Generating prior probabilities for classifiers of brain tumours using belief networks

Abstract

Background: Numerous methods for classifying brain tumours based on magnetic resonance spectra and imaging have been presented in the last 15 years. Generally, these methods use supervised machine learning to develop a classifier from a database of cases for which the diagnosis is already known. However, little has been published on developing classifiers based on mixed modalities, e.g. combining imaging information with spectroscopy. In this work a method of generating probabilities of tumour class from anatomical location is presented.

Multivariate Analysis in Pediatric Brain Tumor

“…Among them, 13 articles reported multivariate analysis methods applied to pediatric brain tumor classification (or segmentation) [3][4][5][6][7][8][9][10][11][12][13][14][15]. Table 1 provides a summary of these papers.…”

“…They found that combined ADC and metabolite ratio features (using water as an internal standard) could discriminate between these tumor groups [4]. Reynolds et al used tumor location information obtained from MRI to construct Bayesian belief networks, incorporated MRS information into the classifiers, and found that inclusion of a priori knowledge improved classification accuracy [5]. Later, Davies et al extracted multiple metabolite profiles from (short-TE PRESS) MRS, classified tumors (e.g., glial-cell vs. non-glial-cell tumors) with LDA and achieved high classification accuracies (>90%) [7].…”

Combining magnetic resonance spectroscopy and molecular genomics offers better accuracy in brain tumor typing and prediction of survival than either methodology alone

“…In many cases, successful differentiation using both linear and nonlinear methods can be made based on single resonance peaks or ratios of resonance ranges (35). More recent work on brain tumors has shown that classification according to histological type and grade is possible using similar approaches, particularly linear discriminant analysis (LDA) after feature extraction with independent components analysis (ICA) in a Bayesian framework (41) or correlation analysis and stepwise LDA (36) or using belief networks (42) or using Support Vector Machines (SVMs) (22). Also, our approach in fusing genomics and MRS to improve the typing and prognosis of human brain tumors agrees with the notion that fusion of different sources of information can improve system performance and facilitate detection, recognition, identification, tracking, change detection, and decision-making in defense, robotics, and Table II.…”

“…While considerable work has been done on the classification of cancers based on genomic data (23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33), and some work has been done using MRS data (22,(34)(35)(36)(37)(38)(39)(40)(41)(42), these two datasets have yet to be integrated along with other clinical features. We hope to improve cancer diagnosis accuracy immediately following biopsy collection by uncovering and exploiting complementary information in the MRS and genomic data.…”

Review on Histopathological Slide Analysis using Digital Microscopy

“…In the citation by G. M. Reynolds et al, [67], this method has been used as a supervised machine learning which has developed a classifier from the database of cases for which this diagnosis has already been known. However, the developing classifiers based on mixed modalities, like combining imaging information which has generated better probability of a class containing set which helps in detecting benign tissue.…”