Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
69 views

Breast Cancer Detection Using Machine Learning

The document summarizes research on using machine learning to detect breast cancer from mammography images. It discusses how machine learning algorithms can be applied to features extracted from medical imaging data to accurately diagnose breast cancer. The document presents results from using an XGBoost model on the MIAS mammography dataset, achieving 95% accuracy in classifying images as normal or abnormal. It concludes machine learning shows potential as a valuable tool for early breast cancer detection but suggests future work with larger datasets and deep learning approaches.

Uploaded by

SMART ENGRZ
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
69 views

Breast Cancer Detection Using Machine Learning

The document summarizes research on using machine learning to detect breast cancer from mammography images. It discusses how machine learning algorithms can be applied to features extracted from medical imaging data to accurately diagnose breast cancer. The document presents results from using an XGBoost model on the MIAS mammography dataset, achieving 95% accuracy in classifying images as normal or abnormal. It concludes machine learning shows potential as a valuable tool for early breast cancer detection but suggests future work with larger datasets and deep learning approaches.

Uploaded by

SMART ENGRZ
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 14

Breast Cancer Detection

Using Machine Learning


Mustafa Mundher Mohammed
Introduction

 Cancer is a complex and pervasive group of diseases characterized by the


uncontrolled growth and spread of abnormal cells within the body.
 It can occur in virtually any tissue or organ, leading to a wide range of
manifestations and symptoms.
 Globally, cancer has been a significant public health concern, with millions of
new cases reported each year. In 2020, the International Agency for Research
on Cancer (IARC), a part of the World Health Organization (WHO), estimated
that there were approximately 19.3 million new cancer cases worldwide.
 Additionally, cancer-related deaths were estimated at around 10 million in
the same year.
 According to the World Health Organization (WHO), cancer is one of the
leading causes of morbidity and mortality worldwide.
Estimated number of incidents cases and
death all cances
Types of cancers

 Lung Cancer
 Breast Cancer
 Colorectal Cancer
 Prostate Cancer
 Skin Cancer
 Ovarian Cancer
 Pancreatic Cancer
 Leukemia
Causes of cancer

Cancer is a complex group of diseases, and its causes are multifactorial, often involving a
combination of genetic, environmental, and lifestyle factors. While not all cancers share
the same set of causes, common contributors to the development of cancer include:
 1. Genetic Factors
 2. Environmental Factors
 3. Lifestyle Choices
 4. Age
 5. Infections
 6. Hormones
 7. Sun Exposure
 8. Occupational Exposures
 9. Chronic Inflammation
Mammography
 Mammography, called Mastography, is the process of using a low dose of X-ray
to test breast screening.
 Mammography is a pivotal key for the early detection in fighting breast
cancer.
 It’s an accurate way to detect tiny bits of calcium, called microcalcifications.
There are two main types of mammography called digital tomosynthesis
having the same executing mechanism.

2D and 3D mammography
Dimensionality reduction techniques
classified
Dataset
 Mammographic Image Analysis Society (MIAS), an organization including UK
research organizations, offers a digital mammograms database focusing on
comprehending mammograms.
 The database is highly accessible and widely utilized due to its ease of
access. It comprises 322 digitized mammography images, including both right
and left breast images of 161 patients (a pair for each patient).
 The images are in Portable Gray Map (PGM) format and have a resolution of 50
microns. The MIAS dataset consists of two categories: normal and abnormal
classes.
 There are a total of 208 photos classified as normal, while 114 images are
classified as abnormal. Among the abnormal images, 63 are classified as
benign and 51 are classified as malignant.
 Each picture is a grayscale image with a dimension of 1024×1024 pixels [33-
35,48]. The database was obtained from the Mammographic Image Analysis
Society (MIAS) database v1.21, which is hosted on the cam.ac.uk website.
MIAS dataset description
Column Description

1st MIAS database reference number

Character of background tissue:

F – Fatty
2nd
G – Fatty-glandular

D – Dense-glandular

Class of abnormality present:

(1) CALC – Calcification

(2) CIRC–Well-defined/circumscribed masses

(3) SPIC – Spiculated masses


3rd
(4) MISC – Other, ill-defined masses

(5) ARCH – Architectural distortion

(6) ASYM – Asymmetry

(7) NORM –Normal

Severity of abnormality:

4th B – Benign

M – Malignant

5th, 6th x, y image-coordinates of center of abnormality.

7th Approximate radius (in pixels) of a circle enclosing the abnormality.


Proposed model
Results - XGBoost
 The XGBoost technique is used to classify breast cancer into two different
classes (normal and abnormal).
 The class labeled abnormal dedicated for cancerous cases has a bit low
precision rate than the class labeled normal dedicated for no cancer case
with rateof 0.01. however, The normal class has a higher recall rate than the
abnormal class with a rate of 0.04.
 The macro average is equal to 0.94 for precision, recall, and F1-score, while
the weighted average for precision, recall, and F1-score is equal to 0.95.

Classification Report:
precision recall f1-score support

normal 0.95 0.96 0.96 79


abnormal 0.94 0.92 0.93 49

accuracy 0.95 128


macro avg 0.94 0.94 0.94 128
weighted avg 0.95 0.95 0.95 128
Results – Confusion Matrix
The value 76 means that the model detects 76 abnormal female patients having breast cancer.
The TN of value 45 means that the model concludes that 45 normal female patients do not
have breast cancer.
The FP of value 3 indicates that the model considers 3 normal female patients having breast
cancer.
The FN of value 4 means that the model is unable to detect breast cancer in 4 female
patients.
The confusion matrix illustrates the cases that are well classified into normal and abnormal.
Conclusion
 This thesis has addressed a critical aspect of healthcare by exploring the application
of Machine Learning for breast cancer detection.
 The study has demonstrated the potential of Machine Learning algorithms,
particularly those based on features extracted from medical imaging data, to
contribute to the accurate and efficient diagnosis.
 This thesis has delved into the realm of breast cancer detection using machine
learning, employing various approaches to enhance the accuracy of diagnosis.
 The success of these approaches, as evidenced by their high accuracy, underscores
their potential as valuable tools in the early detection of breast cancer.
 The integration of advanced machine learning techniques, such as feature extraction
and classification algorithms, has proven instrumental in achieving robust and
reliable results.
 The attained accuracy rate signifies a substantial step forward in leveraging machine
learning for medical applications, paving the way for enhanced patient outcomes and
a more effective approach to combating this prevalent and life-threatening disease.
Future Works
 By tackling the obstacles described in this study, the forthcoming endeavors in
breast cancer identification by Machine Learning exhibit immense potential
for introducing a new era in breast cancer treatment and significantly
influencing patient results.
 An important area for additional investigation includes improving and
optimizing Machine Learning models on larger and more varied datasets.
 Increasing the magnitude and variety of datasets will enable a more rigorous
assessment of model generalization across diverse populations and
differences in breast cancer subtypes.
 Deep Learning architectures have demonstrated considerable promise in
diverse medical imaging applications, and their utilization in the diagnosis of
breast cancer needs more exploration.
 Designing and adapting neural network structures specifically for
mammographic pictures has the potential to improve the accuracy and
precision of breast cancer diagnosis models.

You might also like