Multi-Cancer Identification and Segmentation

Proposed System Architecture

The project aims to be able to identify the presence of tumors in brain and breast scans. The first step of the project is to classify whether a given scan is a brain or breast scan. After determing the type of the scan, we classify whether a tumor is presnet or not, If a tumor is present it is segmented and displayed as a binary image.

Project Aims

• Apply image classification to classify each image as a brain scan or a breast scan and then determine if the scan is normal or if it has a tumor.
• Apply image segmentation to determine the exact location of the infected area.
• Determine the width and height of the tumor in the scan if it exists.

Data Set

The dataset for this project can be found Here Data Set Samples

Brain Scan with Correpsonding Mask

Breast Scan with Correpsonding Mask

Results

Below is breif description of the results of models used

Brain/Breast Classifier

• Model: VGG16-Based Classifier
• Test Loss: 0.0000
• Test Accuracy: 1.0000
• Test Precision: 1.0000
• Test Recall: 1.0000
• Test F1 Score: 1.0000

Brain / Breast Classifier Results

Brain Tumor Classifier

• Model: VGG16-Based Classifier
• Test Loss: 0.189
• Test Accuracy: 0.935
• Test Precision: 0.93
• Test Recall: 0.939
• Test F1 Score: 0.934

Brain Tumor Classifier Results

Breast Tumor Classifier

• Model: VGG16-Based Classifier
• Test Loss: 0.7698
• Test Accuracy: 0.6667
• Test Precision: 0.7197
• Test Recall: 0.6667
• Test F1 Score: 0.6694

Breast Tumor Classifier Results

Brain Tumor Segmentation

• Test Loss = 0.112
• Test Dice Coef. = 0.607
• Test Mean IoU: 0.8336
• Test Precision: 0.8354
• Test Recall: 0.7926

Brain Scans Segmentation Results

Breast Tumor Segmentation

• Test Loss = 0.2063
• Test Dice Coef. = 0.768
• Test Mean IoU: 0.624
• Test Precision: 0.806
• Test Recall: 0.734

Breast Scans Segmentation Results

System Simulation

In this section it will be demonstrated how all of the models will work together

System Simulation With Brain Scan