Amyloid burden identifies neuropsychological phenotypes at increased risk of progression to Alzheimer's disease in mild cognitive impairment patients.

PURPOSE: The extent of amyloid burden associated with cognitive impairment in amnestic mild cognitive impairment is unknown. The primary aim of the study was to determine the extent to which amyloid burden is associated to the cognitive impairment. The secondary objective was to test the relationship between amyloid accumulation and memory or cognitive impairment. MATERIALS AND METHODS: In this prospective study 66 participants with amnestic mild cognitive impairment underwent clinical, neuropsychological and PET amyloid imaging tests. Composite scores assessing memory and non-memory domains were used to identify two clinical classes of neuropsychological phenotypes expressing different degree of cognitive impairment. Detection of amyloid status and definition of optimal amyloid ± cutoff for discrimination relied on unsupervised k-means clustering method. RESULTS: Threshold for identifying low and high amyloid retention groups was of SUVr = 1.3. Aß + participants showed poorer global cognitive and episodic memory performance than subjects with low amyloid deposition. Aß positivity significantly identified individuals with episodic memory impairment with a sensitivity and specificity of 80 and 79%, (χ2 = 21.48; P < 0.00001). Positive and negative predictive values were 82 and 76%, respectively. Amyloid deposition increased linearly as function of memory impairment with a rate of 0.13/ point of composite memory score (R = -44, P = 0.0003). CONCLUSION: The amyloid burden of SUVr = 1.3 allows early identification of subjects with episodic memory impairment which might predict progression from MCI to Alzheimer's disease. TRIAL REGISTRATION: EudraCT 2015-001184-39.

No evidence for an effect on brain atrophy rate of atorvastatin add-on to interferon ß1b therapy in relapsing-remitting multiple sclerosis (the ARIANNA study).

BACKGROUND: A previous phase 2 trial has suggested that statins might delay brain atrophy in secondary progressive multiple sclerosis. OBJECTIVES: The objective of this study was to evaluate the effect of atorvastatin add-on therapy on cerebral atrophy in relapsing-remitting multiple sclerosis. METHODS: This randomised, placebo-controlled study compared atorvastatin 40 mg or placebo add-on therapy to interferon ß1b for 24 months. Brain magnetic resonance imaging, multiple sclerosis functional composite score, Rao neuropsychological battery and expanded disability status scale were evaluated over 24 months. RESULTS: A total of 154 patients were randomly assigned, 75 in the atorvastatin and 79 in the placebo arms, with a comparable drop-out rate (overall 23.4%). Brain atrophy over 2 years was not different in the two arms (-0.38% and -0.32% for the atorvastatin and placebo groups, respectively). Relapse rate, expanded disability status scale, multiple sclerosis functional composite score or cognitive changes were not different in the two arms. Patients withdrawing from the study had a higher number of relapses in the previous 2 years (P=0.04) and a greater probability of relapsing within 12 months. CONCLUSIONS: Our results suggest that the combination of atorvastatin and interferon ß1b is not justified in early relapsing-remitting multiple sclerosis and adds to the body of evidence indicating an absence of significant radiological and clinical benefit of statins in relapsing-remitting multiple sclerosis.

StepBrain: A 3-Dimensionally Printed Multicompartmental Anthropomorphic Brain Phantom to Simulate PET Activity Distributions.

An innovative multicompartmental anatomic brain phantom (StepBrain) is described to simulate the in vivo tracer uptake of gray matter, white matter, and striatum, overcoming the limitations of currently available phantoms. Methods: StepBrain was created by exploiting the potential of fused deposition modeling 3-dimensional printing to replicate the real anatomy of the brain compartments, as modeled through ad hoc processing of healthy-volunteer MR images. Results: A realistic simulation of 18F-FDG PET brain studies, using target activity to obtain the real concentration ratios, was obtained, and the results of postprocessing with partial-volume effect correction tools developed for human PET studies confirmed the accuracy of these methods in recovering the target activity concentrations. Conclusion: StepBrain compartments (gray matter, white matter, and striatum) can be simultaneously filled, achieving different concentration ratios and allowing the simulation of different (e.g., amyloid, tau, or 6-fluoro-l-dopa) tracer distributions, with a potentially valuable role for multicenter PET harmonization studies.

A polynomial regression-based approach to estimate relaxation rate maps suitable for multiparametric segmentation of clinical brain MRI studies in multiple sclerosis.

BACKGROUND AND OBJECTIVE: Relaxation parameter maps (RPMs) calculated from spin-echo data have provided a basis for the segmentation of normal brain tissues and white matter lesions in multiple sclerosis (MS) MRI studies. However, Conventional Spin-Echo (CSE) sequences, once the core of clinical MRI studies, have been largely replaced by faster ones, which do not allow the calculation a-posteriori of RPMs from clinical studies. Aim of the study was to develop and validate a method to estimate RPMs (pseudo-RPMs) from routine clinical MRI protocols (including 3D-Gradient Echo T1w, FLAIR and fast-T2w sequences), suitable for fully automatic multiparametric segmentation of normal-appearing and pathological brain tissues in MS. METHODS: The proposed method processes spatially normalized clinical MRI studies through a multistep pipeline, to collect a set of data points of matched signal intensities (from MRI studies) and relaxation parameters (from a CSE-derived digital template and an MS lesion database), which are then fitted by a multiple and multivariate 4-th degree polynomial regression, providing pseudo-RPMs. The method was applied to a dataset of 59 clinical MRI studies providing pseudo-RPMs that were segmented through a method originally developed for the CSE-derived RPMs. Results of the segmentation in 12 studies were used to iteratively optimize method parameters. Accuracy of segmentation of normal-appearing brain tissues from the pseudo-RPMs was assessed by comparing their age-related changes, as measured in 47 clinical studies, against those measured acquired using CSE sequences in a comparable dataset of 47 patients. Lesion segmentation was validated against manual segmentation carried out by three neuroradiologists. RESULTS: Age-related changes of normal-appearing brain tissue volumes measured using the pseudo-RPMs substantially overlapped those measured using the RPMs obtained from CSE sequences, and segmentation of MS lesions showed a moderate-high spatial overlap with manual segmentation, comparable to that achieved by the widely used Lesion Segmentation Tool on FLAIR images, with a greater volumetric agreement. CONCLUSIONS: The proposed approach allows calculation from clinical studies of pseudo-RPMs, which are equivalent to those obtainable from CSE sequences, avoiding the need for the acquisition of additional, dedicated sequences for segmentation purposes.

Voxel-based analysis of gray matter relaxation rates shows different correlation patterns for cognitive impairment and physical disability in relapsing-remitting multiple sclerosis.

BACKGROUND: Regional analyses of markers of microstructural gray matter (GM) changes, including relaxation rates, have shown inconsistent correlations with physical and cognitive impairment in MS. OBJECTIVE: To assess voxelwise the correlation of the R1 and R2 relaxation rates with the physical and cognitive impairment in MS. METHODS: GM R1 and R2 relaxation rate maps were obtained in 241 relapsing-remitting MS patients by relaxometric segmentation of MRI studies. Correlations with the Expanded Disability Status Scale (EDSS) and the percentage of impaired cognitive test (Brief Repeatable Battery and Stroop Test, available in 186 patients) were assessed voxelwise, including voxel GM content as nuisance covariate to remove the effect of atrophy on the correlations. RESULTS: Extensive clusters of inverse correlation between EDSS and R2 were detected throughout the brain, while inverse correlations with R1 were mostly limited to perirolandic and supramarginal cortices. Cognitive impairment correlated negatively with R1, and to a lesser extent with R2, in the middle frontal, mesial temporal, midcingulate and medial parieto-occipital cortices. CONCLUSION: In relapsing-remitting MS patients, GM microstructural changes correlate diffusely with physical disability, independent of atrophy, with a preferential role of the sensorimotor cortices. Neuronal damage in the limbic system and dorsolateral prefrontal cortices correlates with cognitive dysfunction.

A New Relational Database Including Clinical Data and Myocardial Perfusion Imaging Findings in Coronary Artery Disease.

BACKGROUND: The aim of this study was to test a relational database including clinical data and imaging findings in a large cohort of subjects with suspected or known Coronary Artery Disease (CAD) undergoing stress single-photon emission computed tomography (SPECT) myocardial perfusion imaging. METHODS: We developed a relational database including clinical and imaging data of 7995 subjects with suspected or known CAD. The software system was implemented by PostgreSQL 9.2, an open source object-relational database, and managed from remote by pgAdmin III. Data were arranged according to a logic of aggregation and stored in a schema with twelve tables. Statistical software was connected to the database directly downloading data from server to local personal computer. RESULTS: There was no problem or anomaly for database implementation and user connections to the database. The epidemiological analysis performed on data stored in the database demonstrated abnormal SPECT findings in 46% of male subjects and 19% of female subjects. Imaging findings suggest that the use of SPECT imaging in our laboratory is appropriate. CONCLUSION: The development of a relational database provides a free software tool for the storage and management of data in line with the current standard.

Brain tissue volumes and relaxation rates in multiple sclerosis: implications for cognitive impairment.

OBJECTIVE: Both normal gray matter atrophy and brain tissue relaxation rates, in addition to total lesion volume, have shown significant correlations with cognitive test scores in multiple sclerosis (MS). Aim of the study was to assess the relative contributions of macro- and microstructural changes of both normal and abnormal brain tissues, probed, respectively, by their volumes and relaxation rates, to the cognitive status and physical disability of MS patients. METHODS: MRI studies from 241 patients with relapsing-remitting MS were retrospectively analyzed by fully automated multiparametric relaxometric segmentation. Ordinal backward regression analysis was applied to the resulting volumes and relaxation rates of both normal (gray matter, normal-appearing white matter and CSF) and abnormal (T2-weighted lesions) brain tissues, controlling for age, sex and disease duration, to identify the main independent contributors to the cognitive status, as measured by the percentage of failed tests at a cognitive test battery (Rao's Brief Repeatable Battery and Stroop test, available in 186 patients), and to the physical disability, as assessed by the Expanded Disability Status Scale (EDSS). RESULTS: The R1 relaxation rate (a putative marker of tissue disruption) of the MS lesions appeared the single most significant contributor to cognitive impairment (p < 0.001). On the contrary, the EDSS appeared mainly affected by the decrease in R2 of the gray matter (p < 0.0001), (possibly influenced by cortical plaques, edema and inflammation). CONCLUSIONS: In RR-MS the tissue damage in white matter lesions appears the single main determinant of the cognitive status of patients, likely through disconnection phenomena, while the physical disability appears related to the involvement of gray matter.

Correlation between fatigue and brain atrophy and lesion load in multiple sclerosis patients independent of disability.

BACKGROUND: Fatigue is a major problem in multiple sclerosis (MS), and its association with MRI features is debated. OBJECTIVE: To study the correlation between fatigue and lesion load, white matter (WM), and grey matter (GM), in MS patients independent of disability. METHODS: We studied 222 relapsing remitting MS patients with low disability (scores or=5; n=197) and low-fatigue groups (FSS

Neuropsychological assessment, quantitative MRI and ApoE gene polymorphisms in a series of MS patients treated with IFN beta-1b.

Few trials issued the effect of disease-modifying medications on cognitive functions in multiple sclerosis. We designed an open-label longitudinal study to evaluate, during 2 years, cognitive performance and its relationship with MRI data and ApoE polymorphism findings in a group of relapsing-remitting (RR) multiple sclerosis (MS) Interferon (IFN) beta-1b-treated patients (median age 30 years, median disease duration 3.4 years). Complete neuropsychological battery was grouped into attention, information learning/memory, language and visuo-spatial functions. Fifty-two patients (33 females) were enrolled in the study. Six patients (11.5%) dropped out, mainly due to side effects. At baseline neuropsychological evaluation, we found 54% normal, 42% mildly impaired and 4% moderately impaired patients. At 2 years follow-up, cognitive status was stable in 65%, improved in 33% and worsened in 2% of patients. No significant relations were found between global cognitive outcome vs. EDSS change, clinical disease activity, MRI data or ApoE gene polymorphisms over the 2 years follow-up. EDSS and MRI fractional volumes were found to correlate with the performance at single tests. Twenty-one patients (45.6%) showed active MRI scans throughout the study, without any worsening at the corresponding neuropsychological examination. This ongoing trial suggests a possible beneficial effect of IFN beta-1b treatment on cognitive functions in RRMS patients. Extension of follow-up and further data analyses are needed to confirm and clarify these findings.

Diffusion volume (DV) measurement in endometrial and cervical cancer: A new MRI parameter in the evaluation of the tumor grading and the risk classification.

PURPOSE: A new MRI parameter representative of active tumor burden is proposed: diffusion volume (DV), defined as the sum of all the voxels within a tumor with apparent diffusion coefficient (ADC) values within a specific range. The aims of the study were: (a) to calculate DV on ADC maps in patients with cervical/endometrial cancer; (b) to correlate DV with histological grade (G) and risk classification; (c) to evaluate intra/inter-observer agreement of DV calculation. MATERIALS AND METHODS: Fifty-three patients with endometrial (n=28) and cervical (n=25) cancers underwent pelvic MRI with DWI sequences. Both endometrial and cervical tumors were classified on the basis of G (G1/G2/G3) and FIGO staging (low/medium/high-risk). A semi-automated segmentation procedure was used to calculate the DV. A freehand closed ROI outlined the whole visible tumor on the most representative slice of ADC maps defined as the slice with the maximum diameter of the solid neoplastic component. Successively, two thresholds were generated on the basis of the mean and standard deviation (SD) of the ADC values: lower threshold (LT="mean minus three SD") and higher threshold (HT="mean plus one SD"). The closed ROI was expanded in 3D, including all the contiguous voxels with ADC values in the range LT-HT × 10-3mm(2)/s. A Kruskal-Wallis test was used to assess the differences in DV among G and risk groups. Intra-/inter-observer variability for DV measurement was analyzed according to the method of Bland and Altman and the intraclass-correlation-coefficient (ICC). RESULTS: DV values were significantly different among G and risk groups in both endometrial (p<0.05) and cervical cancers (p ≤ 0.01). For endometrial cancer, DV of G1 (mean ± sd: 2.81 ± 3.21 cc) neoplasms were significantly lower than G2 (9.44 ± 9.58 cc) and G3 (11.96 ± 8.0 cc) ones; moreover, DV of low risk cancers (5.23 ± 8.0 cc) were significantly lower than medium (7.28 ± 4.3 cc) and high risk (14.7 ± 9.9 cc) ones. For cervical cancer, DV of G1 (0.31 ± 0.13 cc) neoplasms was significantly lower than G3 (40.68 ± 45.65 cc) ones; moreover, DV of low risk neoplasms (6.98 ± 8.08 cc) was significantly lower than medium (21.7 ± 17.13 cc) and high risk (62.9 ± 51.12 cc) ones and DV of medium risk neoplasms was significantly lower than high risk ones. The intra-/inter-observer variability for DV measurement showed an excellent correlation for both cancers (ICC ≥ 0.86). CONCLUSIONS: DV is an accurate index for the assessment of G and risk classification of cervical/endometrial cancers with low intra-/inter-observer variability.

Self-Trained Supervised Segmentation of Subcortical Brain Structures Using Multispectral Magnetic Resonance Images.

The aim of this paper is investigate the feasibility of automatically training supervised methods, such as k-nearest neighbor (kNN) and principal component discriminant analysis (PCDA), and to segment the four subcortical brain structures: caudate, thalamus, pallidum, and putamen. The adoption of supervised classification methods so far has been limited by the need to define a representative training dataset, operation that usually requires the intervention of an operator. In this work the selection of the training data was performed on the subject to be segmented in a fully automated manner by registering probabilistic atlases. Evaluation of automatically trained kNN and PCDA classifiers that combine voxel intensities and spatial coordinates was performed on 20 real datasets selected from two publicly available sources of multispectral magnetic resonance studies. The results demonstrate that atlas-guided training is an effective way to automatically define a representative and reliable training dataset, thus giving supervised methods the chance to successfully segment magnetic resonance brain images without the need for user interaction.

A Novel Multiparametric Approach to 3D Quantitative MRI of the Brain.

Magnetic Resonance properties of tissues can be quantified in several respects: relaxation processes, density of imaged nuclei, magnetism of environmental molecules, etc. In this paper, we propose a new comprehensive approach to obtain 3D high resolution quantitative maps of arbitrary body districts, mainly focusing on the brain. The theory presented makes it possible to map longitudinal (R1), pure transverse (R2) and free induction decay ([Formula: see text]) rates, along with proton density (PD) and magnetic susceptibility (χ), from a set of fast acquisition sequences in steady-state that are highly insensitive to flow phenomena. A novel denoising scheme is described and applied to the acquired datasets to enhance the signal to noise ratio of the derived maps and an information theory approach compensates for biases from radio frequency (RF) inhomogeneities, if no direct measure of the RF field is available. Finally, the results obtained on sample brain scans of healthy controls and multiple sclerosis patients are presented and discussed.

Improving Signal-to-Noise Ratio in Susceptibility Weighted Imaging: A Novel Multicomponent Non-Local Approach.

In susceptibility-weighted imaging (SWI), the high resolution required to obtain a proper contrast generation leads to a reduced signal-to-noise ratio (SNR). The application of a denoising filter to produce images with higher SNR and still preserve small structures from excessive blurring is therefore extremely desirable. However, as the distributions of magnitude and phase noise may introduce biases during image restoration, the application of a denoising filter is non-trivial. Taking advantage of the potential multispectral nature of MR images, a multicomponent approach using a Non-Local Means (MNLM) denoising filter may perform better than a component-by-component image restoration method. Here we present a new MNLM-based method (Multicomponent-Imaginary-Real-SWI, hereafter MIR-SWI) to produce SWI images with high SNR and improved conspicuity. Both qualitative and quantitative comparisons of MIR-SWI with the original SWI scheme and previously proposed SWI restoring pipelines showed that MIR-SWI fared consistently better than the other approaches. Noise removal with MIR-SWI also provided improvement in contrast-to-noise ratio (CNR) and vessel conspicuity at higher factors of phase mask multiplications than the one suggested in the literature for SWI vessel imaging. We conclude that a proper handling of noise in the complex MR dataset may lead to improved image quality for SWI data.

In vivo imaging and characterization of [(18)F]DPA-714, a potential new TSPO ligand, in mouse brain and peripheral tissues using small-animal PET.

INTRODUCTION: The translocator protein 18 kDa (TSPO), a biochemical marker of neuroinflammation, is highly expressed in the brain activated microglia and it is also expressed by peripheral inflammatory cells and normal peripheral tissues. Thus, development of radioligands for the TSPO may contribute to further understanding the in vivo TSPO function in central and peripheral inflammatory processes and other pathologies. Here, we report the biodistribution, the specific binding and the radiometabolites of [(18)F]DPA-714, a promising fluorinated PET radiotracer, in normal mice using a microPET/CT scanner. METHODS: The in vivo biodistribution and kinetics of [(18)F]DPA-714 were measured in mice brain and peripheral tissues. Specific binding to TSPO sites was assessed using pharmacological competitive studies by means of saturation experiments performed by i.v. injection of 1mg/kg of unlabeled DPA-714 or 3mg/kg of unlabeled PK11195. A region of interest analysis was performed to generate time-activity curves in the brain, heart, lung, kidney, spleen and liver. Metabolites assay was performed in the plasma and peripheral organs by radio-HPLC. RESULTS: [(18)F]DPA-714 reached high concentration in lung, heart, kidney and spleen, tissues well known to be rich in TSPO sites. [(18)F]DPA-714 kinetics were faster in the lung and slower in the kidney. Pre-injection of unlabeled DPA-714 or PK11195 inhibited about 80% of [(18)F]DPA-714 uptake in the lung and heart (p<0.0005). The percentage of inhibition in the kidney was lower and achieved at later times only with DPA-714 (p<0.05) but not with PK11195. Sixty minutes after radiotracer injection only unmetabolized radioligand was found in the brain, lung, heart and spleen. CONCLUSION: These results suggest that [(18)F]DPA-714 is a suitable PET ligand for imaging in mice brain and peripheral tissues since it binds with high specificity TSPO binding sites and it is almost unchanged at 60 minutes after radiotracer injection in the brain and TSPO-rich regions.

Lesion load may predict long-term cognitive dysfunction in multiple sclerosis patients.

BACKGROUND: Magnetic Resonance Imaging (MRI) techniques provided evidences into the understanding of cognitive impairment (CIm) in Multiple Sclerosis (MS). OBJECTIVES: To investigate the role of white matter (WM) and gray matter (GM) in predicting long-term CIm in a cohort of MS patients. METHODS: 303 out of 597 patients participating in a previous multicenter clinical-MRI study were enrolled (49.4% were lost at follow-up). The following MRI parameters, expressed as fraction (f) of intracranial volume, were evaluated: cerebrospinal fluid (CSF-f), WM-f, GM-f and abnormal WM (AWM-f), a measure of lesion load. Nine years later, cognitive status was assessed in 241 patients using the Symbol Digit Modalities Test (SDMT), the Semantically Related Word List Test (SRWL), the Modified Card Sorting Test (MCST), and the Paced Auditory Serial Addition Test (PASAT). In particular, being SRWL a memory test, both immediate recall and delayed recall were evaluated. MCST scoring was calculated based on the number of categories, number of perseverative and non-perseverative errors. RESULTS: AWM-f was predictive of an impaired performance 9 years ahead in SDMT (OR 1.49, CI 1.12-1.97 p = 0.006), PASAT (OR 1.43, CI 1.14-1.80 p = 0.002), SRWL-immediate recall (OR 1.72 CI 1.35-2.20 p<0.001), SRWL-delayed recall (OR 1.61 CI 1.28-2.03 p<0.001), MCST-category (OR 1.52, CI 1.2-1.9 p<0.001), MCST-perseverative error(OR 1.51 CI 1.2-1.9 p = 0.001), MCST-non perseverative error (OR 1.26 CI 1.02-1.55 p = 0.032). CONCLUSION: In our large MS cohort, focal WM damage appeared to be the most relevant predictor of the long-term cognitive outcome.

Integrated software for the analysis of brain PET/SPECT studies with partial-volume-effect correction.

UNLABELLED: We present software for integrated analysis of brain PET studies and coregistered segmented MRI that couples a module for automated placement of regions of interest (ROI) with 4 alternative methods for partial-volume-effect correction (PVEc). The accuracy and precision of these methods have been measured using 4 simulated (18)F-FDG PET studies with increasing degrees of atrophy. METHODS: The software allows the application of a set of labels, defined a priori in the Talairach space, to segmented and coregistered MRI. Resulting ROIs are then transferred onto the PET study, and corresponding values are corrected according to the 4 PVEc techniques under investigation, providing corresponding corrected values. To evaluate the PVEc techniques, the software was applied to 4 simulated (18)F-FDG PET studies, introducing increasingly larger experimental errors, including errors in coregistration (0- to 6-pixel misregistration), segmentation (-13.7% to 14.1% gray matter [GM] volume change) and resolution estimate errors (-16.9% to 26.8% full-width-at-half-maximum mismatch). RESULTS: Even in the absence of segmentation and coregistration errors, uncorrected PET values showed -37.6% GM underestimation and 91.7% WM overestimation. Voxel-based correction only for the loss of GM activity as a result of spill-out onto extraparenchymal tissues left a residual underestimation of GM values (-21.2%). Application of the method that took into account both spill-in and spill-out effects between any possible pair of ROIs (R-PVEc) and of the voxel-based method that corrects also for the WM activity derived from R-PVEC (mMG-PVEc) provided an accuracy above 96%. The coefficient of variation of the GM ROIs, a measure of the imprecision of the GM concentration estimates, was 8.5% for uncorrected PET data and decreased with PVEc, reaching 6.0% for mMG-PVEc. Coregistration errors appeared to be the major determinant of the imprecision. CONCLUSION: Coupling of automated ROI placement and PVEc provides a tool for integrated analysis of brain PET/MRI data, which allows a recovery of true GM ROI values, with a high degree of accuracy when R-PVEc or mMG-PVEc is used. Among the 4 tested PVEc methods, R-PVEc showed the greatest accuracy and is suitable when corrected images are not specifically needed. Otherwise, if corrected images are desired, the mMG-PVEc method appears the most adequate, showing a similar accuracy.

Segmentation of magnetic resonance brain images through discriminant analysis.

Segmentation (tissue classification) of medical images obtained from a magnetic resonance (MR) system is a primary step in most applications of medical image post-processing. This paper describes nonparametric discriminant analysis methods to segment multispectral MR images of the brain. Starting from routinely available spin-lattice relaxation time, spin-spin relaxation time, and proton density weighted images (T1w, T2w, PDw), the proposed family of statistical methods is based on: (i) a transform of the images into components that are statistically independent from each other; (ii) a nonparametric estimate of probability density functions of each tissue starting from a training set; (iii) a classic Bayes 0-1 classification rule. Experiments based on a computer built brain phantom (brainweb) and on eight real patient data sets are shown. A comparison with parametric discriminant analysis is also reported. The capability of nonparametric discriminant analysis in improving brain tissue classification of parametric methods is demonstrated. Finally, an assessment of the role of multispectrality in classifying brain tissues is discussed.

Parametric MR dynamic imaging for breast lesions characterization and prediction of lymph nodes involvement.

The main purpose of this study was to evaluate Gd-DTPA kinetics for the differential diagnosis between malignant and benign breast lesions. As a secondary aim, Gd-DTPA kinetics in malignant lesions was tested for predicting axillary lymph nodes involvement. Eighty-eight patients who underwent MRI for suspected breast tumor were selected from our database. All patients underwent the same acquisition protocol consisting of pre-contrast and dynamic contrastenhanced MRI. For all of them clinical and histopathological data were available. MR studies were performed on the same 1.5T scanner with a standard dedicated breast coil. Pre- and post-contrast dynamic images were used to calculate R1, R2 relaxation rates and proton density maps. The maximum influx rate (MIR) as well as the corresponding time (TMIR) were derived using R1 relaxation rate maps and relative changes as a function of time. Histopatological analysis led to the diagnosis of 46 breast carcinomas and 42 benign lesions. All breast carcinomas and 24 out of 42 benign lesions showed contrast-enhancement. The mean MIR was 0.75±0.14 (SD) sec-(2) in malignant tumors and 0.53±0.14 (SD) sec-(2) in contrast-enhancing benign breast lesions (p<0.0001). The TMIR was 1.40±0.43 min and 3.01±1.92 min (mean±SD) in enhancing malignant and benign lesions, respectively (p<0.0001). In malignant tumors, TMIR was not significantly different in node negative and node positive carcinomas whereas MIR was significantly lower in node negative carcinomas (0.67±0.11 versus 0.83±0.12 respectively, p<0.0001). Our findings suggest that quantitative assessment of Gd-DTPA kinetics may be an additional tool characterized for enhancing breast lesions and for predicting axillary lymph nodes involvement in malignant breast carcinoma.

Evaluation of supervised methods for the classification of major tissues and subcortical structures in multispectral brain magnetic resonance images.

This work investigates the capability of supervised classification methods in detecting both major tissues and subcortical structures using multispectral brain magnetic resonance images. First, by means of a realistic digital brain phantom, we investigated the classification performance of various Discriminant Analysis methods, K-Nearest Neighbor and Support Vector Machine. Then, using phantom and real data, we quantitatively assessed the benefits of integrating anatomical information in the classification, in the form of voxels coordinates as additional features to the intensities or tissue probabilistic atlases as priors. In addition we tested the effect of spatial correlations between neighboring voxels and image denoising. For each brain tissue we measured the classification performance in terms of global agreement percentage, false positive and false negative rates and kappa coefficient. The effectiveness of integrating spatial information or a tissue probabilistic atlas has been demonstrated for the aim of accurately classifying brain magnetic resonance images.

Descent algorithms on oblique manifold for source-adaptive ICA contrast.

A Riemannian manifold optimization strategy is proposed to facilitate the relaxation of the orthonormality constraint in a more natural way in the course of performing independent component analysis (ICA) that employs a mutual information-based source-adaptive contrast function. Despite the extensive development of manifold techniques catering to the orthonormality constraint, only a limited number of works have been dedicated to oblique manifold (OB) algorithms to intrinsically handle the normality constraint, which has been empirically shown to be superior to other Riemannian and Euclidean approaches. Imposing the normality constraint implicitly, in line with the ICA definition, essentially guarantees a substantial improvement in the solution accuracy, by way of increased degrees of freedom while searching for an optimal unmixing ICA matrix, in contrast with the orthonormality constraint. Designs of the steepest descent, conjugate gradient with Hager-Zhang or a hybrid update parameter, quasi-Newton, and cost-effective quasi-Newton methods intended for OB are presented in this paper. Their performance is validated using natural images and systematically compared with the popular state-of-the-art approaches in order to assess the performance effects of the choice of algorithm and the use of a Riemannian rather than Euclidean framework. We surmount the computational challenge associated with the direct estimation of the source densities using the improved fast Gauss transform in the evaluation of the contrast function and its gradient. The proposed OB schemes may find applications in the offline image/signal analysis, wherein, on one hand, the computational overhead can be tolerated, and, on the other, the solution quality holds paramount interest.