Home

Discrimination of benign and malignant mammographic masses based on BI-RADS attributes and the patient’s age.

This website demonstrates the project made by Rohit Sarkar for the course Digital Medicine taught by Professor Raghupathi Cavale

Check if a tumour is malignant!!

Site Map :

  • Home : Summary of the Project
  • Predict : Input the features and get predictions in real time!
  • Report : A technical report of the Data Analysis, Visualisation and Machine Learning process in the form of a ipython notebook
  • Synopsis : Synopsis of the Project

Brief

Worldwide, breast cancer is the second most common type of cancer after lung cancer and the fifth most common cause of cancer death. Breast Cancer is the leading type of cancer in women globally, accounting for 25% of all cases. Mammography is the most popular method for breast cancer screening available today. However the results of breast biopsies resulting from the Mammogram Interpretations leads to approximately 70% benign outcomes which could have been prevented.

The goal of this Project is to provide a Computer Aided Diagnosis (CAD) method that helps physicians in classifying Mammographic Masses with the help of BI-RADS attributes and the Age of the patient. The machine learning model presented in this Project achieves a false positive rate of 15% using a Artificial Neural Network model which is significantly better than the physicians performance.

Methodology

Dataset

The dataset used for learning is the Mammographic Mass Dataset from the UCI Machine Learning Repository. Data exploration and analysis was carried out to generate useful insights.

Data Analysis and Learning

For the purpose of prediction an optimal classifier (Artificial Neural Network) was selected among 3 classifers namely, Logistic Regression, Support Vector Machine and Artificial Neural Networks. K-fold Cross Validation was carried out to select the optimal hyperparameters.

Deployment

The final model was deployed on a server running Python. A RESTful API was constructed using Python and the Flask microframework. The frontend ie this website was made using HTML, CSS, JS, Jekyll and Github Pages. For prediction the FrontEnd sends a request containing a list of features to the API Endpoint. The API then returns the predictions back to the Frontend.

Code

The code used to analyse the data and build the Neural Network model can be found in the report
The code used to make this website and the webserver running the Neural Network model can be found here

Conclusion and Future Goals

It is clear that Computer Aided Diagnosis methods can significantly aid physicians in taking informed decisions. Thus Digital Medicine is playing a crucial role in the field of Breast Cancer. In the future with more data, better models can be trained which will lead to higher accuracy.

Open Source Libraries Used

Acknowledgements

I would like to thank Professor Raghupathi Cavale and Professor Fayaz Shradgan for their guidance throughout the Digital Medicine Course.