Strojové učení zaznamenalo v posledních letech pozoruhodný pokrok od téměř nepoužitelného rozpoznávání řeči a obrazu k nadlidské přesnosti. Od programů, které nedokázaly porazit jen trochu zkušenějšího hráče go, jsme dospěli k přemožiteli mistra světa. Za pokrokem ve vývoji učících se programů stojí tzv. hluboké učení - deep learning.; Strojové učení zaznamenalo v posledních letech pozoruhodný pokrok od téměř nepoužitelného rozpoznávání řeči a obrazu k nadlidské přesnosti. Od programů, které nedokázaly porazit jen trochu zkušenějšího hráče go, jsme dospěli k přemožiteli mistra světa. Za pokrokem ve vývoji učících se programů stojí tzv. hluboké učení – deep learning.

Cíl práce: Prokázat u nemocných po první epizodě schizofrenie strukturální změny bílé hmoty mozkové pomod fixel-based analýzy (FBA). Metodika: V prospektivnf studii byly pomod FBA studovány difuzně vážené MR obrazy skupiny pacientů po první epizodě schizofrenie (n = 16) a kontrolní skupiny (n = 22). Vyšetření proběhlo na 3T MR tomografu s využitím 64-kanálové hlavové dvky. FBA byla provedena pomocí softwarového balíku MRtrix (verze 3.0.1). Výsledky: Pomod FBA byl u nemocných prokázán pokles metriky fibre density v oblasti commissura anterior. Nebyly nalezeny statisticky významné oblasti se změnami metriky fibre cross-section. Pro kombinovanou metriku fibre density and cross-section byl nalezen jeden statisticky významný fixel v bílé hmotě frontální části pravostranného centrum semiovale. Závěr: Pomocí FBA byl prokázán u nemocných po první epizodě schizofrenie při srovnání s kontrolním souborem velmi malý rozsah změn v bílé hmotě mozkové. Jejich charakter svědd pro poruchu mikrostruktury bílé hmoty u nemocných. Maximum prokázaných změn leží v commissura anterior. Rozsah nalezených změn je při porovnání s publikovanými pracemi využívajícími FBA výrazně menší, což může být podmíněno mj. časnějším stadiem onemocnění probandů ve vyšetřovaném souboru.

Purpose: To assess white matter changes after the first episode of schizophrenia by using fixel-based analysis (FBA). Methods: Diffusion weighted MR images were used to prospectively compare white matter microstructure between subjects after the first episode of schizophrenia (n = 16) and healthy controls (n = 22). The subjects were examined on a 3 Tesla MRI system equipped with a 64-channel head coil. FBA was performed using MRtrix software package (version 3.0.1). Results: A limited area of fibre density decrease was found in the anterior commissure. No fixels of fibre cross-sedion changes were found. For combined metric of fibre density and cross-sedion only one statistically significant fixel was found in the right frontal white matter. Conclusion: Using FBA, only a limited area of white matter microstructure corruption was found after the first episode of schizophrenia. Changes were located mainly in the anterior commissure. Compared to other published studies using FBA, the extent of changes is distinctly smaller which can be among other causes subjected to earlier stages of the illness of subjects included in this study.

• Keywords
• fixel-based analýza,
• MeSH
• White Matter diagnostic imaging   pathology   MeSH
• Anterior Commissure, Brain diagnostic imaging   pathology   MeSH
• Humans MeSH
• Magnetic Resonance Imaging * methods   MeSH
• Prospective Studies MeSH
• Schizophrenia * diagnostic imaging   pathology   MeSH
• Diffusion Tensor Imaging methods   MeSH
• Check Tag
• Humans MeSH
• Publication type
• Research Support, Non-U.S. Gov't MeSH

Background: The research of primary progressive multiple sclerosis (PPMS) has not been able to capitalize on recent progresses in advanced magnetic resonance imaging (MRI) protocols. Objective: The presented cross-sectional study evaluated the utility of four different MRI relaxation metrics and diffusion-weighted imaging in PPMS. Methods: Conventional free precession T1 and T2, and rotating frame adiabatic T1ρ and T2ρ in combination with diffusion-weighted parameters were acquired in 13 PPMS patients and 13 age- and sex-matched controls. Results: T1ρ, a marker of crucial relevance for PPMS due to its sensitivity to neuronal loss, revealed large-scale changes in mesiotemporal structures, the sensorimotor cortex, and the cingulate, in combination with diffuse alterations in the white matter and cerebellum. T2ρ, particularly sensitive to local tissue background gradients and thus an indicator of iron accumulation, concurred with similar topography of damage, but of lower extent. Moreover, these adiabatic protocols outperformed both conventional T1 and T2 maps and diffusion tensor/kurtosis approaches, methods previously used in the MRI research of PPMS. Conclusion: This study introduces adiabatic T1ρ and T2ρ as elegant markers confirming large-scale cortical gray matter, cerebellar, and white matter alterations in PPMS invisible to other in vivo biomarkers.

• Publication type
• Journal Article MeSH

BACKGROUND AND OBJECTIVE: The geodesic ray-tracing method has shown its effectiveness for the reconstruction of fibers in white matter structure. Based on reasonable metrics on the spaces of the diffusion tensors, it can provide multiple solutions and get robust to noise and curvatures of fibers. The choice of the metric on the spaces of diffusion tensors has a significant impact on the outcome of this method. Our objective is to suggest metrics and modifications of the algorithms leading to more satisfactory results in the construction of white matter tracts as geodesics. METHODS: Starting with the DTI modality, we propose to rescale the initially chosen metric on the space of diffusion tensors to increase the geodetic cost in the isotropic regions. This change should be conformal in order to preserve the angles between crossing fibers. We also suggest to enhance the methods to be more robust to noise and to employ the fourth order tensor data in order to handle the fiber crossings properly. RESULTS: We propose a way to choose the appropriate conformal class of metrics where the metric gets scaled according to tensor anisotropy. We use the logistic functions, which are commonly used in statistics as cumulative distribution functions. To prevent deviation of geodesics from the actual paths, we propose a hybrid ray-tracing approach. Furthermore, we suggest how to employ diagonal projections of 4th order tensors to perform fiber tracking in crossing regions. CONCLUSIONS: The algorithms based on the newly suggested methods were succesfuly implemented, their performance was tested on both synthetic and real data, and compared to some of the previously known approaches.

• MeSH
• Algorithms MeSH
• Anisotropy MeSH
• White Matter * diagnostic imaging   MeSH
• Brain diagnostic imaging   MeSH
• Diffusion Tensor Imaging MeSH
• Publication type
• Journal Article MeSH

Remyelination is a naturally occurring response to demyelination and has a central role in the pathophysiology of multiple sclerosis and traumatic brain injury. Recently we demonstrated that a novel MRI technique entitled Relaxation Along a Fictitious Field (RAFF) in the rotating frame of rank n (RAFFn) achieved exceptional sensitivity in detecting the demyelination processes induced by lysophosphatidylcholine (LPC) in rat brain. In the present work, our aim was to test whether RAFF4, along with magnetization transfer (MT) and diffusion tensor imaging (DTI), would be capable of detecting the changes in the myelin content and microstructure caused by modifications of myelin sheets around axons or by gliosis during the remyelination phase after LPC-induced demyelination in the corpus callosum of rats. We collected MRI data with RAFF4, MT and DTI at 3 days after injection (demyelination stage) and at 38 days after injection (remyelination stage) of LPC (n = 12) or vehicle (n = 9). Cell density and myelin content were assessed by histology. All MRI metrics detected differences between LPC-injected and control groups of animals in the demyelination stage, on day 3. In the remyelination phase (day 38), RAFF4, MT parameters, fractional anisotropy, and axial diffusivity detected signs of a partial recovery consistent with the remyelination evident in histology. Radial diffusivity had undergone a further increase from day 3 to 38 and mean diffusivity revealed a complete recovery correlating with the histological assessment of cell density attributed to gliosis. The combination of RAFF4, MT and DTI has the potential to differentiate between normal, demyelinated and remyelinated axons and gliosis and thus it may be able to provide a more detailed assessment of white matter pathologies in several neurological diseases.

• Publication type
• Journal Article MeSH

BACKGROUND: Diffusion kurtosis imaging has been applied to evaluate white matter and basal ganglia microstructure in mixed Parkinson's disease (PD) groups with inconclusive results. OBJECTIVES: To evaluate specific patterns of kurtosis changes in PD and to assess the utility of diffusion imaging in differentiating between healthy subjects and cognitively normal PD, and between PD with and without mild cognitive impairment. METHODS: Diffusion scans were obtained in 92 participants using 3T MRI. Differences in white matter were tested by tract-based spatial statistics. Gray matter was evaluated in basal ganglia, thalamus, hippocampus, and motor and premotor cortices. Brain atrophy was also assessed. Multivariate logistic regression was used to identify a combination of diffusion parameters with the highest discrimination power between groups. RESULTS: Diffusion kurtosis metrics showed a significant increase in substantia nigra (p = 0.037, Hedges' g = 0.89), premotor (p = 0.009, Hedges' g = 0.85) and motor (p = 0.033, Hedges' g = 0.87) cortices in PD with normal cognition compared to healthy participants. Combined diffusion markers in gray matter reached 81% accuracy in differentiating between both groups. Significant white matter microstructural changes, and kurtosis decreases in the cortex were present in cognitively impaired versus cognitively normal PD. Diffusion parameters from white and gray matter differentiated between both PD phenotypes with 78% accuracy. CONCLUSIONS: Increased kurtosis in gray matter structures in cognitively normal PD reflects increased hindrance to water diffusion caused probably by alpha-synuclein-related microstructural changes. In cognitively impaired PD, the changes are mostly driven by decreased white matter integrity. Our results support the utility of diffusion kurtosis imaging for PD diagnostics.

• MeSH
• Atrophy MeSH
• Basal Ganglia diagnostic imaging   MeSH
• White Matter diagnostic imaging   MeSH
• Diffusion Magnetic Resonance Imaging MeSH
• Hippocampus diagnostic imaging   MeSH
• Cognitive Dysfunction diagnostic imaging   physiopathology   psychology   MeSH
• Middle Aged MeSH
• Humans MeSH
• Logistic Models MeSH
• Motor Cortex diagnostic imaging   MeSH
• Brain diagnostic imaging   pathology   MeSH
• Multivariate Analysis MeSH
• Parkinson Disease diagnostic imaging   physiopathology   psychology   MeSH
• Gray Matter diagnostic imaging   MeSH
• Aged MeSH
• Models, Statistical MeSH
• Thalamus diagnostic imaging   MeSH
• Diffusion Tensor Imaging MeSH
• Check Tag
• Middle Aged MeSH
• Humans MeSH
• Male MeSH
• Aged MeSH
• Female MeSH
• Publication type
• Journal Article MeSH
• Research Support, Non-U.S. Gov't MeSH

Diffusion magnetic resonance imaging (dMRI) proved promising in patients with non-myelopathic degenerative cervical cord compression (NMDCCC), i.e., without clinically manifested myelopathy. Aim of the study is to present a fast multi-shell HARDI-ZOOMit dMRI protocol and validate its usability to detect microstructural myelopathy in NMDCCC patients. In 7 young healthy volunteers, 13 age-comparable healthy controls, 18 patients with mild NMDCCC and 15 patients with severe NMDCCC, the protocol provided higher signal-to-noise ratio, enhanced visualization of white/gray matter structures in microstructural maps, improved dMRI metric reproducibility, preserved sensitivity (SE = 87.88%) and increased specificity (SP = 92.31%) of control-patient group differences when compared to DTI-RESOLVE protocol (SE = 87.88%, SP = 76.92%). Of the 56 tested microstructural parameters, HARDI-ZOOMit yielded significant patient-control differences in 19 parameters, whereas in DTI-RESOLVE data, differences were observed in 10 parameters, with mostly lower robustness. Novel marker the white-gray matter diffusivity gradient demonstrated the highest separation. HARDI-ZOOMit protocol detected larger number of crossing fibers (5-15% of voxels) with physiologically plausible orientations than DTI-RESOLVE protocol (0-8% of voxels). Crossings were detected in areas of dorsal horns and anterior white commissure. HARDI-ZOOMit protocol proved to be a sensitive and practical tool for clinical quantitative spinal cord imaging.

• MeSH
• Biomedical Engineering MeSH
• Diffusion Magnetic Resonance Imaging * MeSH
• Adult MeSH
• Spinal Cord Compression diagnostic imaging   pathology   MeSH
• Cervical Vertebrae pathology   MeSH
• Middle Aged MeSH
• Humans MeSH
• Spinal Cord Diseases diagnostic imaging   pathology   MeSH
• Signal-To-Noise Ratio MeSH
• Reproducibility of Results MeSH
• Sensitivity and Specificity MeSH
• Cluster Analysis MeSH
• Case-Control Studies MeSH
• Diffusion Tensor Imaging MeSH
• Check Tag
• Adult MeSH
• Middle Aged MeSH
• Humans MeSH
• Male MeSH
• Female MeSH
• Publication type
• Journal Article MeSH
• Research Support, Non-U.S. Gov't MeSH
• Research Support, N.I.H., Extramural MeSH

Even though MRI visualization of white matter lesions is pivotal for the diagnosis and management of multiple sclerosis (MS), the issue of detecting diffuse brain tissue damage beyond the apparent T2-hyperintense lesions continues to spark considerable interest. Motivated by the notion that rotating frame MRI methods are sensitive to slow motional regimes critical for tissue characterization, here we utilized novel imaging protocols of rotating frame MRI on a clinical 3 Tesla platform, including adiabatic longitudinal, T1ρ, and transverse, T2ρ, relaxation methods, and Relaxation Along a Fictitious Field (RAFF) in the rotating frame of rank 4 (RAFF4), in 10 relapsing-remitting multiple sclerosis patients and 10 sex- and age-matched healthy controls. T1ρ, T2ρ and RAFF4 relaxograms extracted from the whole white matter exhibited a significant shift towards longer relaxation time constants in MS patients as compared to controls. T1ρ and RAFF4 detected alterations even when considering only regions of normally appearing white matter (NAWM), while other MRI metrics such as T1w/T2w ratio and diffusion tensor imaging measures failed to find group differences. In addition, RAFF4, T2ρ and, to a lesser extent, T1ρ showed differences in subcortical grey matter structures, mainly hippocampus, whereas no functional changes in this region were detected in resting-state functional MRI metrics. We conclude that rotating frame MRI techniques are exceptionally sensitive methods for the detection of subtle abnormalities not only in NAWM, but also in deep grey matter in MS, where they surpass even highly sensitive measures of functional changes, which are often suggested to precede detectable structural alterations. Such abnormalities are consistent with a wide spectrum of different, but interconnected pathological features of MS, including the loss of neuronal cells and their axons, decreased levels of myelin even in NAWM, and altered iron content.

• MeSH
• White Matter diagnostic imaging   pathology   MeSH
• Adult MeSH
• Image Interpretation, Computer-Assisted methods   MeSH
• Middle Aged MeSH
• Humans MeSH
• Magnetic Resonance Imaging methods   MeSH
• Brain pathology   MeSH
• Neuroimaging methods   MeSH
• Cross-Sectional Studies MeSH
• Multiple Sclerosis, Relapsing-Remitting diagnostic imaging   pathology   MeSH
• Check Tag
• Adult MeSH
• Middle Aged MeSH
• Humans MeSH
• Male MeSH
• Female MeSH
• Publication type
• Journal Article MeSH
• Research Support, Non-U.S. Gov't MeSH
• Research Support, N.I.H., Extramural MeSH

Despite recent success of deep learning models in numerous applications, their widespread use on mobile devices is seriously impeded by storage and computational requirements. In this paper, we propose a novel network compression method called Adaptive Dimension Adjustment Tucker decomposition (ADA-Tucker). With learnable core tensors and transformation matrices, ADA-Tucker performs Tucker decomposition of arbitrary-order tensors. Furthermore, we propose that weight tensors in networks with proper order and balanced dimension are easier to be compressed. Therefore, the high flexibility in decomposition choice distinguishes ADA-Tucker from all previous low-rank models. To compress more, we further extend the model to Shared Core ADA-Tucker (SCADA-Tucker) by defining a shared core tensor for all layers. Our methods require no overhead of recording indices of non-zero elements. Without loss of accuracy, our methods reduce the storage of LeNet-5 and LeNet-300 by ratios of 691× and 233 ×, respectively, significantly outperforming state of the art. The effectiveness of our methods is also evaluated on other three benchmarks (CIFAR-10, SVHN, ILSVRC12) and modern newly deep networks (ResNet, Wide-ResNet).

• MeSH
• Benchmarking MeSH
• Deep Learning * trends   MeSH
• Data Compression methods   trends   MeSH
• Humans MeSH
• Neural Networks, Computer * MeSH
• Machine Learning MeSH
• Check Tag
• Humans MeSH
• Publication type
• Journal Article MeSH

Background: We hypothesized that right and left temporal lobe epilepsy (RTLE and LTLE, respectively) have distinctive spatial patterns of white matter (WM) changes that can be differentiated and interpreted with the use of multiple diffusion parameters. We compared the global microstructure of fiber bundles with regard to WM alterations in both RTLE and LTLE, addressing some of the methodological issues of previous studies. Methods: Diffusion tensor imaging data from 17 patients with RTLE (age: 40.7 ± 10.4), 15 patients with LTLE (age: 37.3 ± 10.4), and 15 controls (age: 34.8 ± 11.2) were used in the study. WM integrity was quantified by fractional anisotropy (FA), mean diffusivity (MD), longitudinal diffusivity (LD), and radial diffusivity (RD). The diffusion parameters were compared between the groups in tracts representing the core of the fiber bundles. The volumes of hippocampi and amygdala were subsequently compared across the groups, while the data were adjusted for the effect of hippocampal sclerosis. Results: Significantly reduced FA and increased MD, LD, and RD were found bilaterally over widespread brain regions in RTLE. An increase in MD and RD values was observed in widespread WM fiber bundles ipsilaterally in LTLE, largely overlapping with regions where FA was lower, while no increase in LD was observed. We also found a difference between the LTLE and RTLE groups for the right hippocampal volume (with and without adjustment for HS), whereas no significant volume differences were found between patients and controls. Conclusions: It appears that patients with RTLE exhibit a more widespread pattern of WM alterations that extend far beyond the temporal lobe in both ipsilateral and contralateral hemisphere; furthermore, these changes seem to reflect more severe damage related to chronic degeneration. Conversely, more restrained changes in the LTLE may imply a pattern of less severe axonal damage, more restricted to ipsilateral hemisphere. Comprehensive finding of more prominent hippocampal atrophy in the RTLE raises an interesting issue of seizure-induced implications on gray matter and WM microstructure that may not necessarily mean a straightforward causal relationship. Further correlations of diffusion-derived metrics with neuropsychological and functional imaging measures may provide complementary information on underlying WM abnormalities with regard to functional hemispheric specialization.

• Publication type
• Journal Article MeSH