The diagnosis of multiple sclerosis (MS) is usually based on clinical symptoms and signs of damage to the central nervous system, which is assessed using magnetic resonance imaging. The correct interpretation of these data requires excellent clinical expertise and experience. Deep neural networks aim to assist clinicians in identifying MS using imaging data. However, before such networks can be integrated into clinical workflow, it is crucial to understand their classification strategy. In this study, we propose to use a convolutional neural network to identify MS patients in combination with attribution algorithms to investigate the classification decisions. The network was trained using images acquired with susceptibility-weighted imaging (SWI), which is known to be sensitive to the presence of paramagnetic iron components and is routinely applied in imaging protocols for MS patients. Different attribution algorithms were used to the trained network resulting in heatmaps visualizing the contribution of each input voxel to the classification decision. Based on the quantitative image perturbation method, we selected DeepLIFT heatmaps for further investigation. Single-subject analysis revealed veins and adjacent voxels as signs for MS, while the population-based study revealed relevant brain areas common to most subjects in a class. This pattern was found to be stable across different echo times and also for a multi-echo trained network. Intensity analysis of the relevant voxels revealed a group difference, which was found to be primarily based on the T1w magnitude images, which are part of the SWI calculation. This difference was not observed in the phase mask data.

No abstract

BackgroundPreventing sepsis‐associated acute kidney injury (S‐AKI) can be challenging because it develops rapidly and is often asymptomatic. Probability assessment of disease progression for therapeutic follow‐up and outcome are important to intervene and prevent further damage.PurposeTo establish a noninvasive multiparametric MRI (mpMRI) tool, including T1, T2, and perfusion mapping, for probability assessment of the outcome of S‐AKI.Study TypePreclinical randomized prospective study.Animal ModelOne hundred and forty adult female SD rats (65 control and 75 sepsis).Field Strength/Sequence9.4T; T1 and perfusion map (FAIR‐EPI) and T2 map (multiecho RARE).AssessmentExperiment 1: To identify renal injury in relation to sepsis severity, serum creatinine levels were determined (31 control and 35 sepsis). Experiment 2: Animals underwent mpMRI (T1, T2, perfusion) 18 hours postsepsis. A subgroup of animals was immediately sacrificed for histology examination (nine control and seven sepsis). Result of mpMRI in follow‐up subgroup (25 control and 33 sepsis) was used to predict survival outcomes at 96 hours.Statistical TestsMann–Whitney U test, Spearman/Pearson correlation (r), P < 0.05 was considered statistically significant.ResultsSeverely ill septic animals exhibited significantly increased serum creatinine levels compared to controls (70 ± 30 vs. 34 ± 9 μmol/L, P < 0.0001). Cortical perfusion (480 ± 80 vs. 330 ± 140 mL/100 g tissue/min, P < 0.005), and cortical and medullary T2 relaxation time constants were significantly reduced compared to controls (41 ± 4 vs. 37 ± 5 msec in cortex, P < 0.05, 52 ± 7 vs. 45 ± 6 msec in medulla, P < 0.05). The combination of cortical T2 relaxation time constants and perfusion results at 18 hours could predict survival outcomes at 96 hours with high sensitivity (80%) and specificity (73%) (area under curve of ROC = 0.8, Jmax = 0.52).Data ConclusionThis preclinical study suggests combined T2 relaxation time and perfusion mapping as first line diagnostic tool for treatment planning.Level of Evidence2Technical Efficacy Stage2