Search

In radiotherapy, the dose and volumes of the irradiated normal tissues is correlated to the complication rate. We assessed the performances of low-energy proton therapy (ocular PT) with eye-dedicated equipment, high energy PT with pencil-beam scanning (PBS) or CyberKnifeCT-based comparative dose distribution between external beam radiotherapy techniques was assessed using an anthropomorphic head phantom. The prescribed dose was 60Gy_RBE in 4 fractions to a typical posterior pole uveal melanoma. Clinically relevant structures were delineated, and doses were calculated using radiotherapy treatment planning softwares and measured using Gafchromic dosimetry films inserted at the ocular level.Precision was significantly better with ocular PT than both PBS or SBRT in terms of beam penumbra (80%-20%: laterally 1.4 vs. ≥10mm, distally 0.8 vs. ≥2.5mm). Ocular PT duration was shorter, allowing eye gating and lid sparing more easily. Tumor was excellent with all modalities, but ocular PT resulted in more homogenous and conformal dose compared to PBS or SBRT. The maximal dose to ocular/orbital structures at risk was smaller and often null with ocular PT compared to other modalities. Mean dose to ocular/orbital structures was also lower with ocular PT. Structures like the lids and lacrimal punctum could be preserved with ocular PT using gaze orientation and lid retractors, which is easier to implement clinically than with the other modalities. The dose to distant organs was null with ocular PT and PBS, in contrast to SBRT.ocular PT showed significantly improved beam penumbra, shorter treatment delivery time, better dose homogeneity, and reduced maximal/mean doses to critical ocular structures compared with other current external beam radiation modalities. Similar comparisons may be warranted for other tumor presentations.

The decision to irradiate during pregnancy is based on a risk benefit compromise of two kinds: maternal risk and fetal risk. The aim of this work is to determine the foetal risk, and uterine dose measurement in proton therapy. Foetal exposure during treatment is linked to two sources: the treatment phase, and the repositioning phase. An Alderson-Rando anthropomorphic ghost (170cm, 74kg) was positioned on the table in the treatment position. A tissue-equivalent proportional counter (TEPC), adapted to the analysis of complex radiation fields (neutron and photonics), was used to determine the irradiation related to the treatment phase. An AT1123 radiation survey meter was used to measure photons generated by X-ray radiation. I dosimetry was proposed using radio-photoluminescent dosimeters, allowing for a daily check of the dose received in the uterus. The treatment phase produces higher uterine doses than the positioning phase, but these remain very low. The equivalent dose received in the uterus for the entire treatment is estimated at 840 μSv. Using a methodology for measuring the out-of-field dose with pencil beam scanning proton therapy, the foetal dose in the first trimester was well below the acceptance dose of 100 mGy determined by the International Commission on Radiological Protection.

Purpose: Secondary particles produced in the collision of protons with beam modifiers are of concern in proton therapy. Nevertheless, secondary radiation can provide information on the dosimetric parameters through its dependency on the modulating accessories (range shifter and range modulating wheel). Relatively little data have been reported in the literature for low-energy proton beams. The present study aims at characterizing the neutron and photon secondary radiation at the low-energy proton therapy facility of the Centre Antoine Lacassagne (CAL), and studying their correlation to the dosimetric parameters to explore possible practical uses of secondary radiation in the treatment quality for proton therapy. Methods: The Monte Carlo code MCNPX was used to simulate the proton therapy facility at CAL. Neutron and photon fluence, Φ, and ambient dose equivalent per protondose, H*(10)/D, were determined across the horizontal main plane spanning the whole treatment room. H*(10)/D was also calculated at two positions of the treatment room where dosimetric measurements were performed for validation of the Monte Carlo calculations. Calculations and measurements were extended to 100 clinical spread-out Bragg Peaks (SOBPs) covering the whole range of therapeutic dose rates (D/MU) employed at CAL. In addition, the values of D and MU were also calculated for each SOBP and the results analyzed to study the relationship between secondary radiation and dosimetric parameters. Results: The largest production of the secondary particles takes place at the modulating devices and the brasscollimators located along the optical bench. Along the beam line and off the beam axis to 2.5 m away, H*(10)/D values ranged from 5.4μSv/Gy to 5.3 mSv/Gy for neutrons, and were 1 order of magnitude lower for photons. H*(10)/D varied greatly with the distance and angle to the beam axis. A variation of a factor of 5 was found for the different range of modulations (SOBPs). The ratios between calculations and measurements were 2.3 and 0.5 for neutrons and photons, respectively, and remained constant for all the range of SOBPs studied, which provided validation for the Monte Carlo calculations. H*(10)/D values were found to correlate to the protondose rate D/MU with a power fit, both for neutrons and photons. This result was exploited to implement a system to obtain D/MU values from the measurement of the integrated photon ambient dose equivalent H*(10) during treatment, which provides a method to control the dosimetric parameters D/MU and D. Conclusions: The treatment room at CAL is moderately polluted by secondary particles. The constant ratio between measurements and calculations for all SOBPs showed that simulations correctly predict the dosimetric parameters and the dependence of the production of secondary particles on the modulation. The correlation between H*(10)/D and D/MU is a useful tool for quality control and is currently used at CAL. This system works as an indirectin vivodosimetry method, which is so far not feasible in proton therapy. This tool requires very simple instrumentation and can be implemented from the measurement of either photons or neutrons.

Resume Introduction Les carcinomes de la conjonctive sont des neoplasies rares a fort potentiel recidivant. L’obtention d’un bon controle tumoral necessite des traitements adjuvants tels que la cryotherapie, la chimiotherapie locale ou la radiotherapie. Nous presentons la technique et les resultats preliminaires de l’association chirurgie-protontherapie pour la prise en charge de ces carcinomes recidivants et/ou invasifs. Patients et methodes Depuis juin 2001 jusqu’a septembre 2008, quinze patients presentant des carcinomes invasifs ont eu une protontherapie adjuvante apres exerese chirurgicale. Un dispositif specifique de sculpture du faisceau d’irradiation permet une irradiation localisee du site tumoral en preservant au mieux les structures saines adjacentes. Resultats Le suivi moyen des patients etait de 39,1 mois. Selon la classification TNM, ils se repartissaient en 7 T4, 5 T3 et 3 T2 ; ils etaient tous N0M0. Treize patients ont ete controles sur le plan carcinologique. Deux patients ont presente des recidives. Ils presentaient des tumeurs recidivantes multi-operes de grande taille et la protontherapie avait ete realisee plus de 6 mois apres la prise en charge therapeutique initiale. Dans les effets secondaires, il a ete constate des syndromes secs, des cataractes, des dysplasies conjonctivales, des vaisseaux episcleraux dilates inesthetiques, des alopecies ciliaires, une stenose des voies lacrymales et un cas de glaucome unilateral. Conclusion Les traitements classiques ne permettent pas toujours le controle tumoral des carcinomes invasifs recidivants et le traitement final de ces neoplasies reste trop souvent l’enucleation, voire l’exenteration. La protontherapie, si son indication est posee precocement, semble etre un traitement adjuvant efficace qui expose a des effets secondaires acceptables.

Dependence of time-dependent dielectric breakdown with pulsewidth was investigated on MgO magnetic tunnel junctions. Barrier breakdown measurements were performed using different pulsewidths, from 10 ns up 1 s. The equivalence of the breakdown results between DC and pulsed voltage experiments was established using the total cumulated pulse time under electrical stress as the relevant parameter. Two different breakdown regimes were identified: the first one corresponding to pulses longer than 100 ns up to the DC limit and a second regime for pulses shorter than 100 ns. The barrier lifetime is increased for pulsewidths lower than 100 ns pulses. The breakdown process was shown to be thermally activated, but the longer lifetime at shorter pulses cannot be simply explained assuming a lower temperature during the pulse application.

Real time pattern recognition applications often deal with high dimensional data, which require a data reduction step which is only performed offline. However, this loses the possibility of adaption to a changing environment. This is also true for other applications different from pattern recognition, like data visualization for input inspection. Only linear projections, like the principal component analysis, can work in real time by using iterative algorithms while all known nonlinear techniques cannot be implemented in such a way and actually always work on the whole database at each epoch. Among these nonlinear tools, the Curvilinear Component Analysis (CCA), which is a non-convex technique based on the preservation of the local distances into the lower dimensional space, plays an important role. This paper presents the online version of CCA. It inherits the same features of CCA, is adaptive in real time and tracks non-stationary high dimensional distributions. It is composed of neurons with two weights: one, pointing to the input space, quantizes the data distribution, and the other, pointing to the output space, represents the projection of the first weight. This on-line CCA has been conceived not only for the previously cited applications, but also as a basic tool for more complex supervised neural networks for modelling very complex high dimensional data. This algorithm is tested on 2-D and 3-D synthetic data and on an experimental database concerning the bearing faults of an electrical motor, with the goal of novelty (fault) detection.

To determine the relative biological effectiveness (RBE) for initial and delayed inactivation of cells by a modulated proton beam suitable for the treatment of tumours of the eye, within the spread-out Bragg peak and in its distal declining edge.Human tumour SCC25 cells were irradiated with the 65 MeV proton beam at the Cyclotron Medicyc in Nice. Perspex plates of different thickness were used to simulate five positions along the beam line: 2mm corresponding to the entrance beam; 15.6 and 25 mm in the spread-out Bragg peak; 27.2 and 27.8mm for the distal edge. At each position clonogenic survival of the irradiated cells and of their progeny were determined at various dose values. 60Co gamma-rays were used as reference radiation.RBE values evaluated at the survival level given by 2 Gy of gamma-rays increased with increasing depth from close to 1.0 at the proximal to about 1.2 at the distal part of the peak. Within the declining edge it reached the value of about 1.4 at 27.2 and about 2 at 27.8 mm. For the progeny of irradiated cells, the RBE value ranged from 1.0 to 1.1 within the spread-out Bragg peak and then increased up to a value of 2.0 at the last position. The dose-effect curves for the progeny always had a larger shoulder than for the irradiated progenitors, their alpha parameters being lower by a factor of about 4 and their beta parameters always being higher. The alpha/beta ratio was about 50 Gy for the progenitors and about 6 Gy for their progeny. The incidence of delayed effects increased with dose and with the depth within the beam.RBE values for the inactivation of cells irradiated in the spread-out Bragg peak are compatible with the value currently assumed in clinical applications. In the distal declining edge of the beam, the RBE values increased significantly to an extent that may be of concern when the region of the treatment volume is close to sensitive tissues. The yield of delayed reproductive cell death was significant at each position along the beam line.

La reponse des cellules tumorales humaines a l'irradiation in vitro par des particules de TEL eleve depend de plusieurs facteurs. Pour les particules chargees, la radiosensibilite augmente avec le TEL jusqu'a des valeurs de TEL voisines de 200 KeV/μm, puis diminue pour des valeurs de TEL superieures. En clinique, l'augmentation du TEL moyen de la particule chargee en profondeur est responsable d'une efficacite biologique relative (EBR) croissante. La probabilite de mortalite de la cellule par une particule traversant le noyau est proche de 1 pour les particules de TEL tres eleve comme l'Ar de 7 MeV par nucleon (TEL = 1500 KeV/μm); elle n'est que de 0,02 pour les ions d'O de 400 MeV (TEL = 20 KeV/μm). L'EBR est plus elevee pour les petites doses, la relation EBR/dose etant plus marquee dans les cellules ayant un rapport α/β plus faible apres exposition aux photons, selon le modele lineaire-quadratique. Les particules moins efficaces que les photons en terme de mortalite cellulaire moyenne induisent des degâts individuels importants, reveles par la production de micronoyaux (MN) multiples. Concernant le lien entre l'EBR et la radiosensibilite intrinseque, les donnees recentes utilisant a la fois la technique des colonies et la methode des MN, montrent que l'EBR est d'autant plus elevee que les cellules sont radioresistantes aux photons.

We present a new strategy called "curvilinear component analysis" (CCA) for dimensionality reduction and representation of multidimensional data sets. The principle of CCA is a self-organized neural network performing two tasks: vector quantization (VQ) of the submanifold in the data set (input space); and nonlinear projection (P) of these quantizing vectors toward an output space, providing a revealing unfolding of the submanifold. After learning, the network has the ability to continuously map any new point from one space into another: forward mapping of new points in the input space, or backward mapping of an arbitrary position in the output space.

Secondary particles produced in the collision of protons with beam modifiers are of concern in proton therapy. Nevertheless, secondary radiation can provide information on the dosimetric parameters through its dependency on the modulating accessories (range shifter and range modulating wheel). Relatively little data have been reported in the literature for low-energy proton beams. The present study aims at characterizing the neutron and photon secondary radiation at the low-energy proton therapy facility of the Centre Antoine Lacassagne (CAL), and studying their correlation to the dosimetric parameters to explore possible practical uses of secondary radiation in the treatment quality for proton therapy.The Monte Carlo code MCNPX was used to simulate the proton therapy facility at CAL. Neutron and photon fluence, Φ, and ambient dose equivalent per proton dose, H∗(10)∕D, were determined across the horizontal main plane spanning the whole treatment room. H∗(10)∕D was also calculated at two positions of the treatment room where dosimetric measurements were performed for validation of the Monte Carlo calculations. Calculations and measurements were extended to 100 clinical spread-out Bragg Peaks (SOBPs) covering the whole range of therapeutic dose rates (D∕MU) employed at CAL. In addition, the values of D and MU were also calculated for each SOBP and the results analyzed to study the relationship between secondary radiation and dosimetric parameters.The largest production of the secondary particles takes place at the modulating devices and the brass collimators located along the optical bench. Along the beam line and off the beam axis to 2.5 m away, H∗(10)∕D values ranged from 5.4 μSv∕Gy to 5.3 mSv∕Gy for neutrons, and were 1 order of magnitude lower for photons. H∗(10)∕D varied greatly with the distance and angle to the beam axis. A variation of a factor of 5 was found for the different range of modulations (SOBPs). The ratios between calculations and measurements were 2.3 and 0.5 for neutrons and photons, respectively, and remained constant for all the range of SOBPs studied, which provided validation for the Monte Carlo calculations. H∗(10)∕D values were found to correlate to the proton dose rate D∕MU with a power fit, both for neutrons and photons. This result was exploited to implement a system to obtain D∕MU values from the measurement of the integrated photon ambient dose equivalent H∗(10) during treatment, which provides a method to control the dosimetric parameters D∕MU and D.The treatment room at CAL is moderately polluted by secondary particles. The constant ratio between measurements and calculations for all SOBPs showed that simulations correctly predict the dosimetric parameters and the dependence of the production of secondary particles on the modulation. The correlation between H∗(10)∕D and D∕MU is a useful tool for quality control and is currently used at CAL. This system works as an indirect in vivo dosimetry method, which is so far not feasible in proton therapy. This tool requires very simple instrumentation and can be implemented from the measurement of either photons or neutrons.

International audience; In this work, exchange bias was used as a probe to characterise the temperature profile induced by the inelastic relaxation of electrons tunnelling across a MgO barrier. Thermally assisted magnetic random access memory (TA-MRAM) cells comprising a magnetic tunnel junction (MTJ) with a reference pinned layer and a FeMn exchange biased storage layer were used. The pinning direction of the ferromagnetic storage layer is reversed when heated above the blocking temperature of the antiferromagnetic layer (FeMn). The power density required to reach this blocking temperature in the FeMn layer depends on the current polarity, indicating that the heat source term associated with the current flowing through the barrier depends itself on the current direction in contrast to simple Joule heating. This effect is due to the mechanism of energy dissipation in tunnelling. The tunnelling itself is ballistic i.e., without dissipation. However, after tunnelling, the hot electrons very quickly relax to the Fermi energy thereby loosing their excess energy in the receiving electrode. Therefore, the heat is essentially generated on one side of the barrier so that the whole profile of temperature throughout the pillar depends on the current direction. Full 3D thermal simulations also confirmed the temperature profile asymmetry. The proper choice of heating current direction (i.e., voltage polarity applied to the MTJ) can yield a reduction of about 10% in the heating power density required to enable writing in thermally assisted MRAM cells.

Radiation therapy with positively charged particles implies that the Bragg peak be spread out to deliver a homogeneous dose to the tumour. The spread-out Bragg peak (SOBP) has a higher linear energy transfer (LET) than the entrance beam. In addition, there is an LET gradient from proximal to distal SOBP. The aim of this study is to find out whether these small LET variations lead to differences in radiation response. Human melanoma cells (CAL4) were exposed to 65 MeV proton beams produced by the cyclotron Medicyc at five different positions: 2 mm depth corresponding to the entrance, 15, 20, 25 and 26.8 mm depth corresponding to four different positions in the half-modulated SOBP. Survival curves were generated using the in vitro colony method and fitted with the linear-quadratic model. Survival differences were observed at high doses; they were statistically significant at a dose of 8 Gy. With respect to the entrance position (2 mm), the relative biological effectiveness (RBE) at 1% survival was 1.09, 1.12, 1.19 and 1.27 at 15, 20, 25 and 26.8 mm in the SOBP, respectively. Whereas RBE values in the SOBP greater than 1.0 relative to the entrance beam represent a small biological advantage to be added to the well-known physical advantage of high energy proton beams; the RBE gradient along the SOBP would imply that the distal end of the tumour would receive a higher biologically equivalent dose than the proximal end, despite a homogeneous physical dose, especially at the high doses per fraction given in ocular melanomas. Although the increase in effectiveness with depth is mild, it should be kept in mind during eye treatment planning, in case a critical target is present at the extreme end of the SOBP.

MRAM are already used in low density applications such as microcontrollers or for spatial/avionics applications where they benefit from their radiation hardness. Thermally assisted writing allows extending the downsize scalability of both FIMS and STT-MRAM below the 20nm node. Within a few years, MRAM can play an important role in the memory hierarchy of electronic circuits. They may be first used as DRAM replacement below the 20nm node and in a second step even closer the logic units (Fig.10).

In protontherapy, accessories are used in order to adapt the beam to the surface and to the depth of the target. They are positioned close to the patient in order to avoid perturbation effects related to proton scattering. The level of contamination of the beam caused by these accessories may be assessed by examining the dose maps in parallel planes to the beam incidence. The EBT2 radiochromic film is a suitable tool for this task, as it can be cut into small pieces and immersed in water. Prior to use the EBT2 film for dose measurements, its response when exposed to a proton beam must be analysed.The measurements were performed at the Centre Antoine Lacassagne, using the hospital-based MEDICYC isochronous cyclotron which provides 65 MeV protons. Monoenergetic as well as polyenergetic beams were used. Small pieces of EBT2 films were irradiated with proton beams in a small water phantom. Films were exposed at various angles close to the beam incidence and received doses ranging from 0.25 to 500 Gy. The optical density (OD) was studied as a function of angle, dose and linear energy transfer (LET).The effective atomic number of the active layer of the film is close to that of water which prevents disturbances of the measurement. However, the high density and the significant thickness of the Mylar substrate surrounding the active layer affect the use of the film in a parallel orientation to the beam. Therefore, the substrate layer may totally or partially slow down the protons. The measurement is then no longer representative to what happens in water. The measurement errors can be corrected by applying a tilt angle of at least 5° between the film and the beam. The dose analysis reveals that the green channel is the most sensitive in the dose range from 1 to 100 Gy. The OD is accurately described by a Weibull function of the dose with four free parameters. The Weibull function is valid for both monoenergetic and polyenergetic beams if the LET is limited to values below 15 MeV g(- 1) cm(2). When using a film orientation close to the beam incidence angle, increasing LET values are encountered throughout the film axis gradually with the protons slowing down in water. The EBT2 films show an underestimated response for higher LET values. The comparison of data from the present study to data obtained by other authors for EBT films allows modelling the underestimated response as a function of the LET. The definition improvement of the link between OD and LET requires to integrate more closely the beam energy characteristics.The EBT2 film is a suitable dosimeter for analysing dose plans in planes nearly parallel to the beam orientation by compensating the underestimated dose response due to LET.

Invasive squamous cell carcinomas are uncommon neoplasias with high recurrence and mortality rates. The improvement of tumoral control requires additional treatments such as cryotherapy, topical chemotherapy, and radiotherapy. We present the technique and preliminary results of associating treatment with surgery and proton beam therapy for recurrent and invasive squamous cell carcinomas.From June 2001 to September 2008, 15 patients were treated in our ocular oncologic center for squamous cell carcinomas either with recurrences or with invaded resection margins. The treatment combined new surgical resection with protontherapy. Specific improvements in proton beam therapy have been made at the Nice Cyclotron to adapt the treatment to conjunctival tumors. Proton beam carving consists in using a specific device to treat the thickness of the whole lesion site and the adjacent conjunctiva and to spare the surrounding healthy structures.Patients were staged according to the TNM classification of malignant tumors in T2: 3; T3: 5; T4: 7. Mean follow-up was 39.1 months (range, 6-90 months). Our 15 patients included 12 males and three females. Left eyes were involved in eight cases. The mean age at first consultation was 63.7 years (range, 46-80 years). In 13 cases (86.8%), the bulbar and limbic conjunctiva was involved, in five of these cases the cornea was invaded, and the anterior chamber was involved in one case. In one case, the tumor was located on bulbar conjunctiva near the caruncle (6.6%) and in one case in the fornix (6.6%). One patient died of another cancer after 48 months of follow-up. We obtained local tumor control for 13 patients (86.8%) and recurrences for two patients (13.2%). One of them has presented with cervical node metastases. These two patients who presented recurring and extensive tumors had had previous repeated surgeries in other centers. Moreover, proton beam therapy was performed more than 6 months after the initial treatment. Exenteration and enucleation had to be performed to treat these recurrences 6 and 24 months after proton beam therapy. The exentered patient has been lost to follow-up. No patients developed recurrences with additional proton beam therapy performed within 6 months after initial surgical resection. As for side effects, seven patients suffered from sicca syndrome, six needed cataract surgeries, three unesthetic dilatations of episclera vessels, two conjunctival postradiation dysplasias, two experienced eyelash loss, one stenosis of the lacrimal duct, and one glaucoma controlled by monotherapy. Conjunctiva and amniotic grafts had to be performed on one of the patients presenting with dysplasia. Due to the rarity and diversity of these cases, it is nearly impossible to carry out prospective and comparative studies.Traditional adjuvant treatments often failed to control recurring and invasive squamous cell carcinomas. We often ended up performing exenteration to control local recurrences. The preliminary results of the present study suggest that proton beam therapy may be considered as a good alternative to traditional treatments with acceptable side effects.

International audience; Hell in data analysis is paved (at least) with variability and noise. Is there some lost garden of Eden? Is there some way to approach it? In this paper, we take the example of Human visual perception and show how our visual system manages to process visual information in such a highly efficient way that it is able to categorize images or scenes within ranges of 100-150 ms, whatever the viewing conditions. In fact, before any high-level recognition task, the visual system unfolds a series of pre-processing stages to reduce image variability: * In the retina: a first adaptation process to global and local illuminant intensity and color and second one to local contrasts allow to extract information equally in the whole image. A spatio-temporal filter allows a spectral whitening of images so that all spatial frequencies are equally represented. * In the primary visual cortex: the estimation of the local power frequency spectrum allows a relative insensitivity to image translations. Sampling the frequency spectrum on a log-polar basis and by means of Log-Normal filters allows to easily process zooms and rotations and also to estimate local perspective. * In the cortical area V4, a further Fourier Transform of the log-polar spectrum provides insensitivity to zooms and rotations, as well as to perspective transformations. We can consider images as represented by high-dimensional vectors, the components of which are the image's energy extracted by each frequency filter. Taking advantage of all these preprocessing stages, data variability can be expected to be optimally reduced for comparison purposes. In the last part of this paper, we give an example of image categorization by means of CCA, a self-organizing neural network, which, after noise reduction, reduces dimension and unfolds the manifold where the data are embedded. It is shown that the network, taught by human subjects performing the same task, is able to improve its behavior by providing a better separation of classes.

An approach to microstrip antennas is developed that is based on a mixed transient-frequency analysis. The solution of transient integral equations by the method of moments allows one to compute the transient currents on the antenna. The Fourier transform of these transient currents used in asymptotic expressions for the far field of a Hertz dipole leads directly to harmonic parameters of radiation. The relative importance of spherical space waves and cylindrical surface waves and their behavior in terms of substrate parameters and frequency are discussed for some cases. >

The application of diamond to dosimetry is desirable because of its tissue equivalence, chemical inertness and small size, but this has not been commercially viable owing to the non-reproducible response of natural diamond. The chemical vapour deposition (CVD) of diamond permits controlled, reproducible and large-scale production of this material at potentially low cost. An investigation of some clinically relevant features like the depth-dose distribution as well as the absorbed dose profile, obtained using thermoluminescence (TL), is reported for several CVD diamond films. The TL characterisation presented here shows that CVD diamond films should be excellent TL-mode detectors in instances of radiotherapy and in vivo radiation dosimetry.

As soon as an image processing operator can be expressed as a linearly separable Boolean function involving a cell and its neighborhood, there is a way of straightforwardly deriving an equivalent cellular neural network (CNN) operation. An appropriate method had already been introduced for the robust design of uniformly initialized uncoupled CNN operators, and is now applied to the design of binary initialized and coupled CNN operators. A way of implementing in a unique operator two different Boolean functions conditioning the white-to-black and the black-to-white transitions, respectively, is also presented.

The problem of actual vision machines on VLSI is the implementation of a large set of linear filters selective to features such as edges, corners, orientations, motion. This procedure requires large arrays of filters with complex structures. In order to reduce the complexity, we propose to implement a simple basis of filters able to generate more complex filters such as oriented quadrature band-pass filters, quadrature wedge filters, velocity tuned filters. The current basis consists of asymmetrical filters implemented with cellular neural networks having only one layer and interactions with four neighbors. We show that this basis generates relevant filters for vision applications and show how the particular structure of the filters basis allows the direct implementation of spatio-temporal filters with a low additional cost.

SMART (Sparse Matrix Adaptive Recursive Transforms) is a new architecture supporting sparse matrix transforms. Though it is primarily designed for neural network simulations, the approach used is general enough to encompass other domains (e.g. signal processing). SMART efficiently implements algorithms that use a vectorial formalism. A simple sparse matrix compaction procedure is given which parallelizes computations thanks to dedicated data structures. Using this architecture, we design a vector processor that interfaces with a host machine which is thus relieved from intensive computational tasks. >

In this paper, we investigate the self-adaptive source separation problem for convolutively mixed signals. The proposed approach uses a recurrent structure adapted by a generic rule involving arbitrary separating functions. We first analyze the stability of this generic algorithm and we apply these results to some classical rules that were proposed in the literature but only partly analyzed. We then derive the expression of the asymptotic error variance achieved by this rule (for strictly causal mixtures). This enables us to determine the optimum separating functions that minimize this error variance. They are shown to be only related to the probability density functions of the sources. The performance improvement achieved by this approach is illustrated by simulations performed with real mixtures of speech signals.

Introduction The proton therapy facility at Centre Antoine Lacassagne in Nice uses the 65 MeV isochronous cyclotron MEDICYC to treat eye tumors. The passive scattering allows to deliver a ‘‘plateau” percentage depth dose (or Spread-Out-Bragg-Peak), in order to cover the target volume in depth. This modulation is given by a range shifter and a range modulator wheel, which are adapted to each patient treatment. At the end of the beam line, the patient collimator reduces the beam laterally to protect healthy tissues. Secondary particles are produced in the collision of protons with beam line elements: photons and neutrons, which are able to escape from the patient and can be used, in theory, for in vivo dosimetry and quality assurance measurements. The secondary particles production is closely linked through the modulation accessories. Materials and methods The aim of this study is to investigate the relationship between proton dose and secondary particle production. Monte Carlo calculations, using MCNPX code, and measurements were extended to 100 clinical Spread-Out-Bragg-Peak covering the whole range of therapeutic dose rates (D/MU). The measurements were performed for validation of Monte Carlo calculations. Results This work highlights the existence of a correlation function between proton dose and secondary particle production. Knowing this relation in one point of the tratment room, it is possible to develop a quality assurance system which controls the effective proton dose rate delivered to the patient. Conclusion This system has been installed and is now used routinely. It allows the delivered dose to be monitored independently of the beam monitor reading and offers a new approach to the conventional in vivo dosimetry in protontherapy.

The spatial distribution of elementary colour receptors in the mammalian retina [2,3], their spatial coupling and a simple model of the neural circuits of outer and inner plexiforme layers [4, 7, 9, 10, 11] can provide an interesting hypothese of the spatio-temporal processing of colour and luminance signals. Considering the spatial sampling scheme of colours in fovea and parafovea. and applying the spatio-temporal filters of retinal circuitry [1, 5], we derive the colour-opponancy phenomenon and its relation to the spatio-temporal properties of X ganglion cells as shown in [3, 6], and also the achromatic and spatio-temporal properties of Y cells [8]. All these properties are explained by means of signal processing in the spatio-temporal frequency domain, in conjunction with spatial sub-sampling of colour signals. As a consequence of colour-opponancy, it is easy to postulate a scheme for the colour-constancy phenomenon at a higher level.

Accurate beam-direction measurements ((0±3) mrad) have been performed on the 40 Ar projectile fragmentation on a Ni target, at 44 MeV/u, by using the telescopic mode on the double spectrometer LISE at GANIL. Sampling results on Al and Au targets are also presented. Velocity distributions are analysed for fragments of masses from 10 to 45. As expected from a study of the forward-angle physics, the presented beam-direction measurements (0°) and the grazing-angle ones (3°) exhibit important differences. The most probable velocities are found much closer to the beam one at 0° than at 3°. The velocity-integrated mass yields obtained in the beam direction exhibit a regular exponential rise from A ∼18 to A ∼40; they are found larger or much larger than at the grazing angle, at variance with the less populated light proton-rich fragments. Integrated yields are found dependent on these new results. A non-dissipative process has to be invoked in the beam direction. The role of diffraction is to be examined. No unique mechanism can explain the whole set of results now available.

Spin-electronics is a very rapidly expanding area of R&D which merges magnetism and electronics (Nobel Prize 2007). Since the discovery of giant magnetoresistance in 1988, several breakthroughs have further boosted this field (spin-valves 1990, tunnel magnetoresistance 1995, spin transfer 1996, voltage controlled magnetic properties 2004). The phenomenon of spin-transfer is particularly attractive both from fundamental and applied view points. It provides a new way to manipulate the magnetisation of magnetic nanostructures by a spin-polarised current. Spinelectronics has found applications in hard disk drives (1998) and more recently in non-volatile standalone memories (MRAM = Magnetic Random Access Memory). The spin-transfer phenomenon provides a new write scheme in MRAM yielding a much better scalability of these devices towards the 22 nm node. Furthermore, besides MRAMs, hybrid CMOS/magnetic technology can yield a totally new approach in the way electronic devices are designed. Most CMOS devices such as microprocessors are based on Von Neumann architecture in which logic and memories are separate components. The unique set of characteristics combined within magnetic tunnel junctions: cyclability, switching speed, scalability, makes it possible to conceive novel electronic systems in which logic and memory are intimately combined in non-volatile logic components (concept of non-volatile CPU). The spin-transfer phenomenon can also be used to generate large amplitude steady magnetic excitations in magnetic nanostructures. This phenomenon can also be used to conceive new frequency tunable radiofrequency oscillators. The latter are very attractive for wireless applications which require scanning large frequency bandwidth.

The magnetic switching of the exchange biased storage layer in thermally assisted magnetic random access memory cells has been studied in the nanosecond time domain. Under reversed static external field, the magnetic tunnel junctions (MTJs) were subjected to current heating pulses long enough to heat the structure above the blocking temperature of the antiferromagnetic layer. The magnetic response of the storage layer was characterized by single-shot real-time measurement of MTJ resistance. The switching of the storage layer exhibits stochastic fluctuations. Nevertheless, using a heating current density of 4.7×106 A/cm2 corresponding to a bias voltage of 1.8 V, the switching takes place in less than 4 ns under 5 mT. Interestingly, the probability of switching versus pulse duration exhibits characteristic periodic steps which are ascribed to a combined effect of the applied field and spin transfer produced by the heating current pulses.

The stopping powers of eleven gaseous media have been measured for 20–85 MeV/u O, 25–60 MeV/u Ar, 20–40 MeV/u Kr and 25–30 MeV/u Xe ions. The experimental method consists of measuring the energy loss of ions traversing a given thickness of gas contained in a gas cell. The energy measurements are performed using a magnetic spectrometer (LISE). The pressure and temperature of the gases are directly measured by appropriate probes. The stopping powers are thus determined with an accuracy of ± 2% to ± 4%. From the comparison of these data with the stopping powers of solids previously measured, the magnitude of the gas-solid effect is estimated. This effect is still observed for 25 MeV/u Xe and vanishes for lighter projectiles at higher energy, when they tend to be fully stripped.

Connectionnist algorithms are known to be computationally expensive, for two main reasons. Firstly, Artificial Neural Networks (ANN) are made up of a large number of elements, thus neural network architecture is linked to the concept of massive parallelism. Secondly, due to the iterative algorithms, a large number of processing steps is often necessary in order to ensure convergence and stability in the network. Therefore, hardware supporting these algorithms must be developed together with a well-suited software environment. General-purpose “neuro-emulators” must be used for algorithm explorations and simulations in application-driven implementations. Software packages must allow, in a convivial and interactive way, the description and monitoring (of the evolution) of the simulated neural network. This software layer is fundamental; comparable to the role electronic C.A.D. software plays in microelectronic development.

The stopping powers of ten solid media (nine elements from Be to Au and mylar) have been measured for 25–45 MeV/u Kr and 25 MeV/u Mo and Xe ions. The accuracy in stopping power determinations is ± 1–2%. In addition, the charge state distributions at equilibrium of Kr, Mo, Xe ions exiting Be, Al and Au targets have been determined. Average charge states Q have been deduced. For Kr ions exiting any target, a semi-empirical expression has been obtained to predict the mean charge Q in a large energy domain (4–50 MeVu). The stopping power data have been compared with a scaling relative to He2+ ion stopping powers, assuming fully stripped ions inside the degraders. Effective charge parameters γ have been derived and compared with the ionization degree Q /Z 1 . Information concerning the emission of Auger electrons by ions exiting targets has been obtained. The stopping power data have also been compared with values from Ziegler (1980). Discrepancies are observed for light (Be) and heavy media (Ta-Au). A comparison has been made with the very recent semi-empirical calculations from Hubert et al. (1989). Excellent agreement is found with experimental data for all projectiles and media.

Neutron captures on 10 B may be a neat way to provide some selectivity in a fast neutron beam: when the neutrons are thermalized in the tissue, they produce nuclear reaction with 10 B nuclei, releasing energetic α particles and Li ions. Provided that a sufficient amount of 10 B may be carried with a specific molecule inside tumoral cells, a very efficient component of irradiation can be added. Tumors like glioblastomas, soft tissues sarcomas or melanomas, which are highly radioresistant, may benefit of this potentiation to obtain better local control, while sparing more the normal tissues. Results of tumor targeting with boronated compounds, of dosimetry and radiobiology in our fast neutron beam will be presented.

An associative memory circuit that may let designers expand neural networks around a matrix of analog synapses is described. The architecture of the chip and its basic cell are discussed, and some SPICE simulation results are presented and compared with measures provided by the first prototype. In particular, the linearity and dynamic response of the complete chip, which includes an array of 25 synapses and two address decoders used for programming the weights, are examined. >

The stopping powers of 11 solid media have been measured for 20–95 MeV/u oxygen ions from the french GANIL accelerator. The energy measurements were made using the LISE magnetic spectrometer. The accuracy in stopping power determinations is ± 1–2%. The data are in good agreement with semi-empirical values from Hubert et al. [Ann. Phys. (Paris) 5, supp. (1980)] and from Ziegler (Pergamon, New York, 1980) and with a scaling relative to the He2+ ion stopping powers, assuming fully stripped ions inside the degrader.

Starting from the properties of a neural network with backward lateral inhibitions, we define a new algorithm suitable for non-linear image processing. From a convenient choice of the network parameters we obtain, after an initial smoothing effect, either a sharpening effect, or a spatial sampling adapted to the singularities of the input pattern.

An original experimental method has been developed using the LISE magnetic spectrometer at the GANIL accelerator for measuring the angular distributions of heavy ions at the exit of targets. Multiple angular scattering of heavy ions (16,17O, 40Ar, 86Kr and 100Mo) in solid and gaseous media have thus been measured in the energy range 20–90 MeV/u. With these different projectile-target systems and incident energies the Born parameter αB takes values in a wide range (0.5–100). The experimental results have thus been compared to multiple scattering theories based on classical or quantum mechanical calculations. As far as the classical theory is concerned a “universal” relation links the reduced straggling angle \ ga 1 2 and the reduced thickness τ intr Meyer according to the expression \ ∼ ga 1 2 = 1.00 τ 0.55 . In the framework of the quantal Moliere theory a good agreement is fou experiment when αB is approximately unity. Increasing deviations are observed when this parameter increases up to 100.

Part I. Starting from the properties of networks with backward lateral inhibitions, we define an algorithm for adaptive spatial sampling of line‐structured images. Applications to character recognition are straightforward.Part II. Let be an array of n sensors, each sensitive to an unknown linear combination of n sources. This is a classical problem in Signal Processing. But what is less classical is to extract each source signal without any knowledge either about those signals or about their combination in the sensors outputs. The only assumption is that the sources are independent.This problem emerged from recent studies on neural networks where any message appears as an unknown mixing of primary entities which are to be ‘‘discovered’’. According to the model of neural networks, we propose an algorithm based on: i − a network of fully interconnected processors (like neurons in a small volume of the Central Nervous System). ii − A law which controls the weights of the interconnections, derived from the Hebb concept for ‘‘Synaptic plasticity’’ in Physiology, and very close to the well know ‘‘stochastic gradient algorithm’’ in Signal Processing.This asociation result in a permanent selfearning mechanism which leads to a continuously up‐dating model of the sensor array information structure.After convergence, the algorithm provides output signals directly proportional to the independent primitive source signals.

The stopping powers of 12 solids and 10 gases and the associated energy loss straggling have been measured at GANIL for 16-17O, 40Ar,40Ca, 84-86Kr, 100Mo and 129-132Xe ions in the energy domain 24 to 35 MeV/u. After a brief description of the experimental method, the set of data are presented and discussed. As far as stopping powers are concerned, a scaling law in Z2 relative to the 4He stopping powers at the same velocity becomes applicable when the projectile tends to be fully stripped. At the same time, the solid-gas effect, observed at low energy, vanishes. The energy loss straggling is compared to that deduced from Bohr model and some of its refinements. The theoretical value appears to be too small, by a factor 2, when the projectile is fully stripped. For partially stripped ions, the charge exchange straggling might explain the experimental observation.