumetric data to increase productivity.

REFERENCES

1. Louis, D. N. et al. (2021), The 2021 WHO classification of tumors of the central nervous system: a summary, Neuro-Oncology 23(8), 1231 1251.

2. Ismael, S. A. A., Mohammed, A. and Hefny, H. (2020), An enhanced deep learning approach for brain cancer mri images classification using residual networks, Artificial intelligence in medicine 102, 101779.

   (ISLES), MR Brain Segmentation 2013 (MRBrainS13), Neonatal Brain Segmentation 2012 (NeoBrainS12), Medical Image Computing and Computer Assisted Intervention Society (MICCAI), Internet Brain Segmentation Repository (IBSR), and Autism Brain Imaging Data Exchange (ABDE), to ensure accurate assessment metrics. Each dataset provides a unique collection of images, allowing researchers to evaluate their performance. The BRATS dataset competition, from 2012 to 2020, includes a total of 3,470 images available for testing, training, and validation, with the number gradually increasing over time. Training data is provided in one folder and is used to train the proposed algorithm; validation data can be used to modify the algorithms; and testing data is used to analyze the outputs of the models. BraTS is a prominent benchmark in the area of segmenting brain tumors, offering data for training and evaluating new segmentation algorithms The Internet Brain Segmentation Repository is a dataset including 45 brain regions

   that have been accurately labeled by hand annotation. The MRI scans were obtained at Massachusetts General Hospital. The collection consists of T1-weighted images which have undergone skull-stripping. All individuals had a volume of 256 × 256 × 128 voxels. The CE-MRI dataset, also known as the Contrast-Enhanced MRI dataset, contains 3,064 T1 images. These images are taken from various planes, including axial, sagittal, and coronal. The dataset features tumors of three distinct types: glioma, meningioma, and pituitary tumors. All MRI images have a size of 512 × 512 pixels. The

3. P. K. Chahal, S. Pandey, and S. Goel, A survey on brain tumor detection techniques for MR images, Multimed. Tools Appl., vol. 79, no. 2930, pp. 2177121814, 2020, doi: 10.1007/s11042-020-08898-3.

4. B. F. Vedran Omeragic, Vedran Omeragic, Data Acquisition Modalities, Radiology, vol. 243, no. 3, pp. 656667, 2013, doi: 10.1148/radiol.2433060243.

5. Badza, M. M. and Barjaktarovi´c, M. (2020), Classification of brain tumors from mri images using a convolutional neural network, Appl. Sci. 10(6), 1999.

6. Magadza, T. and Viriri, S. (2021), Deep learning for brain tumor segmentation: A survey of state-of-the-art, J. Imag. 7(2), 19.

7. Rao, C. S. and Karunakara, K. (2021), A comprehensive review on brain tumor segmentation and classification of mri images, Multimedia Tools and Applications 80(12), 1761117643.

8. Pei, L., Vidyaratne, L., Rahman, M. M. and Iftekharuddin, K. M. (2020), Context aware deep learning for brain tumor segmentation, subtype classification, and survival prediction using radiology images, Scientific Reports 10(1), 19726.

9. Almahfud, M. A., Setyawan, R., Sari, C. A. and Rachmawanto, E. H. (2018), An effective mri brain image segmentation using joint clustering (k-means and fuzzy c-means), in 2018 International seminar on research of information technology and intelligent systems (ISRITI), IEEE, pp. 1116.

10. Morgello, S. (2020), Coronaviruses and the central nervous system, J. NeuroVirology 26(4), 459473.

11. Bathe, K., Rana, V., Singh, S. and Singh, V. (2021), Brain tumor detection using deep learning techniques, Proceedings of the 4th International Conference on Advances in Science and Technology (ICAST2021).

12. Cheng, J. (2017), Brain tumor dataset, https://figshare.com/articles/dataset/brain_tumor_dataset/1512427.

13. Khan, M. A., Ashraf, I., Alhaisoni, M., Damasevicius, R., Scherer, R., Rehman, A. and Bukhari, S. A. C. (2020), Multimodal brain tumor classification using deep learning and robust feature selection: A machine learning application for radiologists, Diagnostics 10(8), 565.

14. Kang, J., Ullah, Z. and Gwak, J. (2021), Mri-based brain tumor classification using ensemble of deep features and machine learning classifiers, Sensors 21(6), 2222.

15. D´az-Pernas, F. J., Mart´nez-Zarzuela, M., Ant´on-Rodr´guez, M. and Gonz´alez-Ortega, D. (2021), A deep learning approach for brain tumor classification and segmentation using a multiscale convolutional neural network, Healthcare 9(2), 153.

16. Togac¸ar, M., Ergen, B. and C¨omert, Z. (2020), Brainmrnet: Brain tumor detection using magnetic resonance images with a novel convolutional neural network model, Medical hypotheses 134, 109531.

17. Kaplan, K., Kaya, Y., Kuncan, M. and Ertunc¸, H. M. (2020), Brain tumor classification using modified local binary patterns (lbp) feature extraction methods, Medical hypotheses 139, 109696.

18. Ramzan, F., Khan, M. U. G., Rehmat, A., Iqbal, S., Saba, T., Rehman,

    A. and Mehmood, Z. (2020), A deep learning approach for automated diagnosis and multiclass classification of alzheimers disease stages using resting-state fmri and residual neural networks, Journal of medical systems 44, 116.

19. Javed, R., Rahim, M. S. M., Saba, T. and Rehman, A. (2020), A comparative study of features selection for skin lesion detection from dermoscopic images, Network Modeling Analysis in Health Informatics and Bioinformatics 9, 113.

20. Nikam, R. D., Lee, J., Choi, W., Banerjee, W., Kwak, M., Yadav, M. and Hwang, H. (2021), Ionic sieving through one-atom-thick 2d material enables analog nonvolatile memory for neuromorphic computing, Small 17(44), 2103543.

21. Ye, F. and Yang, J. (2021), A deep neural network model for speaker identification, Applied Sciences 11(8), 3603.

22. Saleem, S., Amin, J., Sharif, M., Anjum, M. A., Iqbal, M. and Wang, S.-

    H. (2021), A deep network designed for segmentation and classification of leukemia using fusion of the transfer learning models, Complex and Intelligent Systems 8, 116.

23. Guan, Y., Aamir, M., Rahman, Z., Ali, A., Abro, W. A., Dayo, Z. A. and Hu, Z. (2021), A framework for efficient brain tumor classification using mri images, AIMS 18, 57905815.