See discussions, stats, and author profiles for this publication at: [Link]

net/publication/350936156

Brain Tumor Detection using Deep Learning and Image Processing

Conference Paper · March 2021

DOI: 10.1109/ICAIS50930.2021.9395823

CITATIONS READS

88 5,192

1 author:

Aryan Methil
Narsee Monjee Institute of Management Studies
2 PUBLICATIONS 110 CITATIONS

SEE PROFILE

All content following this page was uploaded by Aryan Methil on 30 April 2021.

The user has requested enhancement of the downloaded file.

 Brain Tumor Detection Using Deep Learning and
Image Processing
Aryan Sagar Methil
Narsee Monjee Institute of
Management Studies
Mumbai, India
aryanmethil@[Link]

Abstract — Brain Tumor Detection is one of the most In recent times, Computer-aided diagnosis of diseases is
difficult tasks in medical image processing. The detection task gaining interest and is helping doctors take swift decisions.
is difficult to perform because there is a lot of diversity in the One such approach is using Convolutional Neural Networks
images as brain tumors come in different shapes and textures. (CNN) to learn the spatial and temporal features from the
Brain tumors arise from different types of cells and the cells given dataset which are necessary to identify the disease. A
can suggest things like the nature, severity, and rarity of the Convolutional Neural Network is a special type of neural
tumor. Tumors can occur in different locations and the network which specializes in handling image datasets. The
location of tumors can suggest something about the type of very fundamental principle of this neural network is
cells causing the tumor which can aid further diagnosis. The
performing a convolution operation between the kernel and
task of brain tumor detection can become aggravating by the
problems which are present in almost all digital images eg.
the image to extract the features. All neural networks learn
illumination problems. Tumor and non-tumor images can have by the iterative updation of the weights matrix. Here, it is
overlapping image intensities which makes it difficult for any required to find the optimal kernel values for all layers of
model to make good predictions from raw images. This paper the CNN model. Hence, the kernel values themselves act as
proposes a novel method to detect brain tumors from various the weights of this model and the optimal values of the
brain images by first carrying out different image kernel are gradually learned through backpropagation and
preprocessing methods ie. Histogram equalization and opening gradient descent. Backpropagation is the computation of
which was followed by a convolutional neural network. The derivatives of the loss function with respect to the weights
paper also discusses other image preprocessing techniques and biases in a backward fashion. Gradient descent is the
apart from the ones that are finalized for training and their periodic updation of the weights such that the loss or the
impact on our dataset. The experimental study was carried on error is decreasing with each iteration. A convolutional layer
a dataset with different tumor shapes, sizes, textures, and is often coupled with a pooling layer and we can connect
locations. Convolutional Neural Network (CNN) was employed multiple such convolutional layer-pooling layer pairs. In the
for the task of classification. In our work, CNN achieved a end, we can have few Dense layers and dropout layers for
recall of 98.55% on the training set, 99.73% on the validation the final learning process. Dropout layers are used for
set which is very compelling. tackling the overfitting problem. The last output layer does
the classification job and can have only 1 neuron in case of a
Keywords— Brain Tumor Detection, Computer-aided
binary classification task or more than 1 for a multi-class
Diagnosis, Computer Vision, Convolutional Neural Networks,
Deep Learning, Image Processing, Transfer Learning.
classification task. Special types of neural networks called
Capsule networks which encode the spatial information well
I. INTRODUCTION along with the probability of the object being there in the
image are gaining popularity and they are used in some
A Brain Tumor is a mass of tissue in which the cells
recent works [18][19].
multiply uncontrollably. It arises from different cells - both
in the brain and outside. Primary tumors are the ones that Brain tumors in MRI scans (or any other scans) are
originate in the brain itself whereas secondary tumors are identified by abnormal blobs in the brain. These blobs or
the ones that metastasize to different parts of the body. regions have a different illumination than the rest of the
Tumors can have different origins and based on the cells or brain and are usually brighter than the background.
the origin obtained from different types of tumors. For However, the process of segmenting the tumors in MRI
example, gangliogliomas are tumors that include neoplastic images is a very difficult task. The tumors have different
neurons and are mostly grade I or low-grade tumors which sizes, textures, and even the positions where they are found.
indicates that the tumor is well-differentiated and has slow If we consider segmenting the tumor by properties such as
growth [16]. Another example is meningioma which illumination, we may face issues such as overlapping pixel
originates from the meninges (The set of 3 membranes intensities with the normal tissues. Identification and
covering the brain and spinal cord) and can be grade I, II, or segmentation of brain tumors in MRI images is important as
III and it is slow-growing [16]. Symptoms of a brain tumor it indicates the presence of the abnormal tissues for
include headache which can be acute and persistent, treatment or patient follow-up purposes. Most brain tumors
muscular disorders, dizziness, cognitive disorders, etc. also cause edema [14] which is also a factor that distorts the
Treatment of the same includes chemotherapy, radiotherapy, nearby structures and can change the pixel intensities around
tomotherapy, and surgery (craniotomy). Although brain the tumor. CNN along with some preprocessing techniques
tumor is comparatively infrequent ie. 1.4% new cases per can give accuracy comparable to or even higher than
year [17], in developed countries, fatalities due to brain humans.
tumors have increased over the past few decades. CNS
tumor cases in India range from 5 to 10 per 100,000 In this paper, we have presented a novel brain tumor
population with an increasing trend and it accounts for 2% detection method in MRI images. The first step was image
of malignancies[11][12][13]. preprocessing involving morphological techniques as well
as histogram equalization to enhance our dataset. Then, the
tumor and non-tumor image classification is carried out 105 non-tumor images from another dataset were used[3].
using the Convolutional Neural Network (CNN). This paper The non-tumor images folder was named “no_tumor” in the
strongly emphasizes the use of image processing as digital original dataset on Kaggle. The images were preprocessed
images suffer from various problems such as the and then a 70%-30% split was performed to get the training
illumination problem discussed above and so without proper and validation dataset. The preprocessing which was applied
image preprocessing techniques, even CNNs can get consisted of histogram equalization followed by opening.
misguided to learn incorrect features and produce wrong The resultant dataset was upsampled to get the final dataset
outputs. It is also necessary that we choose the right of 4222 images consisting of 1861 training tumor images,
techniques because a major problem with image processing 563 training non-tumor images, 1463 validation tumor
for datasets is that the chosen technique might be beneficial images, and 315 validation non-tumor images. Upsampling
to only a particular type of image instead of generalizing was done as the dataset should be large enough for the
well to the entire dataset. As such, while experimenting with model.
the techniques, we found some techniques to be harmful to
the dataset even before fitting a model to it while some Test Dataset consists of 20 randomly picked images
techniques looked promising and so we decided to go ahead from the internet out of which 10 had tumors and 10 didn’t.
with the training process.
In this paper, we start by discussing the methodology we
used – ie starting from the dataset followed by discussing
the effects of different image processing techniques on the
dataset and then preparing the proper dataset format for
training, image augmentation, and training and through the
performance metrics to the result and conclusion.
1.
II. METHODOLOGY
A. Block Diagram

Dataset

2.
Image Preprocessing Figure 1 and 2: Sample non-tumor images from dataset.

Train- Validation Split

3.

Image Augmentation

Training

4.
Figure 3 and 4: Sample tumor images from dataset.

Testing

B. Dataset
The input dataset was mostly made up of a subset of a
dataset[1] consisting of 3762 tumor images and the subset
contained 2297 images. The subset selection was done
based on removing the images which might have
misdirected the model training. Another small dataset of 253
images was added[2]. This dataset has 155 tumor images
and 93 non-tumor images. For more non-tumor images, all
C. Image Preprocessing

5. 8.
Figure 7 and 8: Image Processing on non-tumor images

1. Global Thresholding:
This method proved to be quite ineffective. The
reason being, the different levels of illumination,i.e
the different pixel intensities in different images.
Thus, a global threshold cannot be chosen for all
the images, because of their different pixel
intensities. In the tumor images, the tumor is
brighter than the rest of the brain area. If we set a
threshold value, which separates the bright tumor
portion and the darker brain portion, it disrupts
with some non-tumor images in which the entire
brain portion is bright. Setting a threshold will
classify these non-tumor images as tumor images,
hence proving to be ineffective. See figure 7 where
global thresholding makes the brain MRI image
look like there is a tumor but the true label states
6. otherwise.

Figure 5 and 6: Image Processing on tumor images 2. Adaptive Thresholding:

Adaptive thresholding does not use a fixed
threshold value for all pixels in the image. Instead,
the threshold is calculated based on the range of
intensities in the pixel’s local neighborhood.
Therefore, different regions of the image will have
different threshold values. This allows the
thresholding to be done dynamically for different
images. We can observe from the adaptive
thresholding output that the outlines are
highlighted. The issue is that even in the area of the
tumor, the outlines are about equally abundant as
there are in non-tumor regions. This makes it
difficult to distinguish the tumor from the rest of
the brain.

3. Sobel filter
Sobel filter is an edge detection filter and consists
of a pair of convolutional kernels. The first of the
7. pair detects edges running in the vertical direction
while the other detects edges running in the
horizontal direction.
 which adds pixels to the boundaries of objects.
White regions such as tumors grow in size after
dilation due to the addition of white pixels on the
boundaries and the gaps in the white regions get
filled as well (See the tumor region in Figure 6). If
we take a structuring element of size 3x3
containing all 1s and we convolve on our image,
Figure 9: Sobel Filter Kernels any background pixel that is surrounded by at least
1 white pixel, will get changed to foreground pixel.
This technique has a tragic effect on the images. Thus there is an overall decrease in darker regions
The Sobel filter leads to the introduction of and an increase in the whiter regions thus
unwanted noise resulting in ample false detections. highlighting the tumor more. This technique looks
Hence this technique proved to be completely promising and we will proceed with this image
ineffective. processing technique as well.

4. High Pass filter 8. Erosion

A 3x3 high pass filter kernel consisting of 8 in the Erosion works opposite of dilation as it removes
center and -1 everywhere else is applied to the the pixels from the boundaries of the objects.
images. Since it is a derivative filter, it highlights White regions such as tumors shrink in size after
the edges and turns the background black. Here it erosion and the gaps and holes in the white regions
does a similar job and in doing so, we make the increase in size. If we take a structuring element of
tumor indistinguishable from the background size 3x3 containing all 1s and we convolve on our
which is a tragic effect. So we cannot use this image, any foreground pixel that isn’t surrounded
technique for our task. everywhere by white pixels, will get changed to
background pixel. This can lead to the tumors (the
5. Median Blur main focus of our work) being played down.
The median filter is good at eliminating Salt and Although at first thought, erosion seems to be bad
Pepper noise. It does so by replacing the central for this job, erosion combined with dilation
values of the window of the image under (opening or closing) can prove to be effective.
consideration with the median of all the values in Opening (erosion followed by dilation) removes
that window. Since the pixel intensities of the salt the small objects from the foreground which can be
and pepper noise pixels are on the extreme ends ie. used to put the focus on the tumor region. Closing
around 0 for black and around 255 for white in an (dilation followed by erosion) removes the small
8-bit image, while computing the median, they lie gaps and holes in the foreground which can be used
on the extreme ends of the list of pixel intensities to eliminate the darker central regions of the tumor.
(arranged in ascending order for median
computation) of a window. Therefore they rarely Conclusion: Histogram Equalization, erosion, and dilation
become the median and the non-noise pixels were the promising techniques for this task and we tried
become the central value; eliminating the noise different combinations of these for the preprocessing
pixels. Here we observe that the edges and pipeline.
boundaries are preserved. Here very little Training on histogram equalized images gave an accuracy of
difference is observed between the original and 98% on the train set and 80% on the test set which suggests
preprocessed image. Hence it is not very effective. overfitting.
Training on opened (erosion and dilation) images gave an
6. Histogram Equalization accuracy of 96% on the train set and 84% on the test set.
This preprocessing method results in an overall Training on histogram equalized and closed (dilation and
improvement. This is because histogram erosion) images gave an accuracy of 96% on the train set
equalization normalizes the pixel intensities, and 84% on the test set.
thereby normalizing some of the illumination
problems. The technique does a great job in The preprocessing pipeline opted for in the end was
normalizing the pixel intensities of the brain MRI Histogram Equalization followed by opening. This is the
image in figure 7. The image belongs to the non- novelty of our method as the 2 techniques chosen in the end
tumor class but the whiter patch on the top left part handle the illumination problem very well and also enhance
of the brain looks like a tumor although it is not. the morphological features of brain tumors which will help
Histogram equalization reduced the visual the CNN.
difference between that area and the rest of the
brain. This was an encouraging result and we can D. Split the dataset into Train, Validation, and Test sets
proceed to fit a Convolutional Neural Network
(CNN) model on histogram equalized images.
The processed dataset has to be divided into Train and
Validation datasets. The training dataset is the one on which
7. Dilation
CNN is trained while after each epoch or iteration, the
Dilation is an image preprocessing technique
learned model till that iteration can be tested on a validation
set. The validation set becomes a type of test data as the
model was not trained on that; thereby becoming unseen
data.
Through the metrics we use on the validation dataset, we
can track the progress of the model.
A 70% -30% split on the original dataset gave us the train Figure 10-a: Original Image
and validation set.
Images were duplicated for a good dataset.
Test Dataset:
It has the same folder hierarchy as those of the train and
validation datasets. It contains 20 images downloaded from
the Internet out of which 10 contain tumors whereas 10 do
not. Figure 10-b: Shear
E. Image Augmentation
Image Augmentation is the artificial increase in the
image dataset by applying specified transformations such as
rotation, horizontal flipping, etc. The need for image
augmentation arises when the existing dataset is not
sufficient for the neural networks or if there is a scope
improvement in the neural net’s performance if more data is Figure 10-c: Rotation
provided. Thus augmentation increases the size of the
dataset. It also prevents the neural network from
memorizing the data by adding some spatial variations to
the images, thus preventing overfitting. The
ImageDataGenerator of TensorFlow makes a generator for
images that can be fed into the neural network for training
and testing purposes.
The need for augmentation here is because neural Figure 10-d: Horizontal Flip
networks get better when the datasets are large and since our
dataset contains 4222 images, we should increase it. We can
create multiple variants of the image. Consider a cat image, F. Training
if we mirror it from left to right or top to bottom, it still We used a convolutional neural network as our model as
stays a cat. If we rotate the image, then also it stays a cat. CNNs are the neural networks that are best suited for
We can create such variants manually and store them in the images.
dataset but by doing this, we will require more memory and
time. Transfer learning [15] has been applied which means the
training our neural network will do will be based on a pre-
ImageDataGenerator creates such variants internally trained network. We have used a pre-trained model that has
thereby taking extremely less time and not taking a lot of already learned a lot of complex features. The pre-trained
memory. model used is ResNet101v2 which will become our base
Augmentation applied : model on top of which we will fine-tune our task to classify
tumor and non-tumor images.
1. All images are resized to 150x150 pixels.
Reason for using ResNet :
2. Rotation of image between -30 degree and +30
degree. In the case of normal neural networks, an increase in the
3. Horizontal shift by 20% of the image width to the number of layers will decrease the error up to a particular
left or right. point but will then start increasing again. These additional
4. Vertical shift by 20% of the image height to the top layers should not hamper the training so the solution was
or bottom. provided in ResNet.
5. Shear - Distort the image along an axis. ResNet employs a skip connection where the output of
6. Zoom in the image by 20% previous layers will not only affect the next layer but also
7. Apply horizontal flip to the image layers quite ahead of it. Using Regularization can stop the
adverse effects of additional layers.
On top of the ResNet, a flatten layer was used to flatten
out the matrix output of the previous layer into an array. A
dense layer for further training and to avoid any case of
overfitting, a Dropout layer was used. A Dropout layer
drops nodes with a probability p such that any
overdependence of the output on only particular nodes is Figure 11: Metrics
avoided. The value of p used was 0.2. In the above figure:
The activation function used was Rectified Linear Unit TP - True Positive
(ReLU) and a sigmoid activation in the last layer as it is a FP - False Positive
binary classification-yes or no problem. TN - True Negative
FN - False Negative
ReLU activation function outputs y=x if the input x is
positive and 0 otherwise. It has gained popularity as an III. RESULT
activation function for the non-output layers because it is
easier to train and it also tackles the problem of vanishing We trained our CNN model for 20 epochs and we
gradients. Sigmoid activation function outputs 1/ (1+e^-x) recorded the performance metrics after the 20 th epoch.
for an input x. It outputs a probability between 0 and 1 for
the input to belong to the positive class (1). Sigmoid is used Accuracy Precision Recall F1
in the output layer of the model and since it outputs a single Score
value ie. the probability, there is only 1 neuron in the last (%) (%) (%) (%)
layer.
Training 97.94 98.76 98.55 98.65
The model was compiled using Adam as the optimizer Set
with a learning rate of 0.0002 and the loss used was binary
crossentropy and then the entire model was trained for 20 Validation 99.33 99.45 99.73 99.59
epochs. Set

G. Performance Metrics Testing 95 100 90 94.74

Set
Performance metrics measure the performance of a
model based on the predictions made v/s the true labels. The
3 metrics were accuracy, precision, and recall. F1 score is Table 1: Metrics values
another metric that makes use of precision and recall.
Accuracy is the percentage of correctly classified The high values of the performance metrics are indicators of
data points. Accuracy is not a good metric as it fails to a well-trained model for the given dataset and the absence of
suggest anything in the case of imbalanced classes. Consider underfitting which is good. The recall seems to be lower for
10 images out of which 9 are tumor images and 1 is a non- the test set than on the validation set which can be attributed
tumor image. If the model learns badly and predicts every to the fact that the model was trained on a dataset of MRI
image as tumor images, then also the accuracy would be images of a particular distribution while random images
from the Internet might not necessarily belong to that
90% in this case which is good on paper but it fails to tell us
distribution so the test data can be completely foreign to our
that the model was bad. database.
Precision is a metric that says out of all the images
which the metric classified as tumor images, how many of We also have the graphs below depicting the metrics values
those were tumors. Suppose the model identifies an image associated with the training and validation sets after each
to be a tumor image, the person can consult a doctor to epoch. This can show the entire learning process of the
check if there’s a tumor. In this case, there is no health risk model and even the presence of overfitting or underfitting
ie in case of a false positive, the person will only have to (if any).
spend that extra money and time for consulting a doctor.
Recall says that of all tumor images, how many of
those did the model predict that there is a tumor. This is an
extremely important metric and the one we will focus on in
this task. Suppose if a person had a tumor, and the model
classifies it as non-tumor. The person would not consult a
doctor and could die due to the lack of attention given to
that case. Health risk increases if the model predicts a false
negative.
F1 score is a metric that conveys the balance
between precision and recall. It is the harmonic mean of
precision and recall and penalizes the model a lot even when
only one of them is low.
.
Figure 12-a: Accuracy
 The grayscale black and white image is then eroded. In
some places, Berkeley wavelet transform is applied as an
effective segmentation technique. This pipeline was used by
Nilesh Bhaskarrao Bahadure et al. in their paper [4].
A lot of these papers used machine learning techniques that
did not require as many dataset images as deep learning
techniques such as CNN. Hence the experimentation in
these papers was carried out using very few images.
One of the most commonly used techniques is SVM
(Support Vector Machines) which uses what's known as a
kernel trick to find an optimal boundary between the
Figure 12-b: Precision classes. The reason for its popularity is that it is a robust and
accurate algorithm. In Arakeri et al’s paper [5], the
preprocessing involved rotation correction, image denoising
followed by shape and texture feature extraction, and then
the result was fed into an ensemble learning method that
used SVM, ANN, and KNN. It was used to classify benign
and malignant tumors.
Other papers to use SVM are Mariam Saii et al’s paper [9],
and L. Guo et al’s paper [10]. In the former paper, at the
preprocessing level, an anisotropic diffusion filter is applied
to the image for denoising it. In the latter paper, the immune
algorithm was used to search the optimal weights and
parameters.
Figure 12-c: Recall Principal Component Analysis (PCA) is also a popular
algorithm that has its use in representing multivariate data
(The red line graph - Training set tables into sets of smaller variables. This algorithm is
The blue line graph - Validation set) extensively used for feature selection in a lot of research
papers and this brings down the very large feature count to a
From the graphs, we can observe that the training process reasonable one. In Sachdeva J et al’s paper [6], 856 regions
was smooth and the less gap between the training and of interest (SROI) are extracted by the content-based active
validation lines, indicates high generalization to the contour model. 218 intensity and texture features are
validation images and absence of overfitting. extracted from these SROIs. Then PCA was used for
The test images were plotted and we could see which dimensionality reduction and the result was fed into an
images were correctly labeled and which were not. ANN.
Deep learning techniques such as ANN (Artificial Neural
Network) are also used but it is a very basic and generalized
type of neural network and gets combined with other
techniques such as PCA and KNN (K Nearest Neighbors) to
provide good results.
The metric used was Accuracy in most of the cases and the
best result was observed in [Link] et al’s paper who
performed dimensionality reduction using DCT and then
used a Probabilistic neural network (PNN) for classification.
The accuracy was 100% [7].
Figure 13: Wrongly Predicted Test Image Apart from MRI images, tumor classification has also been
performed on CT scan images as seen in Padma
The 1.0 image title in red indicates that the image should Nanthagopal et al’s paper [8]. A wavelet-based statistical
have been classified as a tumor image (1.0) but it got texture feature set was derived from 2 level DWT. Genetic
classified as non-tumor (0.0). The reason for the algorithm and principal component analysis were used for
misclassification of the above image may be attributed to optimal texture feature selection. Then SVM was used for
the fact that the test images may not necessarily belong to classification purposes.
the same distribution as that of the training or validation set.
Most of the research works done on this topic were on very
Some already existing methods: small datasets whereas our work was done on a
While looking at the research papers, we realized that the comparatively larger dataset giving it a better ability to
standard pipeline for image processing looked something generalize. Our novel method emphasizes the power of
like greyscaling followed by skull stripping ie. all non-brain image processing – easy, simple to implement but very
tissues are removed using contouring or segmentation. This useful at the same time.
is followed by thresholding using histogram analysis to find
out the appropriate valley region for the threshold value.
IV. CONCLUSION
 Brain tumors especially the malignant ones are REFERENCES
considered almost incurable and fatal. The need for early [1] Jakesh Bohaju, (2020, July). Brain Tumor, Version 3. Retrieved
detection arises from the fact that brain tumors can have September 22, 2020, from
symptoms that do not seem to be alarming at first. The most [Link]
common symptom of brain ailments is a headache which [2] Navoneel Chakrabarty, (2019, April). Brain MRI Images for Brain
worsens over time in the case of brain tumors. Hence there Tumor Detection, Version 1. Retrieved September 18, 2020, from
are lots of cases where the fatality from brain tumor [Link]
increased due to the diagnosis not being done early. detection.
[3] Sartaj, (2020, May). Brain Tumor Classification (MRI), Version 2.
Brain tumor diagnosis begins with an MRI scan which is Retrieved September 23, 2020 from
followed by studying a tissue sample for determining the [Link]
type of tumor. MRI scan can also reveal additional details [4] Nilesh Bhaskarrao Bahadure, Arun Kumar Ray, Har Pal Thethi,
such as the size of the brain tumor. "Image Analysis for MRI Based Brain Tumor Detection and Feature
Extraction Using Biologically Inspired BWT and SVM", International
This paper presents a novel method involving image Journal of Biomedical Imaging, vol. 2017, Article ID 9749108, 12
processing techniques for image manipulation which would pages, 2017. [Link]
aid our CNN model to classify tumor and non-tumor images [5] Arakeri, M.P., Ram Mohana Reddy, G. An intelligent content-based
better. Image Processing techniques helped us solve the image retrieval system for clinical decision support in brain tumor
illumination issues and brought the tumor into focus. Data diagnosis. Int J Multimed Info Retr 2, 175–188 (2013).
augmentation was used to reduce the chances of overfitting, [Link]
as it artificially expands the size of a training dataset, thus [6] Sachdeva J, Kumar V, Gupta I, Khandelwal N, Ahuja CK.
Segmentation, feature extraction, and multiclass brain tumor
bringing out an improvement in the performance and the
classification. Journal of Digital Imaging. 2013 Dec;26(6):1141-1150.
ability of the model to generalize. Transfer learning is also DOI: 10.1007/s10278-013-9600-0.
used as a pre-trained model, ResNet101v2 was used as the [7] D. Sridhar and I. Murali Krishna, "Brain Tumor Classification using
base model, upon which further training was applied to tune Discrete Cosine Transform and Probabilistic Neural Network," 2013
our task. The system recorded an adequate accuracy of International Conference on Signal Processing, Image Processing &
97.94% with an excellent training recall of 98.55 % and Pattern Recognition, Coimbatore, 2013, pp. 92-96, doi:
validation recall of 99.73%. 10.1109/ICSIPR.2013.6497966.
[8] Padma Nanthagopal, A., Sukanesh Rajamony, R. Automatic
There are limitations to our work as there are small chances classification of brain computed tomography images using wavelet-
that the image preprocessing applied can damage the based statistical texture features. J Vis 15, 363–372 (2012).
information which makes a tumor image appear non-tumor [Link]
in the eye of the CNN model. The input image should be of [9] Mariam Saii, Zaid Kraitem, Automatic Brain Tumor Detection in
a good enough size because if it’s not, after resizing the MRI Using Image Processing Techniques, Biomedical Statistics and
Informatics. Vol. 2, No. 2, 2017, pp. 73-76. doi:
image to the size we have set in the image augmentation 10.11648/[Link].20170202.16.
step ie. 150x150, the image can become unsuitable for use
[10] L. Guo, L. Zhao, Y. Wu, Y. Li, G. Xu and Q. Yan, "Tumor Detection
(Figure 14-a and 14-b). For future improvements, we can in MR Images Using One-Class Immune Feature Weighted SVMs,"
use ensemble techniques and combine the performance of in IEEE Transactions on Magnetics, vol. 47, no. 10, pp. 3849-3852,
different models for better performance. Oct. 2011, doi: 10.1109/TMAG.2011.2158520.
[11] Nair M, Varghese C, Swaminathan R. Cancer: Current Scenario,
Intervention Strategies and Projections for 2015. NCMH Background
Papers; 2015. [Google Scholar] [Ref list]
[12] Yeole BB. Trends in the brain cancer incidence in India. Asian Pac J
Cancer Prev. 2008;9:267–70.[PubMed][Google Scholar]
[13] Dasgupta, Archya et al. “Indian data on central nervous tumors: A
Figure 14-a: Before Resizing summary of published work.” South Asian journal of cancer vol. 5,3
(2016): 147-53. doi:10.4103/2278-330X.187589.
[14] Papadopoulos MC, Saadoun S, Binder DK, Manley GT, Krishna S,
Verkman AS. Molecular mechanisms of brain tumor edema.
Neuroscience. 2004;129(4):1011-20. doi:
10.1016/[Link].2004.05.044. PMID: 15561416.
[15] Niklas Donges (2019, June 16). What is transfer learning? Exploring
the popular deep learning approach. Builtin. [Link]
science/transfer-learning.
[16] American Association of Neurological Surgeons. (n.d.).
“Classification of Brain Tumors”.
[Link]
Figure 14-b: After Resizing
[17] Charles Patrick Davis. (2020, August 24). “How do you get Brain
In conclusion, image processing proved to be effective Cancer?” [Link]
in solving the illumination problems of the different images [18] Vijayakumar, T. "Classification of Brain Cancer Type Using Machine
and reducing the noisy details thereby bringing the tumor in Learning." Journal of Artificial Intelligence 1, no. 02 (2019): 105-
113.
focus. Different variants of the images were created using
[19] Pandian, A. Pasumpon. "Identification and classification of cancer
image augmentation techniques which augmented the
cells using capsule network with pathological images." Journal of
images and internally created more images for the model. Artificial Intelligence 1, no. 01 (2019): 37-44.
CNN combined with transfer learning proved to be an
effective training model which can be seen in the extremely
good values of the three performance metrics.

View publication stats