Automated Quality Control for Proton Magnetic Resonance Spectroscopy Data Using Convex Non-negative Matrix Factorization

Abstract

Abstract. Proton Magnetic Resonance Spectroscopy (1 H MRS) has proven its diagnostic potential in a variety of conditions. However, MRS is not yet widely used in clinical routine because of the lack of experts on its diagnostic interpretation. Although data-based decision support systems exist to aid diagnosis, they often take for granted that the data is of good quality, which is not always the case in a real application context. Systems based on models built with bad quality data are likely to underperform in their decision support tasks. In this study, we propose a system to filter out such bad quality data. It is based on convex Non-Negative Matrix Factorization models, used as a dimensionality reduction procedure, and on the use of several classifiers to discriminate between good and bad quality data.

Quality of clinical brain tumor MR spectra judged by humans and machine learning tools

“…This was the case when the third class was retrospectively and heuristically introduced using the MER scores as those cases that had been accepted by one expert who had been overridden by two others (Table , #17), but it was also the case where we had prospectively used a 10‐point scoring system that could easily be used as three‐class grading base (Table ). A similar trial of binary and three‐class investigation has recently been reported , where the same dataset as in our investigations had been used. A convex non‐negative matrix factorization for feature extraction and different classifiers had been tried to develop an automated quality control protocol.…”

Embedding MRI information into MRSI data source extraction improves brain tumour delineation in animal models

“…These results were in line to those from Group A (excluding a few cases such as the three sources calculated for mouse C179), in the sense that the sources were not strikingly visually different, except for case C583, in which both the unsupervised tumour (red) and normal (blue) sources corresponded to patterns that matched with low signal-to-noise spectra. This is a problem that has been encountered and characterised before [39]. In the case of the non-tumour (blue) sources, they differed more than the tumour (red) source between the two approaches in all five mice.…”

“…In addition to all the aspects mentioned before, both semi-supervised approaches have shown the ability to learn more meaningful, better-quality sources, as a way to overcome the susceptibility to the presence of artefacts and the lower signal-to-noise spectra issues reported in [39], with SSSE providing more accurate results according to the gold standard.…”

Extraction of artefactual MRS patterns from a large database using non‐negative matrix factorization

“…As mentioned in the introduction, we hypothesize that some of the sources would be identified as artefacts, while others will describe prototypical tumor patterns or normal tissue, as the databases from which the spectra are obtained comprise spectra of both good and poor quality. CNMF was implemented in Python language 20 and run either via Google Cloud Platform, or at the computer cluster at the Institut de Biotecnologia i Biomedicina (IBB) in Cerdanyola del Vallès, Spain.…”