Multicenter Evaluation of AI-generated DIR and PSIR for Cortical and Juxtacortical Multiple Sclerosis Lesion Detection

Background Cortical multiple sclerosis lesions are clinically relevant but inconspicuous at conventional clinical MRI. Double inversion recovery (DIR) and phase-sensitive inversion recovery (PSIR) are more sensitive but often unavailable. In the past 2 years, artificial intelligence (AI) was used to generate DIR and PSIR from standard clinical sequences (eg, T1-weighted, T2-weighted, and fluid-attenuated inversion-recovery sequences), but multicenter validation is crucial for further implementation. Purpose To evaluate cortical and juxtacortical multiple sclerosis lesion detection for diagnostic and disease monitoring purposes on AI-generated DIR and PSIR images compared with MRI-acquired DIR and PSIR images in a multicenter setting. Materials and Methods Generative adversarial networks were used to generate AI-based DIR (n = 50) and PSIR (n = 43) images. The number of detected lesions between AI-generated images and MRI-acquired (reference) images was compared by randomized blinded scoring by seven readers (all with >10 years of experience in lesion assessment). Reliability was expressed as the intraclass correlation coefficient (ICC). Differences in lesion subtype were determined using Wilcoxon signed-rank tests. Results MRI scans of 202 patients with multiple sclerosis (mean age, 46 years ± 11 [SD]; 127 women) were retrospectively collected from seven centers (February 2020 to January 2021). In total, 1154 lesions were detected on AI-generated DIR images versus 855 on MRI-acquired DIR images (mean difference per reader, 35.0% ± 22.8; P < .001). On AI-generated PSIR images, 803 lesions were detected versus 814 on MRI-acquired PSIR images (98.9% ± 19.4; P = .87). Reliability was good for both DIR (ICC, 0.81) and PSIR (ICC, 0.75) across centers. Regionally, more juxtacortical lesions were detected on AI-generated DIR images than on MRI-acquired DIR images (495 [42.9%] vs 338 [39.5%]; P < .001). On AI-generated PSIR images, fewer juxtacortical lesions were detected than on MRI-acquired PSIR images (232 [28.9%] vs 282 [34.6%]; P = .02). Conclusion Artificial intelligence-generated double inversion-recovery and phase-sensitive inversion-recovery images performed well compared with their MRI-acquired counterparts and can be considered reliable in a multicenter setting, with good between-reader and between-center interpretative agreement. Published under a CC BY 4.0 license. Supplemental material is available for this article. See also the editorial by Zivadinov and Dwyer in this issue.

P01.10.B The relation between executive functioning and MEG multilayer network centrality in glioma patients

Background

Many patients with glioma suffer from deficits in executive functioning (EF), regardless of tumor location and size. While some patients experience improving EF after tumor resection, others deteriorate over time. This variability in cognitive trajectories is still poorly understood, but neuroscientific approaches that view the brain as a network may bear correlative and predictive value in this respect. Particularly integrative network connectivity of the fronto-parietal network are crucial for EF, but studies so far have only investigated isolated neural frequency bands of interest. Here, we synergize fronto-parietal network connectivity across frequency bands and test its explanatory power for EF at diagnosis and 1 year after tumor resection.

Material and Methods

Patients with diffuse glioma (n=37) underwent neuropsychological assessments, including three tests for EF, and resting-state magnetoencephalography (MEG) at both time points. Patients’ EF performance was standardized to z-scores using validated norm scores, adjusting for age, sex and education. MEG was source-reconstructed using patients’ MRI in combination with a beamformer approach and time series were filtered into six classical frequency bands. Band-specific functional connectivity between 78 cortical regions was estimated, after which an interconnected multilayer, multi-frequency network was created per patient. Multilayer centrality was then calculated as a measure of integrative network connectivity per region, and averaged over all fronto-parietal network regions to yield a single value reflecting integrative, multi-frequency network connectivity per patient at each time point.

Results

At diagnosis, eight patients had z-scores &lt;-1.5, indicating clinically relevant cognitive deficits. Preoperatively, poorer performance on the Concept Shifting Task was associated with lower multilayer centrality (p=.017), while improving performance on this task from diagnosis to 1 year after resection was associated with an increase in multilayer centrality (p=.022). Multilayer centrality at diagnosis did not significantly predict EF at follow-up.

Conclusion

Our results establish a significant relationship between poorer and decreasing EF with lower and decreasing multilayer, multi-frequency integrative connectivity of the fronto-parietal network in glioma patients. Studies with larger and more homogeneous samples may further explore the relevance of this multilayer approach in understanding and particularly predicting postoperative decline in EF in these patients, as our sample was characterized by large heterogeneity in cognitively relevant patient and tumor characteristics.

Topology, symmetry, phase transitions, and noncollinear spin structures

We use a topological approach to describe the frustration- and field-induced phase transitions exhibited by the infinite-range XY model on the AB2 chain, including noncollinear spin structures. For this purpose, we have computed the Morse number and the Euler characteristic, as well as other topological invariants, which are found to behave similarly as a function of the energy level in the context of Morse theory. In particular, we use a method based on an analogy with statistical mechanics to compute the Euler characteristic, which proves to be quite feasible. We also introduce topological energies which help us to clarify several properties of the transitions, both at zero and finite temperatures. In addition, we establish a nontrivial direct connection between the thermodynamics of the systems, which have been solved exactly under the saddle-point approach, and the topology of their configuration space. This connection allows us to identify the nondegeneracy condition under which the divergence of the density of Jacobian's critical points [jl(E)] at the critical energy of a topology-induced phase transition, proposed by Kastner and Schnetz [Phys. Rev. Lett. 100, 160601 (2008)] as a necessary criterion, is suppressed. Finally, our findings and those available in the literature suggest that the cusplike singularity exhibited both by the Euler characteristic and the topological contribution for the entropy at the critical energy, put together with the divergence of jl(E) , and emerge as necessary and sufficient conditions for the occurrence of the finite-temperature topology-induced phase transitions examined in this work. The general character of this proposal should be subject to a more rigorous scrutiny.