1

A Review : Prediction of Tumor from Brain MRIs

using Different Segmentation Approaches
1st Amit Thakur, 2nd Pawan Kumar Patnaik
Phd Scholar, Department of Computer Science & Engineering, Bhilai Institute of Technology, Durg, India
Associate Professor, Department of Computer Science & Engineering, Bhilai Institute of Technology, Durg, India
amitthakur744@[Link], pawanpatnaik37@[Link]

Abstract—Technology advancements have the power to impact sion tomography (PET), and electroencephalography, are now
every area of human existence. Technology has, for instance, used to diagnose brain tumors (EEG). These imaging methods
significantly benefited human civilization when used in medicine. are used to view the brain and look for any potential abnormal
In this piece, we concentrate on using technology to help treat
brain tumors, one of the most prevalent and deadliest illnesses growths [4]. In addition to imaging methods, machine learn-
[Link] of the malignant tumors that is identified most often in ing algorithms have drawn increasing attention as a way to
people of all ages is a brain [Link] brain MRI data, the most increase the precision and effectiveness of brain tumor identi-
sophisticated deep learning technique, a CNN (Convolutional fication [5]. Large volumes of imaging data may be analysed
Neural Network), was utilised to find a tumor. There are still by these algorithms to find patterns and characteristics that are
problems with the laborious training process, [Link] of
the most difficult components of classifying brain tumors is suggestive of brain tumors. The identification of brain tumors
identifying the kind of tumor and avoiding [Link] this paper, may still be difficult, despite advancements in imaging and
we did a thorough analysis of the existing attempts to apply machine learning technology, since certain tumors may not
various deep learning techniques to MRI data and identified have obvious symptoms or may be hard to tell apart from
the domain’s current obstacles before identifying possible future other neurological diseases. Therefore, there is a constant need
[Link] learning networks’ exponential expansion has
made it possible for humans to handle challenging jobs, even in for study and the creation of fresh methods to increase the
the intricate area of medicine. This article provides an overview of precision and efficiency of brain tumor identification.
studies on the segmentation of brain tumors using MRI images This review paper’s objectives are to provide a summary of
published between 2018 and 2023. However, in order for the the status of brain tumor detection today and to explore the
networks to be highly generalizable and with good performance, many strategies and techniques that have been used. The goal
applying these models needs a big corpus of data. For researchers
in the biological and machine learning sectors, several routes for of the article is to emphasise the benefits and drawbacks of
future study are finally presented. the existing approaches for detecting brain tumors and to point
out areas that need more study. The review article will also
Index Terms—Review, CNN, Deep Learning, Brain Tumor,
Feature based prediction, State of art algorithms, Machine examine how machine learning and deep learning methods are
Learning used to the identification of brain tumors and will analyse both
their potential advantages and disadvantages. Overall, the goal
of the article is to provide readers a thorough grasp of how to
I. I NTRODUCTION identify brain tumors, how important it is to do so, and what
A tumor is created when cells grow abnormally and com- research is being done in this area right now.
bine to produce an unnatural portion that differs in many ways
from regular cells. The four different types of tumors include
II. M EDICAL I MAGING H ISTORY: B RAIN I MAGING
meningioma, pituitary, no tumor, and glioma [17].Finding
aberrant growths in the brain, either malignant or not, that The practise of medicine known as ”medical imaging”
may or may not be tumors is the process of ”brain tumor entails the use of a variety of methods to see the body’s interior
identification.” A patient’s prognosis and quality of life may organs for diagnostic and therapeutic reasons [2] [8]. The use
be considerably enhanced by early identification and treatment, of imaging tools to research the anatomy and function of the
which highlights the significance of brain tumor detection [13]. brain is one of the most significant uses of medical imaging.
Due to their sometimes vague symptoms and ability to mimic X-rays, which enabled medical professionals to see the bone
other neurological diseases, brain tumors may be difficult to structure of the brain, formed the basis for the early types
identify [7]. However, there has been a rise in interest in of brain imaging. The soft tissues of the brain could not be
leveraging these technologies for more precise and effective studied with X-rays because of a lack of detail. New imaging
brain tumor diagnosis as a result of new imaging methods methods, such as computed tomography (CT) and magnetic
and advancements in machine learning algorithms [1]. Early resonance imaging, were created in the middle of the 20th
identification and planning of therapy are essential for brain century to enable medical professionals to see the soft tissues
tumor patients and may improve their prognosis. of the brain (MRI).
Various imaging methods, such as magnetic resonance A 3D picture of the brain is produced by CT scans using X-
imaging (MRI), computed tomography (CT), positron emis- rays, and they are valuable for finding anomalies like bleeding
 2

or swelling in the brain. MRI scans may aid in the diagnosis B. Anomalies in the Brain
of diseases including brain tumors, stroke, and dementia by Brain structural or functional anomalies are referred to
producing very fine-grained pictures of the brain’s soft tissues as anomalies in the brain. These anomalies, which may be
using magnetic fields and radio waves. brought on by a variety of factors such as genetic mutations,
Functional imaging methods may aid in determining the developmental problems, infections, traumas, and illnesses,
regions of the brain that are engaged during certain activities or may impact different areas of brain function.
under particular circumstances, in addition to these structural Typical instances of brain abnormalities include:
imaging methods. For instance, functional magnetic resonance 1. Anomalies of structure: Anomalies in the size, shape, or
imaging (fMRI) and positron emission tomography (PET) are organisation of the brain’s structures are referred to as struc-
effective for determining the areas of the brain that are engaged tural anomalies. Brain abnormalities, such as microcephaly, in
during cognitive activities and for examining the functional which the brain is smaller than usual, or holoprosencephaly,
structure of the brain. in which the two hemispheres of the brain fail to separate cor-
Overall, brain imaging has made a substantial contribution rectly, may be caused by aberrant brain growth during foetal
to our knowledge of the anatomy and function of the brain and development. Injuries to the brain tissue, such as contusions,
is a useful tool for the diagnosis and treatment of a variety of hematomas, or skull fractures, may potentially cause structural
neurological diseases. malformations.
[Link] in functionality: Neurotransmitter imbalances or
changes in neuronal activity patterns are examples of func-
A. Brain Anatomy tional anomalies, which are deviations from normal brain
function. Mood disorders like sadness and anxiety, which are
The term ”brain anatomy” describes the arrangement and linked to imbalances in neurotransmitter levels, and epilepsy,
structure of the brain. Movement, sensation, perception, and in which aberrant electrical activity in the brain results in
thinking are just a few of the numerous bodily processes that convulsions, are a few examples of functional abnormalities.
the brain regulates [3]. It is made up of many functionally 3. Tumors: The abnormal proliferation of brain cells is known
distinct sections linked by a web of nerve fibres. as a brain tumor [10]. They may develop from glial cells,
The cerebellum, the brainstem, and the cerebrum are the neurons, and meningeal cells among other kinds of brain cells,
three major components of the brain. Two hemispheres make and they can be benign or cancerous. According on their size
up the cerebrum, the biggest portion of the brain (left and and location, tumors may produce a variety of symptoms, but
right). The frontal, parietal, temporal, and occipital lobes make if they are not treated, they can cause death or severe brain
up each of each hemisphere’s four divisions [9]. Conscious damage.
cognition, sensation, and voluntary movement are a few of 4. Traumatic Brain Injury (TBI): A traumatic brain injury
the numerous processes that the cerebrum is in charge of. The (TBI) is a brain damage brought on by an external force, such
area of the brain that joins the spinal cord and cerebrum is as a blow to the head or a piercing wound. Depending on the
known as the brainstem. The pons, midbrain, and medulla extent of the damage, a TBI may cause a variety of symptoms,
oblongata make up its three constituent sections. Numerous including long-term abnormalities in cognition and behaviour
bodily processes that happen automatically, such as breathing, [28].
heart rate, and digestion, are controlled by the brainstem. 5. Virus-borne illnesses: Meningitis, encephalitis, and neu-
Underneath the cerebrum and beneath the brainstem is where rosyphilis are just a few examples of the infectious disorders
the cerebellum is found. It controls how the body moves and that may harm the brain. Inflammation, brain tissue destruc-
maintains its equilibrium. tion, and a variety of neurological symptoms may all be
Neurons and glial cells are only two of the many kinds of brought on by these illnesses.
cells that make up the brain. Specialized cells called neurons In general, brain abnormalities may have a significant influ-
are able to connect with one another and transfer electrical ence on how well the brain functions and can cause a variety of
impulses [4]. Glial cells maintain and guard the neurons in neurological and cognitive problems. These abnormalities of-
addition to contributing to brain activity. ten need a multidisciplinary approach combining neurologists,
neurosurgeons, and other experts, as well as a mix of medical
The cerebrospinal fluid that surrounds and cushions the
imaging, genetic testing, and other diagnostic techniques.
brain and aids in maintaining a stable environment for it, as
well as the skull, which protects it, assist keep it in a healthy
state. The internal carotid and vertebral arteries, which are a III. B RAIN TUMOR LOCALIZATION AND DIAGNOSTIC
part of the cerebral circulation, provide additional blood flow PROCEDURES
to the brain. For the brain to get oxygen, nutrients, and waste Critical first stages in the treatment of brain tumors are the
products as well as to be able to remove them, blood flow is location and diagnosis of the tumor. During these operations,
crucial. the tumor’s nature, location, size, and possible effects on brain
Overall, the brain’s architecture is intricate and diverse, and function are all identified.
it is in charge of many essential bodily processes. For the diag- Brain tumors may be located and diagnosed using a variety
nosis and treatment of neurological diseases, an understanding of techniques, such as:
of the anatomy of the brain is necessary. 1. Magnetic Resonance Imaging (MRI): The brain is seen in
 3

great detail using the noninvasive imaging method known as Society sponsored the ”Multimodal Brain Tumor Segmenta-
magnetic resonance imaging (MRI). When diagnosing brain tion Challenge.” Its full name is this. The BraTS2017 dataset
tumors, it is the most often utilised imaging technique. Brain consists of MRI images of 285 people that have brain tumors
tumors may be found via MRI, which can also provide details as well as manually segmented tumors that were done by
about their composition, location, size, and other characteris- skilled medical professionals. Participants in the competition
tics [11]. were charged with creating and analysing algorithms for
2. Computed Tomography (CT) Scan: The brain is imaged automatically segmenting and categorising tumors using this
during a CT scan using X-rays. They may help detect the information [26]. Many brain tumor segmentation and classi-
existence of tumors and provide details about their size and fication algorithms have been evaluated and compared using
location. In comparison to MRI, CT scans are quicker and less the BraTS2017 challenge in the fields of medical imaging and
costly, but they are less sensitive in identifying certain forms machine learning.
of brain cancers. The BraTS2018 dataset, in a similar vein, comprises manual
3. Positron Emission Tomography (PET) Scan: To find tumor segmentations made by knowledgeable doctors coupled
metabolic activity in the brain, PET scans employ a radioactive with Magnetic Resonance Imaging (MRI) images of 285
tracer. They may be used to determine if a brain tumor is individuals with brain tumors. Participants in the challenge
present and to provide details on its metabolic activities. were charged with creating and analysing algorithms for
4. Biopsy: An operation known as a biopsy involves the surgi- automated tumor segmentation and classification based on this
cal removal of a sample of brain tissue to be microscopically dataset. By adding two additional tasks—the segmentation of
analysed. The kind and grade of a brain tumor are usually tumor invading the brain parenchyma and the prediction of
determined by a biopsy, which is often used to confirm the patient overall survival—the BraTS2018 challenge broadened
diagnosis. the scope of prior challenges.
5. Examining the nervous system: A neurological examination ”Ischemic Stroke Lesion Segmentation” competition is
includes testing the patient’s reflexes, muscular power, coor- known as ISLES-2018. By recognising and defining areas
dination, and sensory perception as well as their neurological of the brain that have been impacted by an ischemic
health. About the tumor’s location and size, this might provide stroke on Magnetic Resonance Imaging (MRI) images,
crucial information. the challenge intended to enhance the state-of-the-art in
6. Testing for Neuropsychology: The patient’s cognitive and ischemic stroke lesion segmentation. 43 MRI images from
behavioural function is evaluated during neuropsychological individuals who had just had an acute ischemic stroke
testing. Insightful data on the tumor’s effect on brain function were made available as part of the ISLES-2018 challenge,
may be obtained from this. coupled with manual segmentations of the lesions done by
Overall, imaging, biopsy, and clinical evaluation are often skilled medical professionals. Participants in the competition
used to localise and diagnose brain tumors. For patients with were charged with creating and analysing automated stroke
brain tumors, these procedures are essential for directing lesion segmentation algorithms based on this dataset. The
therapy choices and enhancing results [27]. ISLES-2018 competition, which was followed by several
challenges the following year, was extensively utilised in the
IV. SOURCES OF DATA AND A SEARCH PROCESS machine learning and medical imaging research fields to test
and evaluate stroke lesion segmentation algorithms.
Brain tumor segmentation remains one of the most dif-
ficult jobs in medical imaging because of their unexpected
appearance and form [19]. About 28 research publications
from the years 2018 to 2023 were analysed for this report. V. MRI BRAIN TUMOR SEGMENTATION
These approaches for finding brain tumors were created and COMPLEXITY PROBLEMS
tested on a number of well-known datasets[TABLE 1]. For the segmentation of brain tumors, Magnetic Resonance
BraTS2016 is a benchmark dataset and challenge for Imaging (MRI) is a popular method. Identifying and separating
the segmentation and classification of brain tumors. The tumor areas from healthy brain tissue on MRI images is a tech-
BraTS2016 dataset consists of MRI images of 274 people nique known as segmentation [24]. Brain tumor segmentation
who have brain tumors as well as manually segmented tumors using an MRI offers numerous benefits, but it also has some
that were done by skilled medical professionals. Participants [Link] brain tumor segmentation presents a number
in the competition were charged with creating and analysing of difficulties, such as:
algorithms for automatically segmenting and categorising tu- 1. The variety of malignancies: Different levels of cellularity,
mors using this information [18]. In order to test and evaluate necrosis, edoema, and contrast enhancement may be present
brain tumor segmentation and classification algorithms, the in heterogeneous brain tumors. Particularly in places with
BraTS2016 competition, which was followed by other chal- minor alterations, this heterogeneity may make it difficult to
lenges in later years, was extensively utilised in the medical distinguish precisely between tumor and normal brain tissue.
imaging and machine learning research fields. 2. Traces of an image: Motion artefacts, inhomogeneities
Brain tumor segmentation and classification benchmark in the magnetic field, and picture noise are just a few of
dataset and competition called BraTS2017. The Medical Image the abnormalities that may interfere with MRI images. By
Computing and Computer Assisted Intervention (MICCAI) producing incorrect tumor area categorization, these artefacts
 4

TABLE I on the ground truth overlap. A complete overlap between the

S OURCE OF DATABASE AND M ETHODS two segmentations is indicated by a score of 1, which is scaled
Source Method(s) Papers from 0 to 1. The proportion of properly identified pixels in the
Databases picture is quantified by the statistic known as ”pixel accuracy.”
BraTS2015 Quantitative Survey [R8], [R9] By dividing the number of pixels in the picture that were
Assisted Quadratic Logit
BraTS2016
BoostClassifier (DBNQLBC)
[R1], [R28] properly identified by all of the pixels in the image, it is
Convolutional XGBoost determined.
(C-XGBOOST), [R5], [R10], Recall indicates the proportion of true positives in the
BraTS2017
K-nearest neighbor, Featuresand [R17]
Extreme Learning Machine (ELM) segmentation that was done using the actual data, whereas
VGG16 , VGG19, precision measures the percentage of true positives in the
[R2], [R11],
3D CNN, Discreate segmentation that was done using the projected data. These
BraTS2018 [R18], [R19],
Wavelet Transform (DWT),
Deep CNN
[R20] metrics may be used to assess how well segmentation models
K-Means Clustering, perform on certain kinds of images. It is usual practice to
BraTS2019 [R3], [R25]
Cascade Convolutional Network measure segmentation performance using the F1 score, which
ISLES-2018 CNN and SVM [R23]
is a weighted average of accuracy and recall. It has a 0 to 1
MRI Dataset from
Deep CNN, R-CNN [R14], [R21] scale, with 1 denoting ideal performance [22].
Figshare
Iranian Imaging
CNN and Softmax [R13], [R27]
Processing speed, memory utilisation, and the capacity to
Center generalise to fresh pictures are other aspects that may be taken
Marker Based
MICCAI 2013
Watershed Algorithm
[R12] into account when assessing segmentation performance in
Five Layers Stopping Criteria addition to these measures. To assess the overall effectiveness
WBA Dataset [R16]
and Batch Normalization of an image segmentation model, it is crucial to take these
The Cancer CNN with U-Net
Imaging Archive Segmentation
[R15] variables into account in addition to the assessment metrics.
U-Net Architecture,
BraTS2020 [R22],
DeepLab V3 Model VII. CONCLUSION
[R24], [R26]
A difficult challenge in cancer diagnosis is the automatic
segmentation of brain tumors. The segmenting of brain tumors
might affect the segmentation algorithms’ accuracy. from MRI using cutting-edge data augmentation techniques
3. Different imaging procedures: The imaging procedures T1- was the focus of this work. We thoroughly examined the WBA
weighted, T2-weighted, and FLAIR may all be used to obtain Dataset, MICCAI 2013, the MRI Dataset from Figshare, all
MRI images. Choosing a procedure may impact how well a of the BraTS Database challenges, and ISLES-2018, and we
tumor is segmented since each technique offers a different examined the segmentation and prediction algorithms used in
level of information about the tumor. these systems. Due to their ease of use and ability to create
4. Absence of reality: For tumor segmentation, there is often anatomically accurate brain tumor instances, affine transforma-
no baseline or industry-accepted gold standard. As a con- tions continue to be the most used in practise, according to our
sequence, comparing the findings of various research and [Link] reviewed cutting-edge deep learning techniques in
assessing the precision of segmentation algorithms is difficult. this article and gave a quick rundown of more conventional
5. Complexity of computation: It takes a lot of computing approaches. Selecting highly representative features for classi-
power and memory to segregate brain tumors using an MRI, fiers or converting past information into probabilistic maps is
which is a computationally challenging operation. Applica- a difficult challenge in conventional automated tumor classes
tions of segmentation algorithms in healthcare contexts could segmentation systems. The development of clinically accept-
be constrained by processing times. able automated tumor segmentation techniques for improved
6. Patterns of tumor development: The location of the tumor, diagnosis may result from future changes and enhancements to
the patient’s age, and other clinical parameters may all have an CNN designs and the integration of supplementary data from
impact on the development pattern of brain tumors [21]. De- other imaging modalities.
veloping segmentation algorithms that can precisely recognise
cancers of various forms and at various phases of development R EFERENCES
may be challenging due to these variances. [1] Kumar, V. V., Prince, P. G. K. (2023). Deep belief network Assisted
In general, the difficulties associated with MRI brain tumor quadratic logit boost classifier for brain tumor detection using MR
segmentation underscore the necessity for the creation of images. Biomedical Signal Processing and Control, 81, 104415.
[2] Ramtekkar, P. K., Pandey, A., Pawar, M. K. (2023). Innovative brain
reliable and precise segmentation algorithms that can take tumor detection using optimized deep learning techniques. International
into account the variety of brain tumors and get beyond the Journal of System Assurance Engineering and Management, 1-15.
constraints of MRI imaging [20]. [3] Baji, F. S., Abdullah, S. B., Abdulsattar, F. S. (2023). K-mean clustering
and local binary pattern techniques for automatic brain tumor detection.
Bulletin of Electrical Engineering and Informatics, 12(3), 1586-1594.
VI. PERFORMANCE ANALYSIS [4] Babayomi, M., Olagbaju, O. A., Kadiri, A. A. (2023). Convolutional
XGBoost (C-XGBOOST) Model for Brain Tumor Detection. arXiv
A popular statistic for assessing segmentation performance preprint arXiv:2301.02317.
[5] Archana, K. V., Komarasamy, G. (2023). A novel deep learning-
is intersection over union (IoU). It calculates the degree to based brain tumor detection using the Bagging ensemble with K-nearest
which the expected segmentation and the segmentation based neighbor. Journal of Intelligent Systems, 32(1).
 5

[6] Panigrahi, A., Subasi, A. (2023). Magnetic resonance imagining-based deep learning. Journal of Magnetic Resonance Imaging, 54(5), 1608-
automated brain tumor detection using deep learning techniques. In 1622.
Applications of Artificial Intelligence in Medical Imaging (pp. 75-107). [27] Kokila, B., Devadharshini, M. S., Anitha, A., Sankar, S. A. (2021, May).
Academic Press. Brain tumor detection and classification using deep learning techniques
[7] Maqsood, S., Damaševičius, R., Maskeliūnas, R. (2022). Multi-modal based on MRI images. In Journal of Physics: Conference Series (Vol.
brain tumor detection using deep neural network and multiclass SVM. 1916, No. 1, p. 012226). IOP Publishing.
Medicina, 58(8), 1090. [28] Khan, A. R., Khan, S., Harouni, M., Abbasi, R., Iqbal, S., Mehmood,
[8] Işın, A., Direkoğlu, C., Şah, M. (2016). Review of MRI-based brain Z. (2021). Brain tumor segmentation using K-means clustering and deep
tumor image segmentation using deep learning methods. Procedia Com- learning with synthetic data augmentation for classification. Microscopy
puter Science, 102, 317-324. Research and Technique, 84(7), 1389-1399.
[9] Dong, H., Yang, G., Liu, F., Mo, Y., Guo, Y. (2017). Automatic brain
tumor detection and segmentation using U-Net based fully convolutional
networks. In Medical Image Understanding and Analysis: 21st Annual
Conference, MIUA 2017, Edinburgh, UK, July 11–13, 2017, Proceed-
ings 21 (pp. 506-517). Springer International Publishing.
[10] Kumar, S., Negi, A., Singh, J. N. (2019). Semantic segmentation using
deep learning for brain tumor MRI via fully convolution neural networks.
In Information and Communication Technology for Intelligent Systems:
Proceedings of ICTIS 2018, Volume 1 (pp. 11-19). Springer Singapore.
[11] Anil, A., Raj, A., Sarma, H. A., Chandran, N., Deepa, R. (2019). Brain
Tumor detection from brain MRI using Deep Learning. International
Journal of Innovative Research in Applied Sciences and Engineering
(IJIRASE), 3(2), 458-465.
[12] Khan, M. A., Lali, I. U., Rehman, A., Ishaq, M., Sharif, M., Saba,
T., ... Akram, T. (2019). Brain tumor detection and classification: A
framework of marker-based watershed algorithm and multilevel priority
features selection. Microscopy research and technique, 82(6), 909-922.
[13] Siar, M., Teshnehlab, M. (2019, October). Brain tumor detection
using deep neural network and machine learning algorithm. In 2019
9th international conference on computer and knowledge engineering
(ICCKE) (pp. 363-368). IEEE.
[14] Deepak, S., Ameer, P. M. (2019). Brain tumor classification using deep
CNN features via transfer learning. Computers in biology and medicine,
111, 103345.
[15] Naser, M. A., Deen, M. J. (2020). Brain tumor segmentation and grading
of lower-grade glioma using deep learning in MRI images. Computers
in biology and medicine, 121, 103758.
[16] Kalaiselvi, T., Padmapriya, T., Sriramakrishnan, P., Priyadharshini,
V. (2020). Development of automatic glioma brain tumor detection
system using deep convolutional neural networks. International Journal
of Imaging Systems and Technology, 30(4), 926-938.
[17] Khan, M. A., Ashraf, I., Alhaisoni, M., Damaševičius, R., Scherer, R.,
Rehman, A., Bukhari, S. A. C. (2020). Multimodal brain tumor clas-
sification using deep learning and robust feature selection: A machine
learning application for radiologists. Diagnostics, 10(8), 565.
[18] Rehman, A., Khan, M. A., Saba, T., Mehmood, Z., Tariq, U., Ayesha,
N. (2021). Microscopic brain tumor detection and classification using
3D CNN and feature selection architecture. Microscopy Research and
Technique, 84(1), 133-149.
[19] Sarhan, A. M. (2020). Brain tumor classification in magnetic resonance
images using deep learning and wavelet transform. Journal of Biomed-
ical Science and Engineering, 13(06), 102.
[20] Sultan, H. H., Salem, N. M., Al-Atabany, W. (2019). Multi-classification
of brain tumor images using deep neural network. IEEE access, 7,
69215-69225.
[21] Gunasekara, S. R., Kaldera, H. N. T. K., Dissanayake, M. B. (2021).
A systematic approach for MRI brain tumor localization and segmen-
tation using deep learning and active contouring. Journal of Healthcare
Engineering, 2021, 1-13.
[22] Yan, H., Chen, A. (2021, June). A Novel Improved Brain Tumor Seg-
mentation Method Using Deep Learning Network. In Journal of Physics:
Conference Series (Vol. 1944, No. 1, p. 012011). IOP Publishing.
[23] Sahaai, M. B., Jothilakshmi, G. R. (2021, May). Hierarchical based
tumor segmentation by detection using deep learning approach. In
Journal of Physics: Conference Series (Vol. 1921, No. 1, p. 012080).
IOP Publishing.
[24] Grøvik, E., Yi, D., Iv, M., Tong, E., Nilsen, L. B., Latysheva, A.,
... Zaharchuk, G. (2021). Handling missing MRI sequences in deep
learning segmentation of brain metastases: a multicenter study. NPJ
digital medicine, 4(1), 33.
[25] Ranjbarzadeh, R., Bagherian Kasgari, A., Jafarzadeh Ghoushchi, S.,
Anari, S., Naseri, M., Bendechache, M. (2021). Brain tumor segmen-
tation based on deep learning and an attention mechanism using MRI
multi-modalities brain images. Scientific Reports, 11(1), 1-17.
[26] Jünger, S. T., Hoyer, U. C. I., Schaufler, D., Laukamp, K. R., Goertz,
L., Thiele, F., ... Pennig, L. (2021). Fully automated MR detection and
segmentation of brain metastases in non-small cell lung cancer using