Related Experiment Video
Updated: Apr 26, 2026
Quantifying the Brain Metastatic Tumor Micro-Environment using an Organ-On-A Chip 3D Model, Machine Learning, and Confocal Tomography
Published on: August 16, 2020
Multicentre evaluation of the INTERPRET decision support system 2.0 for brain tumour classification
Margarida Julià-Sapé1, Carles Majós, Àngels Camins
1Centro de Investigación Biomédica en Red en Bioingeniería, Biomateriales y Nanomedicina (CIBER-BBN), Cerdanyola del Vallès, Spain; Departament de Bioquímica i Biologia Molecular, Unitat de Bioquímica de Biociències, Edifici Cs, Universitat Autònoma de Barcelona, UAB, Cerdanyola del Vallès, Spain; Institut de Biotecnologia i de Biomedicina (IBB), Universitat Autònoma de Barcelona, UAB, Cerdanyola del Vallès, Spain.
The International Network for Pattern Recognition of Tumours Using Magnetic Resonance Decision Support System (INTERPRET DSS) 2.0 did not improve diagnostic accuracy for brain tumors compared to version 1.0. Adding a long-TE classifier did not enhance performance, even for clinicians with minimal spectroscopy training.
Area of Science:
- Neuroradiology
- Medical Imaging Analysis
- Machine Learning in Medicine
Background:
- Proton magnetic resonance spectroscopy ((1)H MRS) aids in characterizing adult human brain tumors.
- The International Network for Pattern Recognition of Tumours Using Magnetic Resonance Decision Support System (INTERPRET DSS) 1.0, utilizing a short-TE classifier, previously showed significant value.
- INTERPRET DSS 2.0 was developed with an added long-TE classifier to potentially improve diagnostic capabilities.
Purpose of the Study:
- To evaluate if clinicians with limited spectroscopy experience can achieve comparable results to spectroscopists using the INTERPRET DSS.
- To determine if the updated INTERPRET DSS 2.0, with an additional long-TE classifier, offers superior diagnostic accuracy compared to the initial version.
- To assess the diagnostic performance of DSS 2.0 against DSS 1.0 using a flexible analysis protocol mimicking clinical practice.
Main Methods:
- A second study was conducted with nine evaluators and the same brain tumor cases used in a prior study.
- Evaluators included neuroradiologists and spectroscopists, with minimal training provided for system use.
- The analysis protocol was designed to be flexible, simulating a real-world clinical environment.
Main Results:
- Most tumor classes and superclasses yielded similar diagnostic results between DSS 1.0 and DSS 2.0.
- A notable decrease in performance (AUC) was observed for astrocytomas of World Health Organization (WHO) grade III with DSS 2.0 (0.62) compared to DSS 1.0 (0.87).
- The inclusion of two echo times (TEs) instead of one did not significantly impact the evaluation results for either version.
Conclusions:
- The INTERPRET DSS 2.0, with its added long-TE classifier, did not demonstrate improved diagnostic accuracy for brain tumors over DSS 1.0.
- Clinicians with minimal spectroscopy training performed comparably to spectroscopists when using the DSS system.
- The diagnostic performance for WHO grade III astrocytomas specifically declined with the updated version of the DSS.
