BONE CANCER DETECTION USING

MACHINE LEARNING
SYNOPSIS:
 ABSTRACT
 BONE
 BONE CANCER
 MACHINE LEARNING
 BONE CANCER DETECTION USING MACHINE
LEARNING
 DATASETS USED
 ALGORITHM USED
 METHODOLOGY
 CONCLUSION
 ABSTRACT

In the field of bio-medical, tumor detection in early stages is the trending research
topic as most of the tumors indicates the early stages of cancer. A lot of work have
been done on the tumor detection and identification. So, the point of discussion is to
find out the identification and detection system which is fast and reliable. In this
paper an approach of tumor detection using machine learning have been discussed
and the data set for the performance analysis is MRI images. This paper specifically
dedicated for the bone tumor detection. Also the algorithm have been implemented
in OpenCV so as to make the system more fast and convenient.
 BONE
• The bone is the body's hollow support skeleton. The outer portion of bones is a hard tissue structure
called calcium salt matrix. The hard out layer is made from cortical bones, it covers the trabecular bone
inside and outside the periosteal-covered bone. Some bones are hollow and the soft tissue known as bone
marrow contains the medullary cavity. Endosperm is a lining of the tissue. At each bone end is a area of
bone like cartilage that's thinner than a bone made from the fibrous matrix of a gel-like substance that
doesn't contain much calcium.
• The bulk of the bones are like cartilage. Instead the body puts the calcium on the bone. Some cartilage
can remain at the ends of the bone formation to serve as reinforcement between bones. The cartilage
connects bones to form a joint, along with ligaments and some other tissue. The bone is very rigid and
muscular itself. Bone will carry up to 12,000 livers per centimeter per inch. The strain to crack the thigh
bone takes as much as 1200 to 1800 pounds. There are two cells in the bone. The Osteoclast is the new
bone cell, and the Osteoclast is the old bone cell. Some bones of the marrow are soft tissues.
 BONE CANCER

• There are various types of cancer detected in the human body. If the tumor affects bone
directly then that type of disease is known as bone cancer. The sarcomas are called bone
cancers. Sarcomas arise in muscle , bone, fibrous tissue, blood vessels, fat tissue and several
other tissues. They can grow in the body anyplace. Bone refashion operation is attributed
exclusively to Bone Cancer cells. Standard bone is being tirelessly changed, or conked out and
rebuilt. Cancer cells fault equilibrium for cell growth and bone development. If cancer cells are
in the bones, then as opposed to normal bone density, the bone structure is bent at a higher
rate. The cancer of the bone may either be primary or secondary. Bone cancer occurs
primarily in the bone. Whereas secondary bone cancer develops in every portion of the body.
 MACHINE LEARNING

• Machine Learning is the field of study that gives computers the capability to learn without being
explicitly programmed. ML is one of the most exciting technologies that one would have ever come
across. As it is evident from the name, it gives the computer that makes it more similar to
humans: The ability to learn. Machine learning is actively being used today, perhaps in many more
places than one would expect.
• Machine learning (ML) is the study of computer algorithms that can improve automatically through
experience and by the use of data. It is seen as a part of artificial intelligence. Machine learning
algorithms build a model based on sample data, known as "training data", in order to make
predictions or decisions without being explicitly programmed to do so.Machine learning algorithms
are used in a wide variety of applications, such as in medicine, email filtering, speech recognition,
and computer vision, where it is difficult or unfeasible to develop conventional algorithms to perform
the needed tasks.
 BONE CANCER DETECTION USING MACHINE
LEARNING

• In this day of modern science and age where scientific and technological accomplishments are touching new heights with
every second that is passing, the main step in cancer detection is how to classify tumours into malignant or benign which
is a challenging task. Machine learning techniques can enormously improve the accuracy of diagnosis. We aim to classify
tumour into malignant or benign tumour using different features from several cell images. Machine learning uses the
computer data to learn and then use this data to learn a particular pattern or trend in the data. The increasing cancer
rate all over the world in today’s date is alarming and there is an increased need for efficient cancer detecting techniques.
This is possible using Machine learning. This technique provides early detection of tumour which eventually helps in early
diagnosis which plays an important role in the treatment of tumour patients. According to global statistics breast cancer
is a significant public health problem in today’s society because of its widespread increase in cancer rates. Because of its
unique advantages in critical features detection from complex data sets, machine learning (ML) is widely recognized as
the methodology of choice in cancer pattern classification. This project aims to observe which features are most helpful in
predicting malignant or benign cancer and to see general trends that may aid in accurate cancer detection.
DATASETS USING IN ML FOR BONE CANCER

• This part of the project includes the research about the available dataset that can be used for the
project. The initial dataset that we planned on working had certain drawbacks. CBIS-DDSM dataset:
Neither the analytics nor the experimental validation of the data were sufficient to move forward with
the project. There was no exact data regarding the classification of tumours into benign and malignant
incase of scan images of patients. One the most important demerit of this dataset was the improper
organization of data and huge redundant records which can cause glitches in the training process.
• Hence it was concluded that evaluation based on detection rate and accuracy levels with this data-set is
not appropriate. Breast cancer Wisconsin data-set: This data-set was a text type data-set which had a
classified set of data into malignant and benign tumours based on the cell features threshold values for
various features like cell radius, perimeter, area, symmetry and other such characters. We didn’t go for
this data-set since we planned on a better version of cancer detection model which focused on images
DATASET to detect bone cancer using Detection estimation using machine learning
the technology called MRI
 ALGORITHMS USING IN MACHINE
LEARNING FOR BONE CANCER DETECTION

•  Almost 70% of all reported studies use neural networks as their primary (and sometimes only) predictor. Support
vector machines are a distant second with 9%, while clustering and decision trees each account for about 6%.
Genetic algorithms and other methods (naïve Bayes, fuzzy logic) are rarely used. This is both surprising and a bit
disappointing. ANNs are relatively old machine learning technologies which yield so-called “black-box” results.
That is, their performance and classification processes are not easily explained or rationalized. The existence of
other methods (SVMs, DTs, NBs) which inherently provide easily accessible explanations appears not to be widely
known among cancer informaticians. Overall, many of the papers reviewed for this survey were of generally high
quality. Some of the better papers are discussed in more detail under the “Case Studies” section of this review.
However, a disturbing number of studies lacked sufficient internal or external validation, were trained on far too
few examples, tested on only a single machine learner or had no well-defined standard with which to compare the
performance of the reported algorithm.
A histogram showing the frequency with which different
types of machine learning methods are used to predict
different types of cancer. Breast and prostate cancer
dominate, however a good range of cancers from different
organs or tissues also appear to be compatible with machine
learning prognoses. The “other” cancers include brain,
cervical, esophageal, leukemia, head, neck, ocular,
osteosarcoma, pleural mesothelioma, thoracic, thyroid, and
trophoblastic (uterine) malignancies.
METHODOLOGY
I. Acquisition Of Image
The first stage in any vision system is the acquisition of images. The images are considered the best
because of their high resolution, such as CT scans, the MRI and X-ray. It is commonly used in
scientific applications due to its capacity to create images of the human body that are not
invasively high.

II. Pre-processing
Preprocessing an image increases the precision of an optical inspection significantly.
Preprocessing is the first step towards improving image quality. The filtering technique is used to
start the image processing stage. Image filtering is useful to smoothen, sharpen and remove noise
for many purposes. Filtering eliminates noise or other small image fluctuations. Such noises then
have to be denoted. Gabor filter is used for noise removal and for the deficiency of the pictures in
this methodology. The key advantage of this filter is, as compared with other filters, it achieves
excellent noise reduction with less blurring. The gray conversion will be the next step after
filtering. This is the process by which pixels with RGB level are converted to gray. The picture of
color has more meaning in the process. Therefore, it is appropriate to convert the gray picture.
The main aim of this conversion is to remove the information on the color and saturation by
retaining the light.
III. Super-Pixel Segmentation
Superpixels have a simple primitive from which local image characteristics
can be determined. They record image redundancy and reduce the
complexity of future image processing tasks significantly. The applications,
such as profile estimation, image division, skeletisation, body model
measurement and object localization, have shown their effectiveness. To be
useful, superpixelsis fast, easy to use and produce high-quality segments. It
begins with sampling K cluster centers that are often scattered and moved
to seed sites which value the lower gradient position in a3 [3] very nearby.
very 3 [3] It is also done so that they are not added and the odds are lowered
To select a pixel. Gradients of the image are determined asThe comparison
SUPERPIXEL
vector is the regular L2, where it is equivalent to the pixel ( x , y). Every color
SEGMENTATION
and every colorData on intensity are considered.

I. Feature Extraction
The extraction function of the captured images can be carried out using the number of techniques available. In this
article, we will use a machine learning algorithm to make the device more robust. There are many algorithms in the
machine learning algorithm which are graded on the basis of their results. Specifically, supervised learning of the
Random Forest and the nearest neighbor algorithm is useful, as these algorithms produce a function that maps inputs to
the desired outputs.
I. CLASSIFICATION
Vector support machines are supervised learning models that evaluate data and patterns that can be used
for classification. The basic SVM uses a number of input data and predicts which of the two groups is the
input to construct a non-probabilistic binary linear classification. From a set of training examples, each
classified as belonging to one of the two categories, the SVM training algorithm constructs a model that
assigns new examples to one category or another. In the proposed process, we use a linear classifier

RESULT
We have chosen T1W, T2W, STIR, PD, improved tumor contrast images in light range. Our techniques can be
applied to higher intense tumor images, instead of other tissue or regions of the bone. The reason behind the
selection of these images is because The experimental results of a set of MR images containing two kinds of
tumors, Enchondroma, Giant cells
We have analyzed 2000 MR images for our work and collected 507 MR data samples. They are processed
separately by manual analysis in tumor and non-tumor archives. The remainder have been discarded mostly
because of poor river segmentation. Data 318 in 507 samples MR images are those without tumors and 189 MR
The tumor images are the images.
SUMMARY OF THE PROPOSED METHODOLOGY OF
BONE CANCER DETECTION
RESULT FROM EACH EXPERIMENT FROM EACH DATASETS
 CONCLUSION

• Bone cancer is one form of harmful illness, so cancer must be identified in the early stages. But the most
difficult task is detecting bone cancer. A lot of techniques are used to detect bone cancer from the literature
review but they have some limitations. Our proposed method in the multiple stage as preprocessing, edge
detection, morphological function, segmentation, then extraction of features and finally classification. The
suggested system successfully identifies cancer of the bone from CT scanning images. By the conclusion of
the system , the system is fulfilling the optimal goal.
THANK YOU