Title: - Semantic Segmentation and Breast Cancer Detection in Mammogram Images Using Deep Learning Approach
---------------------------------------------------------------------------------------------------------------
Implementation Plan:
---------------------
Step 1: Initially we load the mammograms images from private(Mammograms - Copy) and public dataset(DDSM).
Step 2: Next we perform an Image Enhancement process, In this process we remove high-quality images that increase the accuracy of segmentation and classification. The image enhancement process includes three sub-process which are explained as follows,
2.1: Noise and Artifacts Removal: Here, we have used Hybrid Filters (HF) which include improved wavelet filters and curvelet filters for noise removal and artifacts removal respectively.
2.2: Image Scaling: Here we used the Pixel based Bilinear Interpolation (PBI) algorithm for image scaling and changed the pixel information of the image.
2.3: Contrast Enhancement: Here, we have used the Election based Optimization (EO) algorithm which optimally adjusts the gamma intensity which increases the quality of the images.
Step 3: Next we perform an Semantic Segmentation process, In this process the quality enhanced images are used for semantic segmentation to increase the segmentation accuracy. Here, we used Reinforcement Learning based Semantic Segmentation (RLSS) which segments the Region of Interest (ROI) based on the pixel wise information of the images. Here, we performed two types of segmentation which are done by the smart agents presented in the reinforcement learning algorithm.
3.1: First, we considered the pectoral muscles for segmentation to increase the accuracy of breast cancer detection.
3.2: In second, we have removed the pectoral muscles and performed segmentation that reduces the misclassification due to removing pectoral muscles. Finally, we get two segmented ROIs from RLSS which are evaluated by the Jordan Closed Curve Theorem.
Step 4: Next, we perform a Multi View based Feature Extraction and Breast Cancer Detection process, In this process we extract the multi view features from the ROI. Here, features are extracted from the segmented ROI as multiple views (ex. 90°, 180°) which increases the detection accuracy. Here, convolutional neural network CNN (CNN) is used for feature extraction and classification.
Step 5: Next, To reduce the computational time and complexity, we have clustered the features based on its similarity. That increases the speed of breast cancer detection. Based on the clustered features the CNN classified the images into three classes such as normal, malignant, and benign.The severity level is denoted into three classes such as mild, moderate, and severe.
Step 6: Finally, The performance of the proposed work is evaluated in terms of following metrics,
Accuracy
Precision
Recall
F-measure
True positive rate
False positive rate
============================================================
Software Requirement:
--------------------------
1. Tool: Matlab-R2020a
2. OS: Windows 10 - (64-bit)
============================================================
Title: - Semantic Segmentation and Breast Cancer Detection in Mammogram Images Using Deep Learning Approach
--------------------------------------------------------------------------------------------------------------
Installl the required software...
Copy the zip file and paste to the D or E drive and Extract the zip file.
Copy the folder "Semantic Segmentation and Breast Cancer Detection" under the code folder and Paste into any drive.
Note: Don't Delete any file or folders project contains...
Then Launch matlab IDE.
Next copy the source code location (Note: Use the source code location avoids space). Then paste the source code location into matlab address location. After paste, press the enter key.