Development and prospective clinical validation of a convolutional neural network for automated detection and segmentation of focal cortical dysplasias.

PURPOSE: Focal cortical dysplasias (FCDs) are a leading cause of drug-resistant epilepsy. Early detection and resection of FCDs have favorable prognostic implications for postoperative seizure freedom. Despite advancements in imaging methods, FCD detection remains challenging. House et al. (2021) introduced a convolutional neural network (CNN) for automated FCD detection and segmentation, achieving a sensitivity of 77.8%. However, its clinical applicability was limited due to a low specificity of 5.5%. The objective of this study was to improve the CNN's performance through data-driven training and algorithm optimization, followed by a prospective validation on daily-routine MRIs. MATERIAL AND METHODS: A dataset of 300 3 T MRIs from daily clinical practice, including 3D T1 and FLAIR sequences, was prospectively compiled. The MRIs were visually evaluated by two neuroradiologists and underwent morphometric assessment by two epileptologists. The dataset included 30 FCD cases (11 female, mean age: 28.1 ± 10.1 years) and a control group of 150 normal cases (97 female, mean age: 32.8 ± 14.9 years), along with 120 non-FCD pathological cases (64 female, mean age: 38.4 ± 18.4 years). The dataset was divided into three subsets, each analyzed by the CNN. Subsequently, the CNN underwent a two-phase-training process, incorporating subset MRIs and expert-labeled FCD maps. This training employed both classical and continual learning techniques. The CNN's performance was validated by comparing the baseline model with the trained models at two training levels. RESULTS: In prospective validation, the best model trained using continual learning achieved a sensitivity of 90.0%, specificity of 70.0%, and accuracy of 72.0%, with an average of 0.41 false positive clusters detected per MRI. For FCD segmentation, an average Dice coefficient of 0.56 was attained. The model's performance improved in each training phase while maintaining a high level of sensitivity. Continual learning outperformed classical learning in this regard. CONCLUSIONS: Our study presents a promising CNN for FCD detection and segmentation, exhibiting both high sensitivity and specificity. Furthermore, the model demonstrates continuous improvement with the inclusion of more clinical MRI data. We consider our CNN a valuable tool for automated, examiner-independent FCD detection in daily clinical practice, potentially addressing the underutilization of epilepsy surgery in drug-resistant focal epilepsy and thereby improving patient outcomes.

Automated detection and segmentation of focal cortical dysplasias (FCDs) with artificial intelligence: Presentation of a novel convolutional neural network and its prospective clinical validation.

PURPOSE: Focal cortical dysplasias (FCDs) represent one of the most frequent causes of pharmaco-resistant focal epilepsies. Despite improved clinical imaging methods over the past years, FCD detection remains challenging, as FCDs vary in location, size, and shape and commonly blend into surrounding tissues without clear definable boundaries. We developed a novel convolutional neural network for FCD detection and segmentation and validated it prospectively on daily-routine MRIs. MATERIAL AND METHODS: The neural network was trained on 201 T1 and FLAIR 3 T MRI volume sequences of 158 patients with mainly FCDs, regardless of type, and 7 focal PMG. Non-FCD/PMG MRIs, drawn from 100 normal MRIs and 50 MRIs with non-FCD/PMG pathologies, were added to the training. We applied the algorithm prospectively on 100 consecutive MRIs of patients with focal epilepsy from daily clinical practice. The results were compared with corresponding neuroradiological reports and morphometric MRI analyses evaluated by an experienced epileptologist. RESULTS: Best training results reached a sensitivity (recall) of 70.1 % and a precision of 54.3 % for detecting FCDs. Applied on the daily-routine MRIs, 7 out of 9 FCDs were detected and segmented correctly with a sensitivity of 77.8 % and a specificity of 5.5 %. The results of conventional visual analyses were 33.3 % and 94.5 %, respectively (3/9 FCDs detected); the results of morphometric analyses with overall epileptologic evaluation were both 100 % (9/9 FCDs detected) and thus served as reference. CONCLUSION: We developed a 3D convolutional neural network with autoencoder regularization for FCD detection and segmentation. Our algorithm employs the largest FCD training dataset to date with various types of FCDs and some focal PMG. It provided a higher sensitivity in detecting FCDs than conventional visual analyses. Despite its low specificity, the number of false positively predicted lesions per MRI was lower than with morphometric analysis. We consider our algorithm already useful for FCD pre-screening in everyday clinical practice.

Comparison of acetylsalicylic acid and clopidogrel non-responsiveness assessed by light transmittance aggregometry and PFA-100® in patients undergoing neuroendovascular procedures.

Objectives: Dual platelet inhibition is commonly used for prevention of cardiovascular events in patients undergoing neuroendovascular procedures. Non-responsiveness to platelet inhibitors may be associated with adverse outcomes. The aim of this study was to evaluate the reliability of the platelet function analyzer PFA-100® in comparison to light transmittance aggregometry (LTA) for monitoring clopidogrel and acetylsalicylic acid (ASA) non-responsiveness in a cohort of patients treated for intracranial aneurysm or cranial artery stenosis. Methods: Non-responsiveness to clopidogrel and ASA was assessed by LTA using adenosine diphosphate (ADP) and arachidonic acid and by PFA-100® with the ADP/prostaglandin E1 (PGE1) and collagen/epinephrine cartridges, respectively. Results: A total of 203 patients (145 females; median age, 57 years) were analyzed. Agreement between the two tests was poor for clopidogrel non-responsiveness (ƙ=0.19) and not better than chance for ASA non-responsiveness (ƙ=0.01). Clopidogrel non-responsiveness by LTA and PFA-100® was associated with higher von Willebrand factor antigen and activity levels. ADP-induced platelet disaggregation was lower in patients with clopidogrel non-responsiveness as assessed by PFA-100®. Clopidogrel non-responsiveness by LTA was associated with a higher prevalence of diabetes and a higher body mass index (BMI). Adverse outcomes (death, thromboembolism, or in-stent thrombosis) occurred in 13% (n=26) of all patients independently of ASA and clopidogrel non-responsiveness as assessed by both devices. Conclusions: Our results show that LTA and PFA-100® are not interchangeable in the assessment of ASA and clopidogrel non-responsiveness in patients undergoing neuroendovascular interventions.

Impact of focal cortical dysplasia Type IIIa on seizure outcome following anterior mesial temporal lobe resection for the treatment of epilepsy.

OBJECTIVE Temporal lobe epilepsy (TLE) is the most common type of pharmacoresistant focal epilepsy, for which anterior mesial temporal lobe resection (AMTLR) is a treatment option. Focal cortical dysplasia Type IIIa (FCD IIIa), a developmental lesion resulting from defects in neuronal formation and migration into the temporal pole (FCD I) combined with hippocampal sclerosis (HS), can be a neuropathological finding. In this study, the authors investigate the impact of FCD IIIa on seizure outcome in patients with TLE who underwent AMTLR. METHODS The authors performed a retrospective analysis of all patients with TLE who underwent AMTLR at their institution between June 2011 and April 2014. Histopathological analysis was used to determine whether patients had HS together with FCD I (FCD IIIa) or HS alone. The groups were compared with regard to age, sex, years of epilepsy, and seizure outcome using the Engel classification. RESULTS A total of 51 patients with TLE underwent AMTLR at the authors' institution. FCD IIIa was diagnosed in 13 cases. The patients experienced seizures for a mean duration of 31.1 years. The mean length of follow-up after the procedure was 18 months. All patients with FCD IIIa had a favorable seizure outcome (Engel Class I or II) compared with 71% of the patients with no pathological findings in the temporal pole (p < 0.01). CONCLUSIONS Patients with histopathologically proven FCD IIIa had a significantly better seizure outcome after AMTLR than patients with HS alone. Further effort should be made during presurgical evaluation to detect FCD IIIa so that the most suitable resection technique can be chosen and postoperative seizure outcome can be predicted for patient counseling.

T1- Thresholds in Black Holes Increase Clinical-Radiological Correlation in Multiple Sclerosis Patients.

BACKGROUND: Magnetic Resonance Imaging (MRI) is an established tool in diagnosing and evaluating disease activity in Multiple Sclerosis (MS). While clinical-radiological correlations are limited in general, hypointense T1 lesions (also known as Black Holes (BH)) have shown some promising results. The definition of BHs is very heterogeneous and depends on subjective visual evaluation. OBJECTIVE: We aimed to improve clinical-radiological correlations by defining BHs using T1 relaxation time (T1-RT) thresholds to achieve best possible correlation between BH lesion volume and clinical disability. METHOD: 40 patients with mainly relapsing-remitting MS underwent MRI including 3-dimensional fluid attenuated inversion recovery (FLAIR), magnetization-prepared rapid gradient echo (MPRAGE) before and after Gadolinium (GD) injection and double inversion-contrast magnetization-prepared rapid gradient echo (MP2RAGE) sequences. BHs (BHvis) were marked by two raters on native T1-weighted (T1w)-MPRAGE, contrast-enhancing lesions (CE lesions) on T1w-MPRAGE after GD and FLAIR lesions (total-FLAIR lesions) were detected separately. BHvis and total-FLAIR lesion maps were registered to MP2RAGE images, and the mean T1-RT were calculated for all lesion ROIs. Mean T1 values of the cortex (CTX) were calculated for each patient. Subsequently, Spearman rank correlations between clinical scores (Expanded Disability Status Scale and Multiple Sclerosis Functional Composite) and lesion volume were determined for different T1-RT thresholds. RESULTS: Significant differences in T1-RT were obtained between all different lesion types with highest T1 values in visually marked BHs (BHvis: 1453.3±213.4 ms, total-FLAIR lesions: 1394.33±187.38 ms, CTX: 1305.6±35.8 ms; p<0.05). Significant correlations between BHvis/total-FLAIR lesion volume and clinical disability were obtained for a wide range of T1-RT thresholds. The highest correlation for BHvis and total-FLAIR lesion masks were found at T1-RT>1500 ms (Expanded Disability Status Scale vs. lesion volume: rBHvis = 0.442 and rtotal-FLAIR = 0.497, p<0.05; Multiple Sclerosis Functional Composite vs. lesion volume: rBHvis = -0.53 and rtotal-FLAIR = -0.627, p<0.05). CONCLUSION: Clinical-radiological correlations in MS patients are increased by application of T1-RT thresholds. With the short acquisition time of the MP2RAGE sequences, quantitative T1 maps could be easily established in clinical studies.

Morphometric MRI analysis enhances visualization of cortical tubers in tuberous sclerosis.

PURPOSE: Focal cortical dysplasias (FCD) type IIb and cortical tubers in tuberous sclerosis complex (TSC) are histopathologically similar and are both epileptogenic lesions frequently causing pharmacoresistant epilepsies. Morphometric analysis of T1- and T2-weighted MRI volume data sets can enhance visualization of FCD. Here, we retrospectively investigated whether morphometric MRI analysis is of equal benefit for visualizing cortical tubers. MATERIALS AND METHODS: Morphometric analysis was applied to T1- and partly also T2-weighted 1.5T or 3T MRI volume data sets of 15 TSC patients using a fully automated MATLAB(®) script (i.e. MAP07) commonly used for FCD detection. In this study, focus was on the most sensitive of the resulting morphometric feature maps (i.e. the 'junction image') which highlights blurring of the gray-white matter junction in comparison to a normal database. The visualization of tubers in these 'junction images' was quantitatively compared with that in conventional MR sequences. RESULTS: In all patients, morphometric analysis visualized almost all tubers detected in the normal MRI, and additionally highlighted on average 23% (range 3-50%) more tubers which were not detected by visual analysis of the conventional MR sequences. When T2 volume data sets from a 3T scanner were available for postprocessing, the rate of additionally detected tubers increased to 29% on average. These formerly overlooked tubers were usually smaller than the tubers already found in the conventional MRI. CONCLUSION: Morphometric analysis of MRIs in TSC can highlight cortical tubers which are likely to be overlooked in conventional MRI sequences alone. Additionally detected tubers may be of potential importance for both presurgical evaluation and initial diagnosis of TSC.

Unexplained loss of vision in a child: consider bilateral primary optic nerve sheath meningioma.

A 4-year-old girl gradually lost her vision to become practically blind at the age of 10 years. Examinations at several medical centers had been unable to establish an etiology. Traditional investigation using cerebral magnetic resonance imaging (MRI) initially showed normal results; however, later on it showed progressive atrophy of both optical nerves without recognizable cause. Subsequently, MRI including adequate orbital sequences, contrast-enhanced sequences, and fat suppression demonstrated bilateral primary optic nerve sheath meningioma, a rare but treatable tumor of childhood. The patient underwent neurosurgery and to date retains minimal vision. Adequate neuroradiological investigation of unexplained optic atrophy is advocated.

Vascular events after transsylvian selective amygdalohippocampectomy and impact on epilepsy outcome.

OBJECTIVE: Epilepsy surgery is a standard treatment option for medically intractable temporal lobe epilepsy. Selective amygdalohippocampectomy (SAH) and anterior temporal lobectomy (ATL) are two of the standard surgical procedures in these cases. We conducted a retrospective analysis of patients treated with SAH via a modified transsylvian approach in our epilepsy center between 2008 and 2011, and we analyzed the impact of adjacent procedure-related infarctions on seizure outcome in these patients. METHODS: Infarctions were detected by magnetic resonance imaging (MRI) within the first week postoperatively and by a second MRI 9 months after surgical intervention. Neuropsychological testing was performed preoperatively. Evaluation of seizure outcome and postoperative neuropsychological testing were conducted approximately 1 year after epilepsy surgery. Correlative clinical data were analyzed by retrospective chart review. RESULTS: The postoperative MRI revealed temporal infarctions in 47.9% (n = 23/48) and frontal infarctions in 10.4% (n = 5/48) of the patients. These vascular events were asymptomatic in terms of focal neurologic deficits. Of the patients, 68.5% (n = 37/54) were free of disabling seizures (Engel class I) 1 year after the procedure. Patients with temporal infarctions were significantly more often free of disabling seizures (Engel class I, p = 0.046) than patients without temporal infarctions. Neuropsychological testing indicated a deterioration in verbal memory after SAH in patients with infarctions on the language-lateralized hemisphere compared to patients without infarction (p = 0.011). All other tested neuropsychological categories showed no significant differences between patients with or without infarctions. SIGNIFICANCE: Our results indicate a surprisingly high number of procedure-related temporal infarctions after transsylvian SAH. Hence, the volume of nonfunctional "eliminated" tissue is enlarged unintentionally, which is a possible explanation for better seizure outcome in these patients. This result supports the notion that ATL is the favorable procedure for temporal lobe epilepsy compared to SAH in the nondominant hemisphere, as neuropsychological deficits are rarely to be expected.

Comparison of morphometric analysis based on T1- and T2-weighted MRI data for visualization of focal cortical dysplasia.

PURPOSE: Focal cortical dysplasias (FCD) are highly epileptogenic lesions frequently accounting for pharmaco-resistant focal epilepsy. Visual MRI analysis combined with morphometric analysis of T1-weighted MRI data was shown to be of higher diagnostic sensitivity in detecting and delineating FCD than conventional visual analysis alone. Here we investigate whether morphometric analysis of T2-weighted MRI volume data sets is of equal benefit or perhaps more helpful for visualizing FCD. MATERIALS AND METHODS: Morphometric analysis was applied to T1- and T2-weighted MRI volume data sets of 20 epilepsy patients with FCD using a fully automated MATLAB script with scanner- and sequence-specific normal databases for T1 and T2 images. For each modality, a new feature map (i.e., 'junction image') highlighting the FCD-typical blurring of the gray-white matter junction and quantifying this feature in comparison to the normal database in terms of z-scores was calculated. The resulting T1 and T2 'junction images' were compared for conspicuity and recognizability of the FCD both qualitatively by visual assessment and quantitatively by analysis of the mean z-scores inside and outside the lesions. RESULTS: In 80% of the cases, the FCD presented with higher contrast and/or clearer delineation in the T2 than in the T1 'junction images' and were thus easier to recognize in these images. The quantitative analysis supported this impression: in 95% of cases, the ratio of mean z-scores inside and outside the FCD was higher in T2- than in T1-based 'junction images'. For the T2 'junction images', this ratio amounted to 8.7 on average and was thus more than twice as high as the corresponding T1 result of 3.7 (p<.003). CONCLUSION: Concerning visualization of FCD by highlighting blurring of the gray-white matter junction, the results of the present study indicate that morphometric analysis of T2-weighted MRI data on average is superior to T1-based morphometry.

MRI correlates of disability in African-Americans with multiple sclerosis.

OBJECTIVES: Multiple sclerosis (MS) in African-Americans (AAs) is characterized by more rapid disease progression and poorer response to treatment than in Caucasian-Americans (CAs). MRI provides useful and non-invasive tools to investigate the pathological substrate of clinical progression. The aim of our study was to compare MRI measures of brain damage between AAs and CAs with MS. METHODS: Retrospective analysis of 97 AAs and 97 CAs with MS matched for age, gender, disease duration and age at MRI examination. RESULTS: AA patients had significantly greater T2- (p = 0.001) and T1-weighted (p = 0.0003) lesion volumes compared to CA patients. In contrast, measurements of global and regional brain volume did not significantly differ between the two ethnic groups (p>0.1). CONCLUSIONS: By studying a quite large sample of well demographically and clinically matched CA and AA patients with a homogeneous MRI protocol we showed that higher lesion accumulation, rather than pronounced brain volume decrease might explain the early progress to ambulatory assistance of AAs with MS.

Brain iron deposition fingerprints in Parkinson's disease and progressive supranuclear palsy.

It can be difficult to clinically distinguish between classical Parkinson's disease (PD) and progressive supranuclear palsy. Previously, there have been no biomarkers that reliably allow this distinction to be made. We report that an abnormal brain iron accumulation is a marker for ongoing neurodegeneration in both conditions, but the conditions differ with respect to the anatomical distribution of these accumulations. We analyzed quantitative T2' maps as markers of regional brain iron content from PD and progressive supranuclear palsy patients and compared them to age-matched control subjects. T2-weighted and T2*-weighted images were acquired in 30 PD patients, 12 progressive supranuclear palsy patients, and 24 control subjects at 1.5 Tesla. Mean T2' values were determined in regions-of-interest in the basal ganglia, thalamus, and white matter within each hemisphere. The main findings were shortened T2' values in the caudate nucleus, globus pallidus, and putamen in progressive supranuclear palsy compared to PD patients and controls. A stepwise linear discriminant analysis allowed progressive supranuclear palsy patients to be distinguished from PD patients and the healthy controls. All progressive supranuclear palsy patients were correctly classified. No progressive supranuclear palsy patient was classified as a healthy control, no healthy controls were incorrectly classified as having progressive supranuclear palsy, and only 6.7% of the PD patients were incorrectly classified as progressive supranuclear palsy. Regional decreases of T2' relaxation times in parts of the basal ganglia reflecting increased brain iron load in these areas are characteristic for progressive supranuclear palsy but not PD patients.

Image-based classification of parkinsonian syndromes using T2'-atlases.

Parkinsonian syndromes (PS) are genetically and pathologically heterogeneous neurodegenerative disorders. Clinical distinction between different PS can be difficult, particularly in early disease stages. This paper describes an automatic method for the distinction between classical Parkinson's disease (PD) and progressive supranuclear palsy (PSP) using T2' atlases. This procedure is based on the assumption that regional brain iron content differs between PD and PSP, which can be selectively measured using T2' MR imaging. The proposed method was developed and validated based on 33 PD patients, 10 PSP patients, and 24 healthy controls. The first step of the proposed procedure comprises T2' atlas generation for each group using affine and following non-linear registration. For classification, a T2' dataset is registered to the atlases and compared to each one of them using the mean sum of squared differences metric. The dataset is assigned to the group for which the corresponding atlas yields the lowest value. The evaluation using leave-one-out validation revealed that the proposed method achieves a classification accuracy of 91%. The presented method might serve as the basis for an improved automatic classification of PS in the future.

Quantitative t2 values predict time from symptom onset in acute stroke patients.

BACKGROUND AND PURPOSE: We hypothesize that in comparison to diffusion-weighted imaging, quantitative T2 values (qT2) are more directly related to water uptake in ischemic tissue, depending on time from symptom onset. We measured the increase of qT2 in the infarct core to quantify the correlation between time from symptom onset and change in qT2. METHODS: Thirty-six patients with acute ischemic stroke in the territory of the proximal middle cerebral artery underwent MRI including diffusion-weighted imaging, fluid-attenuated inversion recovery, and a triple-echo T2 sequence (calculation of T2 maps) within 6 hours after symptom onset. Regions of decreased apparent diffusion coefficient <550 x 10(-9) mm(2)/sec were defined and superimposed onto the corresponding T2 map and the unaffected side in the horizontally flipped maps. Differences of T2/apparent diffusion coefficient values between affected and unaffected side were calculated (differences of T2/differences of apparent diffusion coefficient). Fluid-attenuated inversion recovery images were rated for lesion visibility. RESULTS: Differences of T2 showed a significant correlation with time from symptom onset (R=0.580; P<0.001). T2 values measured in patients with visible fluid-attenuated inversion recovery lesions were significantly higher than in those without visible hyperintensity (P<0.001). The accuracy of qT2 to predict a time from symptom onset <3 hours was 0.794, whereas the corresponding accuracy for visual assessment of fluid-attenuated inversion recovery images was 0.676. CONCLUSIONS: T2 values demonstrated a strong correlation with time from onset, suggesting different pathophysiologic mechanisms than diffusion restriction. Whereas fluid-attenuated inversion recovery only provides binary information on lesion visibility, T2 values correlate well with time from symptom onset, and are free from operator bias, increasing reproducibility to determine time from symptom onset.