Classification of Leukemia White Blood Cell Cancer using Image Processing and Machine Learning
Leukemia, a life-threatening form of blood cancer, has spurred significant research efforts in the quest for accurate diagnosis and treatment. This project presents a robust and accurate system for the classification of leukemia white blood cell cancer using Image Processing and Machine Learning techniques, primarily implemented in MATLAB.
The proposed system exhibits a remarkable achievement, boasting an impressive accuracy rate of 97%. The classification methodology is based on the K-Nearest Neighbors (KNN) algorithm, which leverages the distinctive features extracted from leukemia cell images to make precise categorizations.
The system’s workflow can be divided into several key stages as Image acquisition, Preprocessing, Segmentation, Feature Extraction, Classification and Performance Analysis. Preprocessing several sub-processes, including normalization, contrast enhancement, and noise removal. Normalized results, contrast-enhanced images, and denoised representations are obtained. The segmentation phase comprises color segmentation, initial segmentation, and the final detection of cancer cell nucleus regions. The results include the visualization of these regions and the identification of cancer cell nuclei.
Feature extraction is a pivotal stage where critical information is gleaned from the nucleus regions. Statistical color feature extraction is conducted on the R, G, and B channels, including the calculation of mean and standard deviation. Additionally, texture features, such as contrast, correlation, energy, and homogeneity, are extracted, providing a comprehensive representation of the cell nuclei. The heart of the system’s intelligence lies in its classification module. The final feature vectors are constructed from the extracted features. These vectors are then employed to make accurate classifications based on the KNN algorithm. The results of this stage include the predicted class labels and the classified outcomes, effectively categorizing leukemia cell types into classes such as ALL-L1, ALL-L2, AML-M2, and AML-M5.
In conclusion, the proposed system for the classification of leukemia white blood cell cancer not only demonstrates a remarkable 97% accuracy but also offers a comprehensive and visually interpretable analysis of the image processing and machine learning pipeline. By providing precise categorizations and thorough performance assessments, this system contributes significantly to the early and accurate diagnosis of leukemia, potentially saving lives and improving patient outcomes.
PROJECT OUTPUT VIDEO:
ALGORITHM / MODEL USED:
K-Nearest Neighbors (KNN)
- System : Pentium i3 Processor.
- Hard Disk : 500 GB.
- Monitor : 15’’ LED.
- Input Devices : Keyboard, Mouse.
- Ram : 8 GB.
- Operating system : Windows 10 Pro.
- Coding Language : MATLAB
- Tool : MATLABR2023B
AMREEN BATOOL AND YUNG-CHEOL BYUN, “Lightweight EfficientNetB3 Model Based on Depthwise Separable Convolutions for Enhancing Classification of Leukemia White Blood Cell Images” IEEE Access (Volume: 11), 2023.