An Efficient Cancer Classification Model for CT/MRI/PET Fused Images

Area of Science:

• Medical imaging analysis
• Computational pathology
• Biomedical signal processing

Background:

• Early cancer detection is crucial for effective treatment planning.
• Integrating multi-modal imaging data can improve diagnostic accuracy.
• Existing image fusion techniques have limitations in capturing subtle pathological features.

Purpose of the Study:

• To develop and evaluate a novel image fusion method for enhanced cancer classification.
• To compare the performance of the proposed method against existing fusion techniques.
• To assess the efficacy of different classifiers in determining cancer type from fused images.

Main Methods:

• A Curvelet transform-based image fusion algorithm was developed.
• Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and Positron Emission Tomography (PET) scan images were fused.
• Cancerous regions were segmented using a marker-controlled watershed algorithm.
• Features (color, shape, texture) were extracted and classified using Neural Network, Random Forest, and K-Nearest Neighbors (KNN) classifiers.

Main Results:

• The Curvelet transform fusion method outperformed existing techniques in PSNR, SSIM, Entropy, Mean, and Standard Deviation.
• The K-NN classifier achieved the highest accuracy (94%), sensitivity (88%), and specificity (84%).
• The proposed fusion method combined with KNN demonstrated superior performance in classifying cancer types.

Conclusions:

• The Curvelet transform-based image fusion method is effective for medical image analysis.
• The KNN classifier provides robust performance for cancer classification when using fused imaging data.
• This approach aids in earlier and more accurate diagnosis of cancer, facilitating timely treatment planning.