Deep learning in neuroimaging of epilepsy

In recent years, artificial intelligence, particularly deep learning (DL), has demonstrated utility in diverse areas of medicine. DL uses neural networks to automatically learn features from the raw data while this is not possible with conventional machine learning. It is helpful for the assessment of patients with epilepsy and whilst most published studies have been aimed at the automatic detection and prediction of seizures from electroencephalographic records, there is a growing number of investigations that use neuroimaging modalities (structural and functional magnetic resonance imaging, diffusion-weighted imaging and positron emission tomography) as input data. We review the application of DL to neuroimaging (sMRI, fMRI, DWI and PET) of focal epilepsy, specifically presurgical evaluation of drug-refractory epilepsy. First, a brief theoretical overview of artificial neural networks and deep learning is presented. Next, we review applications of deep learning to neuroimaging of epilepsy: diagnosis and lateralization, automated detection of lesion, presurgical evaluation and prediction of postsurgical outcome. Finally, the limitations, challenges and possible future directions in the application of these methods in the study of epilepsies are discussed. This approach could become an essential tool in clinical practice, particularly in the evaluation of images considered negative by visual inspection, in individualized treatments, and in the approach to epilepsy as a network disorder. However, greater multicenter collaboration is required to achieve the collection of sufficient data with the required quality together with the open access availability of the developed codes and tools.

Effects of interventions on cerebral perfusion in the Alzheimer's disease spectrum: A systematic review

Cerebral perfusion dysfunctions are seen in the early stages of Alzheimer's disease (AD). We systematically reviewed the literature to investigate the effect of pharmacological and non-pharmacological interventions on cerebral hemodynamics in randomized controlled trials involving AD patients or Mild Cognitive Impairment (MCI) due to AD. Studies involving other dementia types were excluded. Data was searched in April 2021 on MEDLINE, Embase, and Web of Science. Risk of bias was assessed using Cochrane Risk of Bias Tool. A meta-synthesis was performed separating results from MCI and AD studies. 31 studies were included and involved 310 MCI and 792 CE patients. The MCI studies (n = 8) included physical, cognitive, dietary, and pharmacological interventions. The AD studies (n = 23) included pharmacological, physical interventions, and phytotherapy. Cerebral perfusion was assessed with PET, ASL, Doppler, fNIRS, DSC-MRI, Xe-CT, and SPECT. Randomization and allocation concealment methods and subject characteristics such as AD-onset, education, and ethnicity were missing in several papers. Positive effects on hemodynamics were seen in 75 % of the MCI studies, and 52 % of the AD studies. Inserting cerebral perfusion outcome measures, together with established AD biomarkers, is fundamental to target all disease mechanisms and understand the role of cerebral perfusion in AD.

The effect of lesion filling on brain network analysis in multiple sclerosis using structural magnetic resonance imaging

Background

Graph theoretical network analysis with structural magnetic resonance imaging (MRI) of multiple sclerosis (MS) patients can be used to assess subtle changes in brain networks. However, the presence of multiple focal brain lesions might impair the accuracy of automatic tissue segmentation methods, and hamper the performance of graph theoretical network analysis. Applying “lesion filling” by substituting the voxel intensities of a lesion with the voxel intensities of nearby voxels, thus creating an image devoid of lesions, might improve segmentation and graph theoretical network analysis. This study aims to determine if brain networks are different between MS subtypes and healthy controls (HC) and if the assessment of these differences is affected by lesion filling.

Methods

The study included 49 MS patients and 19 HC that underwent a T1w, and T2w-FLAIR MRI scan. Graph theoretical network analysis was performed from grey matter fractions extracted from the original T1w-images and T1w-images after lesion filling.

Results

Artefacts in lesion-filled T1w images correlated positively with total lesion volume (r = 0.84, p < 0.001) and had a major impact on grey matter segmentation accuracy. Differences in sensitivity for network alterations were observed between original T1w data and after application of lesion filling: graph theoretical network analysis obtained from lesion-filled T1w images produced more differences in network organization in MS patients.

Conclusion

Lesion filling might reduce variability across subjects resulting in an increased detection rate of network alterations in MS, but also induces significant artefacts, and therefore should be applied cautiously especially in individuals with higher lesions loads.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13244-022-01198-4.

Cerebral topography of vesicular cholinergic transporter changes in neurologically intact adults: A [18F]FEOBV PET study

Acetylcholine plays a major role in brain cognitive and motor functions with regional cholinergic terminal loss common in several neurodegenerative disorders. We describe age-related declines of regional cholinergic neuron terminal density in vivo using the positron emission tomography (PET) ligand [¹⁸F](–)5-Fluoroethoxybenzovesamicol ([¹⁸F] FEOBV), a vesamicol analogue selectively binding to the vesicular acetylcholine transporter (VAChT). A total of 42 subjects without clinical evidence of neurologic disease (mean 50.55 [range 20–80] years, 24 Male/18 Female) underwent [¹⁸F]FEOBV brain PET imaging. We used SPM based voxel-wise statistical analysis to perform whole brain voxel-based parametric analysis (family-wise error corrected, FWE) and to also extract the most significant clusters of regions correlating with aging with gender as nuisance variable. Age-related VAChT binding reductions were found in primary sensorimotor cortex, visual cortex, caudate nucleus, anterior to mid-cingulum, bilateral insula, para-hippocampus, hippocampus, anterior temporal lobes/amygdala, dorsomedial thalamus, metathalamus, and cerebellum (gender and FWE-corrected, P < 0.05). These findings show a specific topographic pattern of regional vulnerability of cholinergic nerve terminals across multiple cholinergic systems accompanying aging.

Regional cerebral cholinergic nerve terminal integrity and cardinal motor features in Parkinson's disease

Clinical effects of anti-cholinergic drugs implicate cholinergic systems alterations in the pathophysiology of some cardinal motor impairments in Parkinson’s disease. The topography of affected cholinergic systems deficits and motor domain specificity are poorly understood. Parkinson's disease patients (n = 108) underwent clinical and motor assessment and vesicular acetylcholine transporter [¹⁸F]-fluoroethoxybenzovesamicol PET imaging. Volumes-of-interest-based analyses included detailed thalamic and cerebellar parcellations. Successful PET sampling for most of the small-sized parcellations was available in 88 patients. A data-driven approach, stepwise regression using the forward selection method, was used to identify cholinergic brain regions associating with cardinal domain-specific motor ratings. Regressions with motor domain scores for model-selected regions followed by confounder analysis for effects of age of onset, duration of motor disease and levodopa equivalent dose were performed. Among 7 model-derived regions associating with postural instability and gait difficulties domain scores three retained significance in confounder variable analysis: medial geniculate nucleus (standardized β = −0.34, t = −3.78, P = 0.0003), lateral geniculate nucleus (β = −0.32, t = −3.4, P = 0.001) and entorhinal cortex (β = −0.23, t = −2.6, P = 0.011). A sub-analysis of non-episodic postural instability and gait difficulties scores demonstrated significant effects of the medial geniculate nucleus, entorhinal cortex and globus pallidus pars interna. Among 6 tremor domain model-selected regions two regions retained significance in confounder variable analysis: cerebellar vermis section of lobule VIIIb (β = −0.22, t = −2.4, P = 0.021) and the putamen (β = −0.23, t = −2.3, P = 0.024). None of the three model-selected variables for the rigidity domain survived confounder analysis. Two out of the four model-selected regions for the distal limb bradykinesia domain survived confounder analysis: globus pallidus pars externa (β = 0.36, t = 3.9, P = 0.0097) and the paracentral lobule (β = 0.26, t = 2.5, P = 0.013). Emphasizing the utility of a systems-network conception of the pathophysiology of Parkinson's disease cardinal motor features, our results are consistent with specific deficits in basal forebrain corticopetal, peduncupontine-laterodorsal tegmental complex, and medial vestibular nucleus cholinergic pathways, against the background of nigrostriatal dopaminergic deficits, contributing significantly to postural instability, gait difficulties, tremor and distal limb bradykinesia cardinal motor features of Parkinson’s disease. Our results suggest significant and distinct consequences of degeneration of cholinergic peduncupontine-laterodorsal tegmental complex afferents to both segments of the globus pallidus. Non-specific regional cholinergic nerve terminal associations with rigidity scores likely reflect more complex multifactorial signalling mechanisms with smaller contributions from cholinergic pathways.

Dopaminergic Nigrostriatal Connectivity in Early Parkinson Disease: In Vivo Neuroimaging Study of 11C-DTBZ PET Combined with Correlational Tractography

Previous histopathologic and animal studies have shown axonal impairment and loss of connectivity of the nigrostriatal pathway in Parkinson disease (PD). However, there are conflicting reports from in vivo human studies. ¹¹C-dihydrotetrabenazine (¹¹C-DTBZ) is a vesicular monoamine type 2 transporter PET ligand that allows assessment of nigrostriatal presynaptic dopaminergic terminal integrity. Correlational tractography based on diffusion MRI can incorporate ligand-specific information provided by ¹¹C-DTBZ PET into the fiber-tracking process. The purpose of this study was to assess the in vivo association between the integrity of the nigrostriatal tract (defined by correlational tractography) and the degree of striatal dopaminergic denervation based on ¹¹C-DTBZ PET. Methods: The study involved 30 subjects with mild to moderate PD (23 men and 7 women; mean age, 66 ± 6.2 y; disease duration, 6.4 ± 4.0 y; Hoehn and Yahr stage, 2.1 ± 0.6; Movement Disorder Society [MDS]-revised Unified Parkinson Disease Rating Scale [UPDRS] [I-III] total score, 43.4 ± 17.8) and 30 control subjects (18 men and 12 women; mean age, 62 ± 10.3 y). ¹¹C-DTBZ PET was performed using standard synthesis and acquisition protocols. Correlational tractography was performed to assess quantitative anisotropy (QA; a measure of tract integrity) of white matter fibers correlating with information derived from striatal ¹¹C-DTBZ data using the DSI Studio toolbox. Scans were realigned according to least and most clinically affected cerebral hemispheres. Results: Nigrostriatal tracts were identified in both hemispheres of PD patients. Higher mean QA values along the identified tracts were significantly associated with higher striatal ¹¹C-DTBZ distribution volume ratios (least affected: r = 0.57, P = 0.001; most affected: r = 0.44, P = 0.02). Lower mean QA values of the identified tract in the LA hemisphere associated with increased severity of bradykinesia sub-score derived from MDS-UPDRS part III (r = -0.42; P = 0.02). Cross-validation revealed the generalizability of these results. Conclusion: These findings suggest that impaired integrity of dopaminergic nigrostriatal nerve terminals is associated with nigrostriatal axonal dysfunction in mild to moderate PD. Assessment of nigrostriatal tract integrity may be suitable as a biomarker of early- or even prodromal-stage PD.

Machine learning in the integration of simple variables for identifying patients with myocardial ischemia

BACKGROUND

A significant number of variables are obtained when characterizing patients suspected with myocardial ischemia or at risk of MACE. Guidelines typically use a handful of them to support further workup or therapeutic decisions. However, it is likely that the numerous available predictors maintain intrinsic complex interrelations. Machine learning (ML) offers the possibility to elucidate complex patterns within data to optimize individual patient classification. We evaluated the feasibility and performance of ML in utilizing simple accessible clinical and functional variables for the identification of patients with ischemia or an elevated risk of MACE as determined through quantitative PET myocardial perfusion reserve (MPR).

METHODS

1,234 patients referred to Nitrogen-13 ammonia PET were analyzed. Demographic (4), clinical (8), and functional variables (9) were retrieved and input into a cross-validated ML workflow consisting of feature selection and modeling. Two PET-defined outcome variables were operationalized: (1) any myocardial ischemia (regional MPR < 2.0) and (2) an elevated risk of MACE (global MPR < 2.0). ROC curves were used to evaluate ML performance.

RESULTS

16 features were included for boosted ensemble ML. ML achieved an AUC of 0.72 and 0.71 in identifying patients with myocardial ischemia and with an elevated risk of MACE, respectively. ML performance was superior to logistic regression when the latter used the ESC guidelines risk models variables for both PET-defined labels (P < .001 and P = .01, respectively).

CONCLUSIONS

ML is feasible and applicable in the evaluation and utilization of simple and accessible predictors for the identification of patients who will present myocardial ischemia and an elevated risk of MACE in quantitative PET imaging.

Topography of Cholinergic Changes in Dementia With Lewy Bodies and Key Neural Network Hubs

OBJECTIVES

The authors investigated the topography of cholinergic vulnerability in patients with dementia with Lewy bodies (DLB) using positron emission tomography (PET) imaging with the vesicular acetylcholine transporter (VAChT) [¹⁸F]-fluoroethoxybenzovesamicol ([¹⁸F]-FEOBV) radioligand.

METHODS

Five elderly participants with DLB (mean age, 77.8 years [SD=4.2]) and 21 elderly healthy control subjects (mean age, 73.62 years [SD=8.37]) underwent clinical assessment and [¹⁸F]-FEOBV PET.

RESULTS

Compared with the healthy control group, reduced VAChT binding in patients with DLB demonstrated nondiffuse regionally distinct and prominent reductions in bilateral opercula and anterior cingulate to mid-cingulate cortices, bilateral insula, right (more than left) lateral geniculate nuclei, pulvinar, right proximal optic radiation, bilateral anterior and superior thalami, and posterior hippocampal fimbria and fornices.

CONCLUSIONS

The topography of cholinergic vulnerability in DLB comprises key neural hubs involved in tonic alertness (cingulo-opercular), saliency (insula), visual attention (visual thalamus), and spatial navigation (fimbria/fornix) networks. The distinct denervation pattern suggests an important cholinergic role in specific clinical disease-defining features, such as cognitive fluctuations, visuoperceptual abnormalities causing visual hallucinations, visuospatial changes, and loss of balance caused by DLB.

Multimodal imaging in nonlesional medically intractable focal epilepsy

Identification and localization of epileptogenic zone (EZ) is vital in patients with medically-intractable focal epilepsy, who may be candidates for potentially curative resective epilepsy surgery. Presence of a lesion on magnetic resonance imaging (MRI) influences both diagnostic classification and selection for surgery. However, the implications for MRI-negative cases are not well-defined for such patients. Most of these patients undergo invasive long-term Electroencephalography recordings before a final decision regarding resection is possible. Recent developments in structural and functional neuroimaging which include quali-quantitative MRI, Positron Emission Tomography, Single Photon Emission Computed Tomography, and functional MRI have significantly changed presurgical epilepsy evaluation. Source analysis based on electrophysiological information, using either EEG or magnetoencephalography are also promising in order to noninvasively localize the EZ and to guide surgery in medically-intractable focal epilepsy patients that exhibit nonlesional MRI. This chapter aims to review the value of the combined use of structural and functional imaging techniques, and how this multimodal approach improves both selection of surgical candidates and post-operative outcomes in medically-intractable nonlesional focal epilepsy.

Cognitive changes after stem cell transplantation in a patient with subcortical stroke

The authors report a case of a 55-year-old Caucasian woman who received autologous bone marrow stem cell transplantation 3 years after a subcortical stroke. She exhibited positive cognitive changes 6 months and 1 year after the surgery without rehabilitation. The blood flow changes, measured with SPECT, were statistical significant in prefrontal areas. During the presurgical neuropsychological assessment, the patient presented a critical speech reduction, reflected in impaired performance in verbal fluency, vocabulary and in each task which required overt verbal response. One year later, she showed improvement in mental flexibility, receptive language, phonological fluency, verbal memory and auditory verbal memory. Positive cognitive changes in verbal and executive functions seem to be contingent on increased blood flow in prefrontal areas. Posterior neuropsychological evaluation 3 and 5 years after transplantation did not show deterioration of the cognitive improvement.