Stefano Diciotti

Neural networks for computer-aided diagnosis: detection of lung nodules in chest radiograms

The paper describes a neural-network-based system for the computer aided detection of lung nodules in chest radiograms. Our approach is based on multiscale processing and artificial neural networks (ANNs). The problem of nodule detection is faced by using a two-stage architecture including: 1) an attention focusing subsystem that processes whole radiographs to locate possible nodular regions ensuring high sensitivity; 2) a validation subsystem that processes regions of interest to evaluate the likelihood of the presence of a nodule, so as to reduce false alarms and increase detection specificity. Biologically inspired filters (both LoG and Gabor kernels) are used to enhance salient image features. ANNs of the feedforward type are employed, which allow an efficient use of a priori knowledge about the shape of nodules, and the background structure. The images from the public JSRT database, including 247 radiograms, were used to build and test the system. We performed a further test by using a second private database with 65 radiograms collected and annotated at the Radiology Department of the University of Florence. Both data sets include nodule and nonnodule radiographs. The use of a public data set along with independent testing with a different image set makes the comparison with other systems easier and allows a deeper understanding of system behavior. Experimental results are described by ROC/FROC analysis. For the JSRT database, we observed that by varying sensitivity from 60 to 75% the number of false alarms per image lies in the range 4-10, while accuracy is in the range 95.7-98.0%. When the second data set was used comparable results were obtained. The observed system performances support the undertaking of system validation in clinical settings.

Lung CT Densitometry in Systemic Sclerosis: Correlation With Lung Function, Exercise Testing, and Quality of Life

BACKGROUND To ascertain if analysis of lung density histograms in thin-section CT was more reproducible than visual assessment of lung changes in systemic sclerosis (SSc), and if such density histogram parameters as mean lung attenuation (MLA), skewness, and kurtosis could more closely reflect pulmonary function as well as exercise and quality of life impairment. METHODS The intraoperator and interoperator reproducibility of visual and densitometric lung CT analysis in 48 SSc patients examined with CT were evaluated by means of weighted kappa statistics. Univariate and multivariate regression analyses were applied to evaluate the relationship of visual and densitometric CT measurements with functional parameters including functional residual capacity (FRC), FVC, FEV(1), diffusion capacity of the lung for carbon monoxide (Dlco), 6-min walking testing (6MWT), and health-related quality of life questionnaire (QLQ) parameters. RESULTS The intraoperator and interoperator reproducibility of MLA (intraobserver weighted kappa = 0.97; interobserver weighted kappa = 0.96), skewness (intraobserver weighted kappa = 0.89; interobserver weighted kappa = 0.88), and kurtosis (intraobserver weighted kappa = 0.89; interobserver weighted kappa = 0.88) were higher than those of visual assessment (intraobserver weighted kappa = 0.71; interobserver weighted kappa = 0.69). In univariate analysis, only densitometric measurements were correlated with some exercise and QLQ parameters. In multivariate analysis, MLA (square regression coefficient corrected [R(2)c] = 0.70), skewness (R(2)c = 0.78), and kurtosis (R(2)c = 0.77) were predicted by FRC, FVC, Dlco, 6MWT, and QLQ parameters, while visual assessment was associated only with FRC and FVC (R(2)c = 0.40). CONCLUSIONS In SSc, densitometric analysis is more reproducible than visual assessment of lung changes in thin-section CT and more closely correlated to pulmonary function testing, 6MWT, and QLQ. Density histogram parameters may be useful for cross-sectional and longitudinal studies of lung involvement in SSc.

Whole-Brain Histogram and Voxel-Based Analyses of Diffusion Tensor Imaging in Patients with Leukoaraiosis: Correlation with Motor and Cognitive Impairment

BACKGROUND AND PURPOSE: Cerebral white matter changes, termed leukoaraiosis (LA), appearing as areas of increased signal intensity in T2-weighted MR images, are common in elderly subjects, but the possible correlation of LA with cognitive or motor deficit has not been established. We hypothesized that histogram and voxel-based analyses of whole-brain mean diffusivity (MD) and fractional anisotropy (FA) maps calculated from diffusion tensor imaging (DTI) could be more sensitive tools than visual scales to investigate the clinical correlates of LA. MATERIALS AND METHODS: Thirty-six patients of the Leukoaraiosis and Disability Study were evaluated with fluid-attenuated inversion recovery for LA extension, T1-weighted images for volume, and DTI for MD and FA. The extent of LA was rated visually. The normalized total, gray, and white matter brain volumes were computed, as well as the 25th percentile, 50th percentile, kurtosis, and skewness of the MD and FA maps of the whole brain. Finally, voxel-based analysis on the maps of gray and white matter volume, MD, and FA was performed with SPM2 software. Correlation analyses between visual or computerized data and motor or neuropsychologic scale scores were performed using the Spearman rank test and the SPM2 software. RESULTS: The visual score correlated with some MD and FA histogram metrics (P < .01). However, only the 25th and 50th percentiles, kurtosis, and skewness of the MD and FA histograms correlated with motor or neuropsychologic deficits. Voxel-based analysis revealed a correlation (P < .05 corrected for multiple comparisons) between a large cluster of increased MD in the corpus callosum and pericallosal white matter and motor deficit. CONCLUSIONS: These results are consistent with the hypothesis that histogram and voxel-based analyses of the whole-brain MD and FA maps are more sensitive tools than the visual evaluation for clinical correlation in patients with LA.

3-D Segmentation Algorithm of Small Lung Nodules in Spiral CT Images

Computed tomography (CT) is the most sensitive imaging technique for detecting lung nodules, and is now being evaluated as a screening tool for lung cancer in several large samples studies all over the world. In this report, we describe a semiautomatic method for 3-D segmentation of lung nodules in CT images for subsequent volume assessment. The distinguishing features of our algorithm are the following. 1) The user interaction process. It allows the introduction of the knowledge of the expert in a simple and reproducible manner. 2) The adoption of the geodesic distance in a multithreshold image representation. It allows the definition of a fusion--segregation process based on both gray-level similarity and objects shape. The algorithm was validated on low-dose CT scans of small nodule phantoms (mean diameter 5.3-11 mm) and in vivo lung nodules (mean diameter 5--9.8 mm) detected in the Italung-CT screening program for lung cancer. A further test on small lung nodules of Lung Image Database Consortium (LIDC) first data set was also performed. We observed a RMS error less than 6.6% in phantoms, and the correct outlining of the nodule contour was obtained in 82/95 lung nodules of Italung-CT and in 10/12 lung nodules of LIDC first data set. The achieved results support the use of the proposed algorithm for volume measurements of lung nodules examined with low-dose CT scanning technique.

Structural and functional evaluation of cortical motor areas in Amyotrophic Lateral Sclerosis

The structural and functional data gathered with Magnetic Resonance Imaging (MRI) techniques about the brain cortical motor damage in Amyotrophic Lateral Sclerosis (ALS) are controversial. In fact some structural MRI studies showed foci of gray matter (GM) atrophy in the precentral gyrus, even in the early stage, while others did not. Most functional MRI (fMRI) studies in ALS reported hyperactivation of extra-primary motor cortices, while contradictory results were obtained on the activation of the primary motor cortex. We aimed to investigate the cortical motor circuitries in ALS patients by a combined structural and functional approach. Twenty patients with definite ALS and 16 healthy subjects underwent a structural examination with acquisition of a 3D T1-weighted sequence and fMRI examination during a maximal force handgrip task executed with the right-hand, the left-hand and with both hands simultaneously. The T1-weighted images were analyzed with Voxel-Based Morphometry (VBM) that showed several clusters of reduced cortical GM in ALS patients compared to controls including the pre and postcentral gyri, the superior, middle and inferior frontal gyri, the supplementary motor area, the superior and inferior parietal cortices and the temporal lobe, bilaterally but more extensive on the right side. In ALS patients a significant hypoactivation of the primary sensory motor cortex and frontal dorsal premotor areas as compared to controls was observed. The hypoactivated areas matched with foci of cortical atrophy demonstrated by VBM. The fMRI analysis also showed an enhanced activation in the ventral premotor frontal areas and in the parietal cortex pertaining to the fronto-parietal motor circuit which paralleled with disease progression rate and matched with cortical regions of atrophy. The hyperactivation of the fronto-parietal circuit was asymmetric and prevalent in the left hemisphere. VBM and fMRI identified structural and functional markers of an extended cortical damage within the motor circuit of ALS patients. The functional changes in non-primary motor cortices pertaining to fronto-parietal circuit suggest an over-recruitment of a pre-existing physiological sensory-motor network. However, the concomitant fronto-parietal cortical atrophy arises the possibility that such a hyper-activation reflects cortical hyper-excitability due to loss of inhibitory inter-neurons.

Automated Segmentation Refinement of Small Lung Nodules in CT Scans by Local Shape Analysis

One of the most important problems in the segmentation of lung nodules in CT imaging arises from possible attachments occurring between nodules and other lung structures, such as vessels or pleura. In this report, we address the problem of vessels attachments by proposing an automated correction method applied to an initial rough segmentation of the lung nodule. The method is based on a local shape analysis of the initial segmentation making use of 3-D geodesic distance map representations. The correction method has the advantage that it locally refines the nodule segmentation along recognized vessel attachments only, without modifying the nodule boundary elsewhere. The method was tested using a simple initial rough segmentation, obtained by a fixed image thresholding. The validation of the complete segmentation algorithm was carried out on small lung nodules, identified in the ITALUNG screening trial and on small nodules of the lung image database consortium (LIDC) dataset. In fully automated mode, 217/256 (84.8%) lung nodules of ITALUNG and 139/157 (88.5%) individual marks of lung nodules of LIDC were correctly outlined and an excellent reproducibility was also observed. By using an additional interactive mode, based on a controlled manual interaction, 233/256 (91.0%) lung nodules of ITALUNG and 144/157 (91.7%) individual marks of lung nodules of LIDC were overall correctly segmented. The proposed correction method could also be usefully applied to any existent nodule segmentation algorithm for improving the segmentation quality of juxta-vascular nodules.

Progression of brain atrophy in the early stages of Parkinson's disease: A longitudinal tensor‐based morphometry study in de novo patients without cognitive impairment

The presence of brain atrophy and its progression in early Parkinsons disease (PD) are still a matter of debate, particularly in patients without cognitive impairment. The aim of this longitudinal study was to assess whether PD patients who remain cognitively intact develop progressive atrophic changes in the early stages of the disease. For this purpose, we employed high‐resolution T1‐weighted MR imaging to compare 22 drug‐naïve de novo PD patients without cognitive impairment to 17 age‐matched control subjects, both at baseline and at three‐year follow‐up. We used tensor‐based morphometry to explore the presence of atrophic changes at baseline and to compute yearly atrophy rates, after which we performed voxel‐wise group comparisons using threshold‐free cluster enhancement. At baseline, we did not observe significant differences in regional atrophy in PD patients with respect to control subjects. In contrast, PD patients showed significantly higher yearly atrophy rates in the prefrontal cortex, anterior cingulum, caudate nucleus, and thalamus when compared to control subjects. Our results indicate that even cognitively preserved PD patients show progressive cortical and subcortical atrophic changes in regions related to cognitive functions and that these changes are already detectable in the early stages of the disease. Hum Brain Mapp 35:3932–3944, 2014.

ACCURACY AND REPRODUCIBILITY OF A NOVEL DUAL-BEAM VECTOR DOPPLER METHOD

Conventional Doppler ultrasound (US) investigations are limited to detect only the axial component of the blood velocity vector. A novel dual-beam method has been recently proposed in which the Doppler angle is estimated through a reference US beam, and the velocity magnitude through a measuring US beam, respectively. In this study, the performance of such a method has been assessed quantitatively through in vitro and in vivo measurements made in different experimental conditions. In vitro, more than 300 acquisitions were completed using seven transducers to insonify a straight tube phantom at different Doppler angles. In steady laminar flow conditions, the velocity magnitude was measured with mean error of -1.9% (95% confidence interval: -2.33% to -1.47%) and standard deviation of 3.4%, with respect to a reference velocity. In pulsatile flow conditions, reproducibility tests of the entire velocity waveforms provided an average coefficient of variation (CV) of 6.9%. For peak velocity measurements made at five Doppler angles and three flow rates, the intrasession and intersession CVs were in the range 0.8-3.7% and 2.9-10.6%, respectively. The peak systolic velocities (PSVs) in the common carotid arteries of 21 volunteers were estimated with 95% limits of agreement of +/- 9.6 cm/s (intersession). This analysis shows that the proposed dual-beam method is capable of overcoming the Doppler angle ambiguity by producing reliable velocity measurements over a large set of experimental conditions.

Decreased and increased cortical activation coexist in de novo Parkinson's disease

Previous fMRI studies using motor tasks yielded conflicting results concerning the activation pattern in Parkinsons disease (PD) patients. Possible explanations of these discrepancies include differences in the clinical features of the examined patients and in the executed tasks and incomplete task monitoring. We evaluated with fMRI 20 patients with untreated de-novo PD and 11 healthy controls with a simple motor task consisting of self-paced continuous right hand-tapping. The task was monitored on-line with a dedicated device which measures the strength and frequency of the tapping. Fifteen patients performed the task correctly. The frequency was not significantly different, whereas force was slightly different between patients (26.4+/-3.0 N) and controls (28.5+/-2.4 N) (p=0.046, Mann-Whitney U-test). After insertion of the subjects frequency and force as covariate variables in the model, PD patients compared to controls showed areas of significantly [Z statistic image>5.1 and p< or =0.05 (corrected) cluster significance] lower activation in the left primary sensorimotor (SM1) cortex and cerebellum and higher activation in the left temporal-parietal cortex adjacent to the SM1 and in right SM1. Furthermore in PD patients the disease severity evaluated with the Hoehn and Yahr staging system correlated significantly [Z statistic image>2.3 and p< or =0.05 (corrected) cluster significance] with activation of left SM1 and supplementary motor area and cingulum, bilaterally. The mixed pattern of decreased and increased cortical activation in de novo PD patients possibly reflects the coexistence of cortical deafferentation, and compensatory phenomena by cortico-cortical circuits.

Combining functional and structural brain magnetic resonance imaging in Huntington disease.

Objective: To concurrently investigate with magnetic resonance (MR) the brain activation and regional brain atrophy in patients with Huntington disease (HD). Methods: Nine symptomatic HD patients and 11 healthy subjects underwent an MR study including functional MR acquisition during finger tapping of the right hand and high-resolution T1-weighted images. Functional and structural data were analyzed using Statistical Parametric Mapping 2 software. Results: As compared with control subjects, HD patients showed decreased activation in the left caudate nucleus and medial frontal and anterior cingulate gyri and increased activation in the right supplementary motor area and supramarginal gyrus and left intraparietal sulcus. The pattern of atrophy included thinning of the gray matter (GM) in the insula, inferior frontal gyrus, caudate, lentiform nucleus, and thalamus, bilaterally, in the left middle frontal, middle occipital, and middle temporal gyri, and of periventricular, subinsular, right temporal lobe, and left internal capsule white matter. Only the decreased activation in the caudate nucleus correlated topographically with the caudate GM loss. Conclusion: The cortical areas of functional changes do not correspond to those of GM atrophy in patients with HD and are likely to reflect decreased output of the motor basal ganglia-thalamo-cortical circuit and compensatory recruitment of accessory motor pathways.