Neural Network-Support Vector Machine for Sinusitis Classification

Computed Tomography (Ct) Scan Assisted Machine Learning in the Management of Artifacts Related to Paranasal Sinuses and Anterior Cranial Fossa

“…For example, when using iterative algorithms, matrices are recalculated in all possible directions to obtain new sinograms. When using other statistical methods, corrections are made in the geometry using the incident ray beam approach, which also requires more computational work and, as a result, has a higher GPU cost [20][21][22]. From the foregoing, we can infer the followings: (i) High contrast and superior quality images (ii) Requires fewer projection data for image reconstruction (iii) Considerable reduction in exposure dose (iv) Longer reconstruction time (v) High computational costs Hybrid architectures that use both reconstruction methods and parallel algorithms are being developed to create more efficient reconstruction mechanisms.…”

“…Higher image resolution means a larger initial matrix and more equations per pixel, which equals more tomograph detectors. Parallel algorithms and filtering techniques such as Gaussian, median, Wiener, and bilateral filters are being implemented to improve data acquisition efficiency [21].…”

“…As we have discussed, several ways to "clean" an image efficiently with statistical methods require complex adaptation to current hardware systems. It works with GPU systems, a graphic processing unit, to speed up iterative data processing, implement new algorithms, and achieve maximum reduction [17,20,21]. Statistical methods have the advantage of not requiring a large number of projections and data to reconstruct later.…”

“…To build a hyperplane that separates two classes, labelled x, 1, +1, so that the distance between the optimal hyperplane and the closest training pattern margin is the maximum [21]. Within the classification problems are the linearly separable and nonlinearly separable cases.…”