Breast Cancer Prediction Project

Project Overview

The goal of this project is to develop a Python-based program that processes, visualizes, and classifies data from a dataset containing indicators of breast cancer patients. The project involves exploring various Machine Learning models to predict (diagnose) whether a patient is likely to have breast cancer.

Objectives

• Data Extraction and Visualization: Extract and visualize key characteristics from a dataset of breast cancer indicators.
• Model Development: Develop and compare multiple classification models (SVM, Random Forest, ANN/DNN) to predict whether a patient is likely to have breast cancer.
• Performance Evaluation: Utilize appropriate metrics (e.g., Accuracy, Precision, Recall) to evaluate the performance of the models and determine the best-performing classifier.

Dataset

The dataset used in this project is named bcdr_f01_features.csv, which contains 44 variables: 16 integer fields, 27 real (float) fields, and 1 string field. The dataset provides indicators collected from breast cancer patients. Here's a brief description of each variable:

Column Name	Description
patient_id	Identifier for each patient.
study_id	Identifier for each study associated with a patient.
series	Series number within the study.
lesion_id	Identifier for each lesion within a study.
segmentation_id	Identifier for each segmentation of a lesion.
image_view	The view in which the image was taken (e.g., craniocaudal, mediolateral).
mammography_type	Type of mammography used (e.g., screening or diagnostic).
mammography_nodule	Indicates the presence of a nodule (binary).
mammography_calcification	Indicates the presence of calcification (binary).
mammography_microcalcification	Indicates the presence of microcalcifications (binary).
mammography_axillary_adenopathy	Indicates the presence of axillary adenopathy (binary).
mammography_architectural_distortion	Indicates the presence of architectural distortion (binary).
mammography_stroma_distortion	Indicates the presence of stromal distortion (binary).
age	Age of the patient.
density	Breast density category.
i_mean	Mean intensity value of the image.
i_std_dev	Standard deviation of the image intensity.
i_maximum	Maximum intensity value in the image.
i_minimum	Minimum intensity value in the image.
i_kurtosis	Kurtosis of the image intensity distribution.
i_skewness	Skewness of the image intensity distribution.
s_area	Area of the segmented region.
s_perimeter	Perimeter of the segmented region.
s_x_center_mass	X-coordinate of the center of mass of the segmented region.
s_y_center_mass	Y-coordinate of the center of mass of the segmented region.
s_circularity	Circularity of the segmented region.
s_elongation	Elongation of the segmented region.
s_form	Form factor of the segmented region.
s_solidity	Solidity of the segmented region.
s_extent	Extent (ratio of area to bounding box area) of the segmented region.
t_energ	Texture energy of the segmented region.
t_contr	Texture contrast of the segmented region.
t_corr	Texture correlation of the segmented region.
t_sosvh	Sum of squares variance of the texture.
t_homo	Texture homogeneity of the segmented region.
t_savgh	Sum average of the texture.
t_svarh	Sum variance of the texture.
t_senth	Sum entropy of the texture.
t_entro	Entropy of the texture.
t_dvarh	Difference variance of the texture.
t_denth	Difference entropy of the texture.
t_inf1h	First information measure of correlation.
t_inf2h	Second information measure of correlation.
classification	Classification of the lesion as either "Malign" (malignant) or "Benign".

The target variable in the dataset is classification, which indicates whether a patient's lesion is "Malign" (malignant) or "Benign".

Program

The program implements a command-line interface (CLI) with the following functionalities:

1. LOAD: Load a specified dataset file and display summarized information.
2. LOADF: Load the provided bcdr_f01_features.csv file and display summarized information.
3. CLEAR: Clear the loaded data from memory.
4. QUIT: Exit the program.
5. DESCRIBE: Provide a statistical summary of the numerical data and count benign and malignant cases.
6. SORT: Sort the data by patient ID, handle missing values, and encode the classification labels.
7. CORRELATION: Remove irrelevant features and visualize correlations using a heatmap.
8. SPLITSCALE: Split the dataset into training and test sets and scale features.
9. SVM: Train and test a Support Vector Machine (SVM) classifier (already fine-tuned with Random Search).
10. RANDOMFOREST: Train and test a Random Forest classifier (already fine-tuned with Random Search).
11. ANN: Train and test an Artificial Neural Network (ANN) classifier (already fine-tuned with Random Search).
12. METRICS: Evaluate and display classification performance metrics (Confusion Matrix, Accuracy, Precision, Recall) for all the models developed.

Project Structure

• /scripts: Contains the main script main.py that implements the above functionalities.
• /data: Directory to store datasets.
• /results: Directory to save results for each run, such as visualizations and model outputs.
• /logs: Directory to save logs for each run, including prints, error messages and code details.
• Dockerfile: Instructions to containerize the application.
• requirements.txt: List of Python dependencies.
• README.md: This file.

Requirements

The code is written in Python and requires the following libraries:

• Pandas
• Requests
• NumPy
• Scikit-learn
• Matplotlib
• Seaborn

Installation

1. Clone the repository:

   git clone https://github.com/HugoTex98/Breast-Cancer-Prediction.git
   cd Breast-Cancer-Prediction

2. Create a virtual environment and activate it:

   python3 -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install the required dependencies:

   pip install -r requirements.txt

Building and Running the Docker Container

To build and run the Docker container for this project, follow the steps below:

1. Build the Docker Image

First, navigate to the root directory of the project (where Dockerfile is located) and run the following command to build the Docker image:

docker build -t breast-cancer-prediction .

This command builds the Docker image using the instructions in the Dockerfile and tags it as breast-cancer-prediction.

2. Run the Docker Container

Once the image is built, you can run the Docker container using the following command:

docker run -it --rm breast-cancer-prediction

• -it: Runs the container in interactive mode, allowing you to interact with the terminal inside the container.
• --rm: Automatically removes the container once it stops running, keeping your environment clean.
• breast-cancer-prediction: The name of the Docker image you built.

Since the project is not a web application and does not expose any ports, the output from the script will be directly visible in the terminal where the container runs. If the script generates output files, they will be saved in the container's file system.

To stop the container while it’s running, you can do so by pressing Ctrl + C in the terminal where the container is running.

Usage

Run the program using the command-line interface:

python main.py

Follow the prompts to load data, process it, explore the features, and use Machine Learning models to predict Breast Cancer.

Future Improvements

In the future, I plan to implement the following improvements and features to enhance the functionality and performance of this project:

1. Expand Model Selection:

   • Integrate additional machine learning models, such as Gradient Boosting Machines (GBM) or XGBoost, to compare performance with the current models.

2. Model Comparison module:

   • Implement a module inside Metrics program to evaluate which model is the best (considering the best metric for the use case).

3. Hyperparameter Optimization:

   • Implement a module to use Optuna to optimize the performance of the classifiers.

4. Data Augmentation:

   • Apply data augmentation techniques (SMOTE) to increase the dataset and potentially improve model accuracy.

5. Containerization and Deployment:

   • Refine the Docker setup to allow for easy deployment on cloud platforms like AWS or Azure, including CI/CD pipeline integration.

6. Testing and Validation:

   • Implement unit tests and continuous integration (CI) pipelines to ensure code quality and detect potential issues before deployment (maybe ML Flow for model monitoring).

7. AutoML

   • Maybe implement some AutoML like PyCaret to speed up modelling.