The goal of this project is to develop an artificial-intelligence model able to identify the malignancy of tumour masses in breast-cancer medical images.
The main milestones include:
Data Collection and Preparation
: A dataset of medical images labelled with positive and negative breast-cancer cases was collected and prepared.Data Reduction
: Transfer learning was used to reduce the size of the initial dataset, leveraging pre-trained models.Exploratory Data Analysis
: An exploratory analysis was performed to understand the characteristics of the dataset.Convolutional Neural Network (CNN) Models
: CNN-based models were built to detect breast cancer in medical images, extracting relevant features.Training and Parameter Tuning
: The model was trained on the prepared data and the network parameters were tuned. Accuracy was measured on the training set.Hyperparameter Optimisation
: The optimal hyperparameter configuration — number of neurons, layers and iterations — was searched for the images.Results Validation
: Results were validated using a separate test dataset.
- Master's Thesis ReportDownload PDF here
- CodeModels Code,Web APP
- PlatformWeb, Jupyter Notebook
- StackPython, ML Libraries (PyTorch, Keras, TensorFlow, Scikit-Learn), Flask
Multimedia
