| Full text | |
| Author(s): |
Souza, S. A. S.
;
Guassu, R. A. C.
;
Alves, A. F. F.
;
Alvarez, M.
;
Pitanga, L. C. C.
;
Reis, F.
;
Vacavant, A.
;
Miranda, J. R. A.
;
Filho, J. C. S. Trindade
;
Pina, D. R.
Total Authors: 10
|
| Document type: | Journal article |
| Source: | MULTIMEDIA TOOLS AND APPLICATIONS; v. N/A, p. 13-pg., 2023-10-03. |
| Abstract | |
We propose a machine learning (ML) approach applied to texture features to differentiate primary brain tumors and solitary brain metastasis. Magnetic resonance imaging (MRI) exams of 96 patients were divided into primary tumors (38) and solitary brain metastasis (58). MRI sequences used: diffusion-weighted image (DWI), fluid-attenuated inversion recovery, T1-weighted, T1-weighted SE gadolinium-enhanced, and T2-weighted images. Regions of interest (ROIs) of 10 x 10 pixels were positioned within the tumors. For each ROI, 40 texture features were extracted and applied to five different ML methods: naive bayes, support vector machine (SVM), stochastic gradient descent, random forest, and tree. The ML methods classified the groups with good differentiation of up to 97.5% of the area under the receiver operator characteristics (ROC) for SVM as the best classifier, especially in the DWI sequence. The method has a reliable classification for the investigation of tumor lesions. (AU) | |
| FAPESP's process: | 20/05539-9 - Quantifications of COVID-19 radiological pulmonary structures (Coronavirus infection) |
| Grantee: | Diana Rodrigues de Pina Miranda |
| Support Opportunities: | Regular Research Grants |