Medical Imaging of Brain Hemorrhage CT Scan Using PDE-Based Filtering with Feature Extraction and Classification

Resumen

Objectives: To propose a suitable technique for employing Computed Tomography (CT) scans to identify brain hemorrhage.

Methods:  Images of CT scans of the brain were collected from the open-source Kaggle website. Two hundred photos total—one hundred normal and one hundred affected—are included in that dataset. In image filtering, Ω, f, ƞ, u, and λ were the parameters that indicated the test image function, noise-containing picture, extra noise, image solution, and smoothing function. An analysis was conducted to compare the accuracy of various feature extraction methods. Features were extracted from the head CT scans using the NGTDM, LGP, and GLSZM methods.

Findings: The study found that the GLSZM feature extraction method achieved the highest precision in detecting brain hemorrhage using a Logistic Regression classifier, outperforming NGTDM and LGP. Noise reduction and evaluation metrics validated the method's effectiveness across various datasets.

Novelty: The dataset on brain hemorrhage is processed using the ROF filter, a PDE-based filter. The accuracy of the results produced by this novel method is 95.01%.