Fusion of Images to Detect Brain Tumors Using DWT, IDWT and VGG19 Algorithms

Detecting brain tumors from medical images is a complex task in medical image analysis and is crucial for accurate diagnosis and treatment planning. Over the past few years, there has been significant development in the field of medical image processing. One such technique that has been employed is the multi-scale decomposition of brain images, which is achieved through the discrete wavelet transform (DWT). DWT decomposes an image into two different components, structural and textural information, without compressing the image. As a result, it provides high accuracy details and it preserves edge and texture details when reconstructing the image from its original frequency, and reduces problems like blocking and ringing artifacts. The low frequency sub-band coefficients are fused by selecting the coefficient with the maximum spatial frequency, indicating the overall active level of an image, while the high frequency subband coefficients are fused by selecting the coefficient with the maximum code value. Brain tumor detection using wavelet transform involves the use of both CT scan and MRI scan images. CT scan uses X-ray images taken from different angles of a specific part of the human body to provide detailed information about its internal structure, while MRI scan employs strong magnetic fields, radio waves, and field gradients to generate images of the inner human body. Finally, the fused image is reconstructed by taking the inverse Inverse Discrete Wavelet Transformation (IDWT) of two varied frequency sub-bands. The methods such as Visual Geometry Group (VGG19) which is a sub class of Convolutional Neural Network (CNN), watershed algorithm, Procrustes Analysis Algorithm, data argumentation are used in registration, fusion and segmentation process to detect brain tumor. Overall, the application of DWT and the fusion of multi-scale components of brain images has significant potential in improving the accuracy and quality of medical image processing, especially in the detection of brain tumors.