Purpose
Infections and tumor are common brain diseases that, sometimes, may have similar pattern on brain MRI. Neuroinfections can be caused by diverse etiological agents and affect millions of people worldwide [1, 2]. Altogether, benign and malignant tumors of the Central Nervous System (CNS) are accounted for 3.5/100.000 new cases and 2.5 deaths worldwide [3], exhibiting life-style and tumor type-dependent prognostic (5% to 92% with 5 years survival rate) [4].
Neuroinfections and CNS tumors may produce severe complications, disability, and economic burden [2]. The diagnosis of...
Methods and materials
Image Bank
Retrospective examinations were collected from a 3.0 Tesla MRI scanner in Botucatu and Campinas Medical School Hospital (UNESP and UNICAMP – Brazil). Patients were selected according to the following criteria:
Inclusion criteria: patients who underwent MRI exams between 2010 and 2017 with diagnosis of inflammation or tumors. All selected patients had their diagnosis confirmed by histopathological or CSF analysis examinations and follow-up exams.
Exclusion criteria: patients who had brain biopsy or surgery before the MRI acquisition, lesions smaller than 10mm, and patients with...
Results
The 63 features extracted are presented in Table 2 .All 63 extracted features were analyzed with the statistical methods Information Gain Ratio and Gini Decrease, and 22 presented identical values in all ROIs and sequences (Gain Ratio=0 and Gini=0). Therefore, all were excluded in the ML classification.
The five best-ranked features, calculated on Orange Canvas software, was different in each MRI images, however EntropyWv and Mean remained constant in all five groups.
After the classification, we obtained AUC, CA, F1, Precision, and Recall. To each...
Conclusion
The values of AUC, CA, F1, Precision and Recall, the results confusion matrix and ROC curves and the comparisons between classification methods and MRI sequences, suggests that the algorithm obtained textures capable of differentiate brain tumors of neuroinfections in MRI images. The best sequence for classification with this method of texture extraction was T1-weighted images without contrast. Sequences T2 (T2-weighted and FLAIR) and T1 with contrast did not have a good classification performance. We also conclude that the five best-ranked features were more efficient than...
Personal information and conflict of interest
A. F. Alves; Botucatu/BR - nothing to disclose L. deM. Feitoza; Campinas/BR - nothing to disclose F. Reis; Campinas/BR - nothing to disclose L. L. Alves; Botucatu/BR - nothing to disclose D. R. de Pina; Botucatu/BR - nothing to disclose T. Castro; Campinas/BR - nothing to disclose
References
[1] World Health Organization. (2016) What are neurological disorders? In: World Heal. Organ. https://www.who.int/features/qa/55/en/. Accessed 4 Nov 2018
[2] World Health Organization. (2006) Neurological Disorders: Public Health Challenges. World Health Organization
[3] International Agency for Research on Cancer (2019) Estimated number of new cases in 2018, worldwide, both sexes, all ages. In: Glob. Cancer Obs. https://gco.iarc.fr/today/home. Accessed 25 Apr 2019
[4]Team TACS medical and editorial content (2017) Survival Rates for Selected Adult Brain and Spinal Cord Tumors. In: Am. Cancer Soc. https://www.cancer.org/cancer/brain-spinal-cord-tumors-adults/detection-diagnosis-staging/survival-rates.html. Accessed 29 Apr...