Medical Image Processing
Medical Image Processing
What is Medical Image Processing?
Medical imaging processing is the special subclass of digital image processing for enlightening the actual functioning of interior human body parts (organ/tissue/bone/muscle) by processing their medical images.
What is medical imaging used for?
Medical imaging is defined as the process involved in acquiring the images of various internal human parts (i.e., medical images) through various digitalized medical equipment and technologies. Once the images or bio-signals are collected, then the
image processing techniques like preprocessing, segmentation, enhancement, classification are applied
to
predict, monitor, and diagnosis the patient’s clinical state.
What kind of medical scans are there?
Types of Medical Imaging Scans
X-Ray
Ultrasound
Computerized Tomography (CT)
Magnetic Resonance Imaging (MRI)
Computerized Axial Tomography (CAT)
In recent days, the use of Artificial intelligence in medical image processing is extensively increasing. By integrating this technology with image processing, the medical disorders or abnormalities are easy to detect and assess efficiently.
Artificial Intelligence Algorithms for Medical Image Processing
Deep Learning
Zero-shot Learning
Federated Learning
Transfer Learning
Self-Supervised Learning
Multimodal Deep Learning
Reinforcement Learning
Geometric Deep Learning
Distributed Machine Learning
Deep Learning Model Complexity
Weakly Supervised Learning
Incremental Deep Learning
Physics-based Machine Learning Model
Deep Neural Network (Optimization)
Latest Medical Image Processing Project
Project Title: Implementation of effective CNN based active contour model for Early Cancer Detection
Datasets: Wisconsin Dataset (Breast Cancer) and MICCAIBraTS 2013 (Brain Tumor)
Input Images: MRI or PET/CT scans Images
Images: MRI or PET/CT scans Images
The main objective of this proposed work is to find the early-stage cancer tumor on medical images through advanced detection, segmentation, and classification image processing techniques. Also, it is intended to analyze the deep CNN-based Active Contour Model. In this proposed work, there are 5 phases that need to be performed to reach our aim. And they are given as follows,
Pre-processing
Tumor Segmentation
Feature Extraction
Feature Selection
Classification
Now, let’s see the task allocated for each phase and what are techniques and algorithms are used in these phases. As well, we can recognize how the flow of work is maintained throughout the execution.
Pre-Processing
Fast Bilateral Filter – To eliminate the noise and smoothen the image
Zig-Zag Order CLAHE – To improve the contrast of the image
Average Intensity Replacement
Patches Conversion
Tumor Segmentation
Convolutional Neural Network (CNN) – To segment the tumor part
Active Contour Models (ACMs) – To find false positive and false negative
Parameters: heterogeneous lesions, intensity inhomogeneous lesions and low contrast lesions
Deep Learning – To improve accuracy of segmentation
Feature Extraction
Fisher Vector Encoding (FVE) – To select the features for extraction
8-directions (3150, 2700, 2250, 1800, 1350, 900, 450 and 00)
Feature Selection
Particle Swarm Optimization (PSO) – To identify the optimal features
Classification
Support Vector Machine (SVM) – To classify the cancer tumor
At the end of implementation, we show that our proposed solution brings the best outcome in comparison to existing methodologies through the following performance metrics,
Computation Time
Classification Accuracy
Segmentation Accuracy
Compiled by
Ms Naresh kuwar
Comments
Post a Comment