Intracranial lesion classification: Semi-automated MR spectroscopy software package (INTERPRET) versus histology/clinical follow up. Abstract: Background: Over the last decade, the use of MR Spectroscopy has increased significantly. However it is not routinely used in clinical practice. Semi-automated software packages containing validated data have been created to try and increase the use of MR Spectroscopy. This project aims to retrospectively analyse Magnetic Resonance (MR) Spectra, acquired in clinical practice, comparing the semi-automated software package (INTERPRET) with histology/clinical follow up. The project study incorporates MR spectra collected over a 5 year period and formal histology from biopsy. In those cases where biopsy was not undertaken clinical/radiological follow up was used as a comparator. Results: 55 patients with adequate MR spectra were identified. Cases were split into tumour and non-tumour groups. Within both groups, the INTERPRET package was unable to replicate accuracy rates published in the literature. An accuracy of 51.0% was found for the tumour group. Within the smaller non-tumour group, INTERPRET could not identify whether the lesion was tumour or pseudo-tumour in approximately half the cases. Conclusion: Within this retrospective study, incorporating data acquired in clinical practice, I was unable to replicate accuracy rates published in the literature for the semi-automated INTERPRET MR Spectroscopy software package. Project type: Analysis of existing data Imaging keywords: Magnetic resonance spectroscopy (MRS) Application / disease keywords: Tumour Supervisor(s): Professor D Hadley Dr Michael Thrippleton Programme: Neuroimaging for Research MSc Year: 13-14 This article was published on 2024-08-22