2. INTRODUCTION
• Overview
• Breast cancer is one of the main causes of cancer death worldwide. Early diagnostics
significantly increases the chances of correct treatment and survival, but this process is
tedious and often leads to a disagreement between pathologists. Computer-aided
diagnosis systems showed potential for improving diagnostic accuracy. But early
detection and prevention can significantly reduce the chances of death. It is important
to detect breast cancer as early as possible
• Purpose
• The goal of this project is to create a web-based tool that can batch analyze
histopathology image patches and predict if breast cancer is present. We will create a
CNN model, this model will be loaded into a web app. A pathologist will be able to get
an instant prediction indicating whether or not breast cancer is present in those
images.
3. LITERATURE SURVEY
• Existing Problem
• The process that's used to detect breast cancer is time consuming and
small malignant areas can be missed. In order to detect cancer, a tissue
section is put on a glass slide. A pathologist then examines this slide
under a microscope.The pathologist needs to visually scan large regions
where there's no cancer in order to ultimately find malignant areas.
• Proposed Solution
• By deploying a machine learning solution as a web app, it can be made
available to medical personnel anywhere in the world for free.
4. THEORETICAL ANALYSIS
• Breast Cancer occurs as a results of abnormal growth of cells in the breast tissue,
commonly referred to as aTumor. A tumor does not mean cancer - tumors can be
benign (not cancerous), pre-malignant (pre-cancerous), or malignant
(cancerous).Tests such as MRI, mammogram, ultrasound and biopsy are
commonly used to diagnose breast cancer performed. Since this build a model
that can classify a breast cancer tumor using two training classification:
• 1= Malignant (Cancerous) - Present
• 0= Benign (Not Cancerous) -Absent
• Since the labels in the data are discrete, the predication falls into two categories,
(i.e. Malignant or Benign). In machine learning this is a classification problem.
Thus, if we can classify whether the breast cancer is benign or malignant..
6. Software Designing
• The training and testing folders have 2010 images in total, equal images in
each category.
• CNN Architecture
• Convolution
• Pooling
• Fully Connected
• Image Classification
7.
8. Steps to execute flask files
• 1- from spyder, edit and save app.py and index.html in a location.
• 2- open Anaconda Prompt, go to that location and execute app.py
• 3- open default browser - type: localhost:5000 - to see if the html page is
working or not.
• 4- in browser on same page- we have to select an image and predict
outcomes.
9. RESULT
We have successfully trained and tested more than 2000 clinical
histopathology image samples, and achieved more than 75% accuracy. By
saving its predictive capabilities as a model we have created a interactive web
application for advanced clinical purposes.
10. ADVANTAGES AND DISADVANTAGES
• While the purpose of such a tool would not be to replace physician advice or
mammograms, risk predictions derived from our models could contribute to
both early breast cancer detection and breast cancer prevention. Women who
receive high estimated risks could be motivated to seek out a doctor or take
other preventative actions. Models could be used to guide immediate personal
decisions such as screening. Doctors could also use these models to inform
decisions on whether and when to recommend long-term preventative actions
such as chemoprevention and hormone replacement therapy.
• One limitation was that models were trained and tested on separate portions
of the same data set. Ideally, models would be trained on one data set and
validated on another data set from a separate study. Such external validation
can prove the generalizability of models
11. APPLICATION
• By deploying a machine learning solution as a web app, it can be made available to medical
personnel anywhere in the world for free.
• A pathologist will be able to get an instant prediction indicating whether or not breast cancer is
present in those images.
• These models could be used to inform and guide screening and preventative measures.
• Our models could easily be incorporated into phone application or website breast cancer risk
prediction tools.
• Using our models as such would be convenient and cost-effective as our personal health data
inputs are easy and inexpensive to obtain from electronic health records or an office visit.
12. CONCLUSION
The analysis of AI methods have great potential for implementation in clinical
analysis and improve the diagnostic capability of existing CAD systems. From
the literature, it can be found that heterogeneous breast densities make
masses more challenging to detect and classify compared with calcifications.
The traditional ML methods present confined approaches limited to either
particular density type or datasets.
13. FUTURE SCOPE
Although the DL methods show promising improvements in breast cancer
diagnosis, there are still issues of data scarcity and computational cost, which
have been overcome to a significant extent by applying data augmentation
and improved computational power of DL algorithms. Future studies could
also determine if logistic regression, linear discriminant analysis, and neural
network models with our selected inputs predict breast cancer risk even better
when trained on a larger data set.
14. REFERENCES
• - Breast Histopathology Images 198,738 IDC(-) image patches; 78,786 IDC(+) image
patches https://www.kaggle.com/paultimothymooney/breast-histopathology-images
• - Convolutional Neural Network for Breast Cancer Classification https://towardsdatascience.com/convolutional-neural-
network-for-breast-cancer-classification-52f1213dcc9
• - Pfeiffer RM, ParkY, Kreimer AR, Lacey JV Jr, Pee D, Greenlee RT, et al. Risk prediction for breast, endometrial, and
ovarian cancer in white women aged 50 y or older: derivation and validation from population-based cohort studies. PLoS
Med. 2013 Jul 30;10(7):e1001492. pmid:23935463
• - Evans DG, Howell A. Breast cancer risk-assessment models. Breast Cancer Res. 2007 Sep 12;9(5):213. pmid:17888188
• - U. S. Preventive ServicesTask Force [Internet]. Final Update Summary: Breast Cancer: Screening; 2019 May [cited 2019
Sep 20]. Available from: https://www.uspreventiveservicestaskforce.org/Page/Document/UpdateSummaryFinal/breast-
cancer-screening