Improving the classification of brain tumors in mice with perturbation enhanced (PE)-MRSI

Abstract

Classifiers based on statistical pattern recognition analysis of MRSI data are becoming important tools for the non-invasive diagnosis of human brain tumors. Here we investigate the potential interest of perturbation-enhanced MRSI (PE-MRSI), in this case acute hyperglycemia, for improving the discrimination between mouse brain MRS patterns of glioblastoma multiforme (GBM), oligodendroglioma (ODG), and non-tumor brain parenchyma (NT). Six GBM-bearing mice and three ODG-bearing mice were scanned at 7 Tesla by PRESS-MRSI with 12 and 136 ms echo-time, during euglycemia (Eug) and also during induced acute hyperglycemia (Hyp), generating altogether four datasets per animal (echo time + glycemic condition): 12Eug, 136Eug, 12Hyp, and 136Hyp. For classifier development all spectral vectors (spv) selected from the MRSI matrix were unit length normalized (UL2) and used either as a training set (76 GBM spv, four mice; 70 ODG spv, two mice; 54 NT spv) or as an independent testing set (61 GBM spv, two mice; 31 ODG, one mouse; 23 NT spv). All Fisher's LDA classifiers obtained were evaluated as far as their descriptive performance-correctly classified cases of the training set (bootstrapping)-and predictive accuracy-balanced error rate of independent testing set classification. MRSI-based classifiers at 12Hyp were consistently more efficient in separating GBM, ODG, and NT regions, with overall accuracies always >80% and up to 95-96%; remaining classifiers were within the 48-85% range. This was also confirmed by user-independent selection of training and testing sets, using leave-one-out (LOO). This highlights the potential interest of perturbation-enhanced MRSI protocols for improving the non-invasive characterization of preclinical brain tumors.

Convex Non-Negative Matrix Factorization for Brain Tumor Delimitation from MRSI Data

“…The VOI of the same mouse is further detailed in figure 6, enlarged and overlaid with the 10×10 MRSI spectral matrix at LTE. The areas delimited by red and blue lines correspond to characteristic tumor and non-tumor labels, respectively, labeled as in [32]. These labels will be referred to, later on in the study, as “supervised”.…”

“…An additional way of validating the obtained sources makes use of the labeling procedure described in [32], to compare the sources obtained with the mean spectra of tumor and non-tumor regions, similarly to the Ki-67 threshold validation described above. In [32], subsets of tumoral and non-tumoral regions were labeled, for each of the investigated mice, according to the following criteria: first, the spectra should not correspond to voxels at the edge of the PRESS-VOI, where signal to noise ratio (SNR) tends to be lower; and second, as in [33], they had not been collected over, or close to, the tumor borderline, to avoid as much as possible voxel ‘bleeding’ between tumor/non-tumor regions.…”

“…In [32], subsets of tumoral and non-tumoral regions were labeled, for each of the investigated mice, according to the following criteria: first, the spectra should not correspond to voxels at the edge of the PRESS-VOI, where signal to noise ratio (SNR) tends to be lower; and second, as in [33], they had not been collected over, or close to, the tumor borderline, to avoid as much as possible voxel ‘bleeding’ between tumor/non-tumor regions.…”

Metabolomics of Therapy Response in Preclinical Glioblastoma: A Multi-Slice MRSI-Based Volumetric Analysis for Noninvasive Assessment of Temozolomide Treatment

“…Furthermore, the rich information contained in MRS/MRSI signals makes them ideally suited to the application of statistical pattern recognition (PR) techniques [10], which are used nowadays to perform automatic categorization of individual MRSI data obtained from different types of tissue. This approach can be applied to MRSI data from human brain tumors [11] and preclinical animal models [12]. PR techniques have also been proven useful to detect and characterize tumor response to therapy [13] with generation of nosological images of therapy response.…”

Semi-supervised source extraction methodology for the nosological imaging of glioblastoma response to therapy

“…The robustness of this assumption was later on evaluated using the test set described before. Most of the voxels in the edges of the grid were discarded (except in case C586) due to low signal to noise ratio (SNR); and those located in the tumor boundaries with normal tissue were also discarded to avoid spectral pattern mixing or contamination from different tissue types essentially as described in [4], [13].…”