Brain Tumor Detector

Brain Tumor Detector is a data science and machine learning project that is the 5th and final Project of the Code Institute's Bootcamp in Full Stack Software Development with specialization in Predictive Analytics. The business goal of this project is the differentiation of the healthy brain and the one with the tumor based on the brain MRI scan images. The project is realised with the Streamlit Dashboard and gives to the client a possibility to upload the MRI brain scan in order to predict the possible tumor diagnosis. The dashboard offers the results of the data analysis, description and the analysis of the project's hypothesis, and details about the performance of the machine learning model. The project includes a series of Jupyter Notebooks that represent a pipeline that includes: importation and cleaning of the data, data visualization, development and evaluation of the deep learning model. Quick

Project Summary

MRI Visualizer - The first business objective of the project is addressed by the MRI Visualizer page, which focuses on Data Analysis. This page includes plots that can be toggled on and off using the built-in toolbar. Examples of these plots are provided below.

MRI Visualizer

Additionally, this app page offers a tool for creating image montages. Users can select a label class (tumor or non-tumor) and view a montage generated through graphical presentation of random validation set images.

Montage

Model Performance - The dataset size and label frequencies, which indicate the initial imbalance of the target, are documented on this page. Additionally, the history and evaluation of the project's machine learning model are provided. The paired graphs display the validation loss and accuracy per epoch, showcasing the model's progress over time. Furthermore, a confusion matrix illustrating the predicted and actual outcomes for the test set is presented.

Model Performance

Brain Tumor Detection - tool fulfills the second ML business objective of the project. It provides access to the original raw dataset, allowing users to download MRI brain scans. These images can then be uploaded to receive a class prediction output generated by the model.

Brain Tumor Detection

Here are some examples of the outputs, namely, a binary class prediction, a graphical representation showing percentages, and the option to download the output DataFrame as a CSV file.

Brain Tumor Detection Outputs

Project Hypothesis This application page showcases written documentation of the project's hypotheses and analysis of the findings, demonstrating their alignment with the aforementioned hypotheses. The contents is similar to the one in this documentation.

Project Hypothesis