Automatic speech analysis to early detect functional cognitive decline in elderly population.

This study aimed at evaluating whether people with a normal cognitive function can be discriminated from subjects with a mild impairment of cognitive function based on a set of acoustic features derived from spontaneous speech. Voice recordings from 90 Italian subjects (age >65 years; group 1: 47 subjects with MMSE>26; group 2: 43 subjects with 20≤ MMSE ≤26) were collected. Voice samples were processed using a MATLAB-based custom software to derive a broad set of known acoustic features. Linear mixed model analyses were performed to select the features able to significantly distinguish between groups. The selected features (% of unvoiced segments, duration of unvoiced segments, % of voice breaks, speech rate, and duration of syllables), alone or in addition to age and years of education, were used to build a learning-based classifier. The leave-one-out cross validation was used for testing and the classifier accuracy was computed. When the voice features were used alone, an overall classification accuracy of 0.73 was achieved. When age and years of education were additionally used, the overall accuracy increased up to 0.80. These performances were lower than the accuracy of 0.86 found in a recent study. However, in that study the classification was based on several tasks, including more cognitive demanding tasks. Our results are encouraging because acoustic features, derived for the first time only from an ecologic continuous speech task, were able to discriminate people with a normal cognitive function from people with a mild cognitive decline. This study poses the basis for the development of a mobile application performing automatic voice analysis on-the-fly during phone calls, which might potentially support the detection of early signs of functional cognitive decline.

Duckneglect: video-games based neglect rehabilitation.

BACKGROUND: Video-games are becoming a common tool to guide patients through rehabilitation because of their power of motivating and engaging their users. Video-games may also be integrated into an infrastructure that allows patients, discharged from the hospital, to continue intensive rehabilitation at home under remote monitoring by the hospital itself, as suggested by the recently funded Rewire project. OBJECTIVE: Goal of this work is to describe a novel low cost platform, based on video-games, targeted to neglect rehabilitation. METHODS: The patient is guided to explore his neglected hemispace by a set of specifically designed games that ask him to reach targets, with an increasing level of difficulties. Visual and auditory cues helped the patient in the task and are progressively removed. A controlled randomization of scenarios, targets and distractors, a balanced reward system and music played in the background, all contribute to make rehabilitation more attractive, thus enabling intensive prolonged treatment. RESULTS: Results from our first patient, who underwent rehabilitation for half an hour, for five days a week for one month, showed on one side a very positive attitude of the patient towards the platform for the whole period, on the other side a significant improvement was obtained. Importantly, this amelioration was confirmed at a follow up evaluation five months after the last rehabilitation session and generalized to everyday life activities. CONCLUSIONS: Such a system could well be integrated into a home based rehabilitation system.

Using virtual reality to rehabilitate neglect.

PURPOSE: Virtual Reality (VR) platforms gained a lot of attention in the rehabilitation field due to their ability to engage patients and the opportunity they offer to use real world scenarios. As neglect is characterized by an impairment in exploring space that greatly affects daily living, VR could be a powerful tool compared to classical paper and pencil tasks and computer training. Nevertheless, available platforms are costly and obstructive. Here we describe a low cost platform for neglect rehabilitation, that using consumer equipments allows the patient to train at home in an intensive fashion. METHOD: We tested the platform on IB, a chronic neglect patient, who did not benefit from classical rehabilitation. RESULTS: Our results show that IB improved both in terms of neglect and attention. Importantly, these ameliorations lasted at a follow up evaluation 5 months after the last treatment session and generalized to everyday life activities. CONCLUSIONS: VR platforms built using equipment technology and following theoretical principles on brain functioning may induce greater ameliorations in visuo-spatial deficits than classical paradigms possibly thanks to the real world scenarios in association with the "visual feedback" of the patient's own body operating in the virtual environment.

Bayesian denoising in digital radiography: a comparison in the dental field.

We compared two Bayesian denoising algorithms for digital radiographs, based on Total Variation regularization and wavelet decomposition. The comparison was performed on simulated radiographs with different photon counts and frequency content and on real dental radiographs. Four different quality indices were considered to quantify the quality of the filtered radiographs. The experimental results suggested that Total Variation is more suited to preserve fine anatomical details, whereas wavelets produce images of higher quality at global scale; they also highlighted the need for more reliable image quality indices.

Automatic detection of neurons in large cortical slices.

The analysis of neuron distribution inside the cerebral cortex is getting more and more attention. It allows assessing, for instance, age-related and pathological decay and preferential connections; moreover, it complements well studies on functional morphology aimed to discovering information coding in neuron assemblies. A large obstacle to these studies is the huge amount of time required by an operator to manually mark the single neurons. We present here an innovative solution for automaticize the entire process: starting from a set of tile images of a given cortical slice, the system stitches all the tiles together, identifies the grey areas and cover them with a mesh. Neurons are automatically identified and their local distribution determined. Key element of the method is a reliable neuron identification algorithm based on a novel multilayer shape analysis of the blobs identified in the tiles images. This allows identifying on average 87+/-6% of the total neurons in the slice, with a false positive ratio of 14+/-9%, in a relatively short processing time. The algorithm was tested on Nissl-stained cortical slices of the BA4 Human area, 10 microm thick, acquired as a meander of tiles ( approximately 3000 images for a slice of medium size) at 40 x magnification, which gives a resolution of 0.264 microm/pixel. Preliminary results on cortical lamination of Human BA4 area are reported. This method is the first automated algorithm for the analysis of a large high-resolution cortical slice.

Compression and smart coding of offset and gain maps for intraoral digital x-ray sensors.

The response of indirect x-ray digital imaging sensors is often not homogenous on the entire surface area. In this case, calibration is needed to build offset and gain maps, which are used to correct the sensor output. The sensors of new generation are equipped with an on-board memory, which serves to store these maps. However, because of its limited dimension, the maps have to be compressed before saving them. This step is critical because of the extremely high compression rate required. The authors propose here a novel method to achieve such a high compression rate, without degrading the quality of the sensor output. It is based on quad tree decomposition, which performs an adaptive sampling of the offset and gain maps, matched with a RBF-based interpolation strategy. The method was tested on a typical intraoral radiographic sensor and compared with traditional compression techniques. Qualitative and quantitative results show that the method achieves a higher compression rate and produces images of superior quality. The method can be adopted also in different fields where a high compression rate is required.

Statistical based impulsive noise removal in digital radiography.

A new filter to restore radiographic images corrupted by impulsive noise is proposed. It is based on a switching scheme where all the pulses are first detected and then corrected through a median filter. The pulse detector is based on the hypothesis that the major contribution to image noise is given by the photon counting process, with some pixels corrupted by impulsive noise. Such statistics is described by an adequate mixture model. The filter is also able to reliably estimate the sensor gain. Its operation has been verified on both synthetic and real images; the experimental results demonstrate the superiority of the proposed approach in comparison with more traditional methods.

3D analysis of the body center of mass in rock climbing.

The purpose of the present study was to search for common patterns and for differences in climbing strategies in a group of recreational climbers. Twelve participants were involved in the study. Each participant climbed a simple indoor route consisting of a 3m horizontal shift followed by a 3m ascent for five times. Climbers could choose their own style, their preferred speed and holds. Their motion was recorded through motion capture based on passive markers. Results suggested that two main climbing strategies were used: the first preferring agility over force and the second preferring force over agility. We also found that our best climbers tried to minimize power during all trials.

A new real time filter for local exposure correction in panoramic radiographs.

A new real time filter for local exposure correction in panoramic radiographs is presented here. The filter, called PaRSEC, allows eliminating the exposure artifacts, mainly introduced by Automatic Exposure Control (AEC) systems. These artifacts reduce the image readability and its diagnostic utility. The PaRSEC filter operates a local exposure equalization, based on a reliable estimate of the column mean gray level. Qualitative and quantitative results are reported for typical panoramic radiographs. They show a complete removal of the artifacts. The method compares favorably with other classical methods targeted to exposure correction.

Enhancing digital cephalic radiography with mixture models and local gamma correction.

We present a new algorithm, called the soft-tissue filter, that can make both soft and bone tissue clearly visible in digital cephalic radiographies under a wide range of exposures. It uses a mixture model made up of two Gaussian distributions and one inverted lognormal distribution to analyze the image histogram. The image is clustered in three parts: background, soft tissue, and bone using this model. Improvement in the visibility of both structures is achieved through a local transformation based on gamma correction, stretching, and saturation, which is applied using different parameters for bone and soft-tissue pixels. A processing time of 1 s for 5 Mpixel images allows the filter to operate in real time. Although the default value of the filter parameters is adequate for most images, real-time operation allows adjustment to recover under- and overexposed images or to obtain the best quality subjectively. The filter was extensively clinically tested: quantitative and qualitative results are reported here.

A functional-anatomical model for lipreading.

Regional cerebral blood flow (rCBF) PET scans were used to study the physiological bases of lipreading, a natural skill of extracting language from mouth movements, which contributes to speech perception in everyday life. Viewing connected mouth movements that could not be lexically identified and that evoke perception of isolated speech sounds (nonlexical lipreading) was associated with bilateral activation of the auditory association cortex around Wernicke's area, of left dorsal premotor cortex, and left opercular-premotor division of the left inferior frontal gyrus (Broca's area). The supplementary motor area was active as well. These areas have all been implicated in phonological processing, speech and mouth motor planning, and execution. In addition, nonlexical lipreading also differentially activated visual motion areas. Lexical access through lipreading was associated with a similar pattern of activation and with additional foci in ventral- and dorsolateral prefrontal cortex bilaterally and in left inferior parietal cortex. Linear regression analysis of cerebral blood flow and proficiency for lexical lipreading further clarified the role of these areas in gaining access to language through lipreading. The results suggest cortical activation circuits for lipreading from action representations that may differentiate lexical access from nonlexical processes.

Distortion correction for x-ray image intensifiers: local unwarping polynomials and RBF neural networks.

In this paper we present two novel techniques, namely a local unwarping polynomial (LUP) and a hierarchical radial basis function (HRBF) network, to correct geometric distortions in XRII images. The two techniques have been implemented and compared, in terms of residual error measured at control and intermediate points, with local and global methods reported in the previous literature. In particular, LUP rests on a locally optimized 3rd degree polynomial applied within each quadrilateral cell on the rectilinear calibration grid of points. HRBF, based on a feed-forward neural network paradigm, is constituted by a set of hierarchical layers at increasing cut-off frequency, each characterized by a set of Gaussian functions. Extensive experiments have been performed both on simulated and real data. In simulation, we tested the effect of pincushion, sigmoidal and local distortions, along with the number of calibration points. Provided that a sufficient number of cells of the calibration grid is available, the obtained accuracy for both LUP and HRBF is comparable to or better than that of global polynomial technique. Tests on real data, carried out by using two different (12 in. and 16 in.) XRIIs, showed that the global polynomial accuracy (0.16+/-0.08 pixels) is slightly worse than that of LUP (0.07+/-0.05 pixels) and HRBF (0.08+/-0.04 pixels). The effects of the discontinuity at the border of the local areas and the decreased accuracy at intermediate points, typical of local techniques, have been proved to be smoothed for both LUP and HRBF.

Different brain correlates for watching real and virtual hand actions.

We investigated whether observation of actions reproduced in three-dimensional virtual reality would engage perceptual and visuomotor brain processes different from those induced by the observation of real hand actions. Participants were asked to passively observe grasping actions of geometrical objects made by a real hand or by hand reconstructions of different quality in 3D virtual reality as well as on a 2D TV screen. We found that only real actions in natural environment activated a visuospatial network including the right posterior parietal cortex. Observation of virtual-reality hand actions engaged prevalent visual perceptual processes within lateral and mesial occipital regions. Thus, only perception of actions in reality maps onto existing action representations, whereas virtual-reality conditions do not access the full motor knowledge available to the central nervous system.

A fast method for calibrating video-based motion analysers using only a rigid bar.

Video-camera systems are widely used in biomechanics and clinical fields to measure the 3D kinematic measurements of human motion. To be used, they need to be calibrated, that is the parameters which geometrically define the cameras have to be determined. It is shown here how this can be achieved by surveying a rigid bar in motion inside the working volume, and in a very short time: less than 15 s on a Pentium III. The exterior parameters are estimated through the coplanarity constraint, the camera focal lengths through the properties of epipolar geometry and the principal points with a fast evolutionary optimisation which guarantees convergence when the initial principal points cannot be adequately estimated. The method has been widely tested on simulated and real data. Results show that its accuracy is comparable with that obtained using methods based on points of known 3D coordinates (DLT): 0.37 mm RMS error over a volume with a diagonal approximately 1.5 m. A preferential absolute reference system is obtained from the same bar motion data and is used to guide an intelligent decimation of the data. Finally, the role that the principal points play in achieving a high accuracy, which is questioned in the computer vision domain, is assessed through simulations.

Complete calibration of a stereo photogrammetric system through control points of unknown coordinates.

This paper presents a new method for calibrating a video 3D stereo-photogrammetric system. The external parameters and the focal lengths of the cameras are determined from the epipolar constraint and the principal points are computed through the minimisation of a cost function carried out through an evolutionary optimisation. The method has been made more robust with a deterministic annealing procedure of the search region amplitude. Calibration is carried out by moving a rigid bar, carrying two markers on its extremities, inside the working volume. The distance between the two markers is the only measure required. Tests on real data are reported which show that the obtained accuracy is comparable to the one achieved calibrating with control points of known 3D coordinates.

Posture and movement: coordination and control.

Studies are reviewed that address the problem of the variables controlled in the maintenance of body posture and generation of limb movement. Vestibulospinal and neck reflexes cancel each other in response to roll, but not in response to pitch of the animal. In pitch trunk orientation is not effectively stabilized in space. Instead, limb length and orientation relative to the vertical are accurately controlled in normal cats pitched statically and dynamically by variable angles. Control of limb geometry may even take precedence over the control of the projected centre of mass. Coordinate transformation results in a constraint of planar covariation of the elevation angles at all limb segments in cat posture. Because the same constraint applies also to human locomotion, we suggest that sharing the same laws of intersegmental coordination for the control of posture and locomotion helps to assure the maintenance of dynamic equilibrium during movement. Moreover, because several neural sites encode posture and movement in gravity-based reference frames, alignment in register of spatial information derived from multiple sensors and directed to multiple effectors is made possible.

Statistical comparison of DLT versus ILSSC in the calibration of a photogrammetric stereo-system.

This paper compares the DLT and ILSSC approaches in the geometrical calibration of a photogrammetric stereo-system in terms of accuracy and speed. To come up with an unbiased quantitative evaluation of the accuracy of the algorithms, the concept of reliable estimate has been introduced: the statistical distribution of the accuracy is assessed over different calibration experiments performed with the same data but with different noise distribution and different test sets. Results show that in the simulations where the only error on the two-dimensional points was Gaussian, zero mean, and on real data which were corrected for distortions through polynomial or linear interpolation, the accuracy of the two methods was quite similar. DLT showed more accurate than ILSSC on simulated data with residual distortion errors and on real data which were not corrected for distortions. As far as speed is concerned, a fast triangulation algorithm is associated with ILSSC while the simultaneous solution of two pairs of DLT equations is associated to DLT. The first algorithm is much faster, requiring 113 flops per point versus 259 of DLT; the fast triangulation with DLT parameters does not achieve the same accuracy on the reconstructed three-dimensional position. Taken all together the results suggest that ILSSC can be theoretically considered the best approach to three-dimensional reconstruction, provided that distortions are corrected in advance. The statistical evaluation of the accuracy allows a fair judgement of the performances of the algorithms to be obtained, unbiased by particular distributions of measurement errors and test points.

A callosal disconnection syndrome of vascular origin.

OBJECTIVE: To study the after-effects of a selective lesion of the corpus callosum on the capacity in transferring spatial information on arm posture derived from kinesthesia and vision. DESIGN: Neuropsychological and psychophysical evaluations of a patient with callosal damage. SETTING: Institute of Neurology, University of "Tor Vergata", Rome, Italy and Research Centre, IRCCS "S. Lucia", Rome, Italy. PATIENT: P.V., a 59 years old man suffering from a vascular lesion of the left half of corpus callosum and left callosal radiations. MEASURES: Sensori-motor and cognitive tasks. Exo- and egocentric pointing abilities tasks. RESULTS: Impairments due to a disconnection between the two cerebral hemispheres were pointed out. About pointing abilities, errors were greater when each hand pointed to the remembered location of the other hand. Errors decreased dramatically with eyes open for right hand pointing to remembered location of left and right hand, and left hand pointing to remembered location of left hand. Open eyes left hand pointing to remembered location of right hand remained as poor as with eyes closed. CONCLUSIONS: The results appear compatible with the following network: visual information from striate and extrastriate cortex is passed forward to parietal cortex bilaterally (even in PV due to the splenium sparing). Kinesthetic information from parietal cortex of left and right hemisphere converge on a single visuokinesthetic center, lateralized to the left hemisphere. Information about limb position in three-dimensional space is then encoded in body-centered coordinates and passed forward to motor and premotor cortex in the frontal lobe. These frontal regions are those disconnected in PV.

Automatic integrated analysis of jaw and lip movement in speech production.

A noninvasive methodology for studying the kinematics of speech production is presented. It is based on the tracking of very small and light passive markers attached to the subjects' face. Using a pair of TV cameras, the 3-D markers' positions are computed in real time, at a subpixel accuracy, by a dedicated hardware. From these data, the time course of a set of parameters which describe lip and jaw movement is computed; in addition, a semiautomatic procedure that identifies the exact onset and offset of the investigated sequences has been developed. To compare the results over different productions, a time normalization procedure based on a continuous inverse Fourier transform has been implemented.