Selecting an Optimizer to Detect the Field of View for CT and MRI Brain Images

• Published in: 2021 IEEE 18th India Council International Conference (INDICON) pp. 1 - 5
• Main Authors: Chandrashekar, Leena, Dinesh, M N
• Format: Conference Proceeding
• Language: English
• Published: IEEE 19.12.2021
• Subjects: ADAM optimizer; Computed tomography; Convolutional neural networks; Deep Learning; Feature extraction; Image registration; Image resolution; Magnetic resonance imaging; Task analysis
• ISSN: 2325-9418
• DOI: 10.1109/INDICON52576.2021.9691763

Medical image fusion is the latest development in the field of medical imaging that deals with integrating multimodal and multiresolution images. However, before integrating these images, they must be aligned for field of view, resolution and scale. Image registration deals with aligning the images with respect to reference images. Feature extraction, feature matching, optimization and measure of similarity metric are steps involved in the registration. Deep learning approaches can be adopted to align the images with lesser algorithm complexity and in absence of reference images. optimizers are significant in design of classifiers, as they influence the convergence of the local minima, weights of the network and learning process.The objective of the paper is to understand the role of optimizer and choose a suitable one for detecting the field of view angle for CT and MRI images.The ADAM optimizer provides the classification accuracy of 91.4% for 10 classes of field of view angles ranging from 0 to 20 degrees. Images with 0 degrees, 4 degrees and 6 degrees field of view angle are detected with 100%, 97.4% and 96.2% respectively.