EP1861791A1

EP1861791A1 - Three-dimensional imaging method, installation implementing said method, method for configuring such an installation, computer programme implementing said method

Info

Publication number: EP1861791A1
Application number: EP06726088A
Authority: EP
Inventors: David Brasse; Bernard Humbert; Nicolas Rudolf; Benoit Speckel; Denis Staub
Original assignee: Centre National de la Recherche Scientifique CNRS; Universite Louis Pasteur Strasbourg I
Current assignee: Centre National de la Recherche Scientifique CNRS; Universite Louis Pasteur Strasbourg I
Priority date: 2005-03-15
Filing date: 2006-03-14
Publication date: 2007-12-05
Also published as: FR2883393B1; WO2006097621A1; US20110149025A1; FR2883393A1

Abstract

The invention concerns a method for three-dimensional imaging of a volume to be imaged executed by several computers connected in parallel: an emitting machine (13c) transmits to a network enabling transfer rates at least equal to 100 megabits per second (Mb/s), data delivered by an acquisition system corresponding to a set of projections of the volume, acquired in accordance with various incidences. The data are duplicated inside the network towards a plurality of processing machines (14a, , 14h), which receive each the received data. In each processing machine the correspondence between the received data and the transmitted data is ascertained. Each processing machine processes the data to reconstruct a fraction of the three-dimensional image of the volume to be imaged. At the end of the acquisition, and hence of the inline real-time reconstruction, a dedicated machine collects the set of fraction of images to reconstitute the three-dimensional image. The resulting improved systolic processing compared to the known multicast MPI (message passing interface) enables the speed, reliability and flexibility of the real-time tomographic restitution of moving small laboratory animals to be increased at a lower cost.

Description

THREE DIMENSIONAL IMAGING METHOD, INSTALLATION PROVIDING

A METHOD, A METHOD OF CONFIGURING SUCH INSTALLATION, A COMPUTER PROGRAM IMPLEMENTING THE SAME

Such a method

The present invention relates to three-dimensional imaging methods, to installations implementing such methods, to methods for configuring such installations, and to computer programs implementing these methods.

In particular, the invention relates to a three-dimensional imaging method from a plurality of images representative of a volume to be imaged taken at different incidences by a sensor, such as the projection of a volume - three-dimensional effect on an encoder.

In particular, two-dimensional images obtained by a planar sensor can be used.

There are two major categories of installation implementing these methods. A first category consists of a dedicated machine comprising a dedicated hardware, developed for each application. Such installations are not very flexible, in particular are too specific to the sensor and the algorithm used to allow to be used directly with another type of sensor, or another algorithm. Therefore, these facilities are very expensive, and in fact are intended for equipment that can bring a high profitability (medical imaging of the human being in particular).

A second major category consists of networked PC farms. These PCs operate in parallel. In this case, it is possible, for example, to provide that each PC operating in parallel receives from the acquisition device a subset of the projections, and independently reconstruct the contribution of this subset of the projections to the whole volume, then add together the volumes obtained by each PC and each corresponding to a subset of projections. Alternatively, it is possible to transmit to each PC all the projections, each PC being assigned to reconstruct a portion of the volume to be imaged, these portions being subsequently assembled. In these methods, the speed of calculation is directly related to the capacity of the network to quickly and reliably transmit a large amount of data, and the number of machines.

The increase in the speed of reconstruction necessarily involves an increase in the number of machines, which in turn inevitably translates into an increase in the throughput on the network. It is this criterion which then becomes the limiting factor of the reconstruction speed of the three-dimensional image.

The good management of information (in particular data coming from the sensor) then becomes a major factor in the improvement of existing three-dimensional image reconstruction methods.

Data dissemination techniques, which reduce the volume of data streams when broadcast to many receivers, are already used on the Internet, particularly for multimedia applications, and are grouped together under the name of "multicast". ".

Only the use of the UDP protocol allows the simultaneous broadcasting of data to a number of receivers. Nevertheless, the transport in non-connected mode carried out by this protocol does not make it possible to ensure that the transmitted data have been received in full by all the receivers. This defect is not really binding for multi-media applications, because it is not essential to receive all the data in this type of application.

To ensure the effective reception of all data to all receivers, the alternative is to use existing "multicast" methods for standard PC networks implementing the MPI function library (acronym for the English Message Passing Interface, ie "Message Transmission Interface"). These require several successive steps during which several point-to-point transmissions using the TCP / IP protocol are performed. Thus, in a first step, a transmitting machine sends the data to a first receiving machine. In a second step, the sending machine and the first receiving machine transmit the data to a new receiving machine each. This is done in succession, each receiving machine becoming in the next step a transmitting machine. The transmission of data to 5 machines thus requires three steps, to 9 machines, 4 steps, etc ..., i If such a transfer of information has advantages, which make it fit for its current use, for example in system of reconstruction without time constraint, it is not likely to be used to quickly obtain a three-dimensional image of a volume to be imaged using a plurality of processing machines. Indeed, for data flows so important, it is necessary to be able to provide the necessary data as soon as possible to all the processing machines. if we want to move towards real-time imaging. For example, US 2003 / 206,609, presenting a parallel computation implementing two machines, already describes "data bottlenecks".

To overcome these drawbacks, the invention proposes an imaging method, three-dimensional from one volume to imager (4), comprising at least:

(a) a transmission step in which at least one transmitting machine transmits on a network allowing transfer rates at least equal to 100 megabits per second (Mb / s), to a plurality of transmission machines; treatment, data transmitted corresponding to a projection of said volume, acquired according to an incidence,

(b) a receiving step in which each processing machine receives received data relating to the transmitted data, said received data being obtained, from the transmitted data, by duplication within the network to each processing machine,

(c) a control step during which ^'ensures for each processing machine, that. data correspond to the data transmitted, and

(d) a processing step in which each processing machine processes said received data to reconstruct a three-dimensional image of the volume to be imaged.

Thanks to these arrangements, the network ensures the almost simultaneous transfer of all the data to each processing machine, which can then process the information to reconstruct the three-dimensional image of the volume to be imaged. So we obtain a process that is both reliable and fast.

In preferred embodiments of the invention, one or more of the following provisions may also be used: in the course of step (a), the transmitting machine transmits the data issued in the form of data blocks each comprising a plurality of datagrams, and wherein during step (c), (cl) each processor transmits, to the machine ^• -émettrice, UNE .requête of _. ... reissue - missing datagrams for the data block,

(c2) the sending machine lists all the missing datagrams for the data block, and

(c3) the sending machine sends all the missing datagrams for the data block to all the machines of. treatment ; step (a) is repeated with other data corresponding to a projection of said volume, step (a) being repeated only after a positive result of said checking step; steps (a) to (d) are repeated for a plurality of data corresponding to a plurality of projections of said volume, each acquired at an incidence, said incidences being distinct two by two; during the receiving step, in each processing machine received data is received corresponding to at least a part of said transmitted data, and during the checking step, each processing machine ensures that the received data correspond to the data transmitted in the following manner: each processing machine estimates a quality of reception of the received data, and transmits towards an issuing control machine an acknowledgment of receipt relating to the quality of reception of said received data, and said control transmitter machine waits for said acknowledgment of receipt of each of the processing machines; - said data comprises a ^{predetermined} number of data datagrams, during the monitoring step each processing machine transmits said acknowledgment after receiving said predetermined number of datagrams (the predetermined number of datagrams may be configurable by the user); in each processing machine, said transmitted data correspond to said first received data and to non-received data, during the checking step, said processing machine issues a request for retransmission of the data not received, and during a retransmission step, at least said non-received data are re-transmitted on said network; the imaging method further comprises, before ^'the diffusion step, a step (z) call setup in which each processor sends to the network a request to be receiver of the broadcast of said transmitted data , said

^• transmitted data being duplicated within the network, during the broadcast step, to the processing machines that have issued a request to be a receiver of the broadcast of the transmitted data; the method further comprises an acquisition step (e), prior to the diffusion step, during which a sensor acquires raw data corresponding to said projection; the acquisition step is triggered according to a signal emitted by the volume to be imaged; the method further comprises a pre-processing step (f) between said acquisition step and said transmitting step, during which said transmitted data is generated by applying to the raw data a corrective preprocessing taking into account physical characteristics of the acquisition; during the processing step, each processing machine processes a respective processing volume corresponding to a fraction of the volume to be imaged; during the processing step, each processing machine retro-projects said data received in. said volume de-treatment to obtain a a three-dimensional, retro-projected image corresponding to said projection, and adding said three-dimensional image to an earlier partial result obtained from previous projections of the volume to be imaged; (Of course, when acquiring the first projection, the previous partial result to which the three-dimensional image is added is zero). the method further comprises an assembly step (g) in which a collection machine receives from said processing machines said respective processing volumes and assembles said processing volumes to reconstruct a three-dimensional image of the volume at _. imaged.

In another aspect, the invention relates to a computer program comprising program codes for implementing such a three-dimensional imaging method, when executed on at least one programmable machine.

According to another aspect, the invention relates to a three-dimensional imaging installation of a volume to be imaged, comprising at least:

at least one transmitting machine, adapted to transmit on a network supporting transfer rates at least equal to 100 megabits per second, to a plurality of data processing machines transmitted corresponding to a projection of said acquired volume according to an incidence, said processing machines, each adapted to receive received data relating to the transmitted data, said received data being obtained, from the transmitted data, by duplication within the network to each processing machine, said installation being adapted to make sure, for, each. processing, the data received correspond to the transmitted data, each processing machine being adapted to process said received data to reconstruct a three-dimensional image of the volume to be imaged.

In preferred embodiments of the present apparatus, one and / or other of the following provisions may further be provided: each transmitting machine is adapted to transmit the data transmitted in the form of data blocks

Each comprising a plurality of datagrams, wherein each processing machine is adapted to transmit, to the transmitting machine, a request for retransmission of the missing datagrams for the data block,

The transmitting machine is adapted to list all the missing datagrams for the received data block and to issue all the missing datagrams for the data block destined for the set of processing machines;

Each of said processing machines is adapted to perform a processing of said first data to reconstruct a three-dimensional image of a respective processing volume corresponding to a fraction of the volume to be imaged;

The installation further comprises a collection machine, intended to be connected to each processing machine through the network, said collection machine being adapted to receive said three-dimensional images of respective treatment volumes;

Each processing machine, and to assemble them to form a three-dimensional image of the volume to be imaged; the installation further comprises at least one pre-processing machine intended to be connected to at least one transmitting machine through the network, said

3.5.processing machine is adapted to apply a preprocessing function to a projection for generating data to be transmitted, and for transmitting said data to be transmitted to said transmitting machine; said pre-processing machine is adapted to receive a projection from a sensor at a sensor rate, and the network rate is selected based on said sensor rate; the installation further comprises a radiation source adapted to emit radiation in

10 direction of the volume to be imaged, and a sensor suitable for

^'Detecting said radiation after passing through the volume to be imaged, and generating said projection;

said machines are PC central units. ^•

According to another aspect, the invention relates to a method of configuring a group of machines in a three-dimensional imaging installation of a volume to be imaged, comprising at least:

at least one transmitting machine, adapted to transmit on a network supporting de | transfer rate at least equal to 100 megabits per second, to a plurality of processing machines, data transmitted corresponding to a projection of said acquired volume according to an incidence,

Said processing machines, each adapted to receive received data relating to the transmitted data, said received data being obtained, from the transmitted data, by duplication within the network for each processing machine, said installation being adapted to ensure that, for each processing machine, the received data correspond to the transmitted data, each processing machine being adapted for - -2ê-fe-r-ai-feeE-1-esd-it-es-data -Reç-ues to reconstruct- an image three-dimensional volume of the volume to be imaged, said configuration method comprising a configuration step in which each machine of said group of machines is assigned to a function selected from a group of functions comprising at least one transmission function, a processing function and a reception function.

In embodiments of this configuration method, it is furthermore possible to provide: the transmission function comprises a function for transmitting the data transmitted in the form of data blocks each comprising a plurality of datagrams, in which the function of reception comprises a transmission function, to the transmitting machine, of a request for retransmission of the missing datagrams for the data block, and in which the transmission function, also comprises a listing function of all the missing datagrams and a transmission function of all missing datagrams destined for all the processing machines; the function group further comprises a pre-processing function, and a collection and assembly function.

According to another aspect, the invention relates to a computer program comprising program codes for implementing such a configuration method, when it is executed on at least one programmable machine. ^'

In order to be able to apply the processing to the data, it may be necessary to pre-process them to take into account the specificities of the sensor and / or to filter detected parasitic components. In US 2003 / 206,609 and conventionally, the raw data acquired are stored in non-volatile memory (for example disk) before being subjected to these pre-treatments ... This step Intermediate also consumes time and resources, since it is then necessary to read the raw data in the non-volatile memory and transmit it to a pre-processing machine. Also the invention relates to an acquisition machine comprising a computer and a connection member to a sensor comprising an output delivering to said computer raw data representative of a projection detected by said sensor, said computer comprising: a daughter card d input / output which comprises a preprocessing processor programmed to apply preprocessing to said raw data output from said connection member, to generate pre-processed data, - a random access memory, and

a bus connecting said preprocessing processor and said random access memory, said computer being adapted to write said pre-processed data from the processor to said random access memory.

Such an acquisition machine makes it possible to apply

"On the fly" the pre-processing on the raw data at the output of the sensor, before writing the pre-processed data into RAM. As a result, resources that were previously dedicated to preprocessing data can then be used directly for reconstruction. For the same number of machines connected to the network as the above method, it is thus possible to maximize the number of machines connected in parallel for the rear projection.

According to preferred embodiments of this acquisition machine, it is also possible to provide one and / or the other of the following provisions: the calculator further comprises: a connection member to the outside, and a processor adapted to read in said random access memory said pre-processed data, to apply to them another pre-processing function for generating data to be transmitted and to write the data to be transmitted in the RAM of a transmitting machine. via said connecting member to the outside; said preprocessing processor is a reconfigurable processor adapted to be configured according to the sensor. Other features and advantages of the invention will become apparent from the following description of one of its embodiments, given by way of non-limiting example, with reference to the accompanying drawings.

In the drawings: FIG. 1 is a schematic view of an imaging installation,

FIG. 2 is an explanatory diagram of the rear projection of a pixel in a voxel,

FIG. 3 is an explanatory diagram of the configuration of the installation of FIG. 1 according to a first embodiment, and FIG. 4 is a detailed diagram of a second embodiment of the preprocessing part of FIG. 3. In these different figures, the same references denote identical or similar elements.

Figure 1 is a schematic view of a three-dimensional imaging installation 1. This conventionally comprises a radiation source 2 and a sensor 3 disposed on either side of a volume to be imaged 4 which may for example comprise a fixed support 5 on which is placed an object to be imaged 6. A For example, the radiation source 2 is an X-ray source, such as, for example, a Hamamatsu L860-01 micro-focus tube, and the sensor _. J3. is an X-ray sensor corresponding, such as for example a CsI scintillator coupled to photodiodes and. distributed by Hamamatsu under serial number T7942. Nevertheless, other types of radiation / detector pairs could be used in the invention, such as in the context of positron emission imaging, or single photon emission, or other.

To take a projection of the object to be imaged 6, according to a first incidence, the radiation source 2 ^' emits radiation in the direction of the object to be imaged 6, and the radiation passing through is detected by the sensor 3 which is, for example formed by a plane matrix of 2048X2048 elementary detectors. The radiation detected by each of the sensors of the sensor 3 is then read by a computerized system 7, which will be described in more detail later and which reconstructs a three-dimensional image from the data read in the sensor 3.

To reconstruct a three-dimensional image of the object to be imaged 6, several planar projections of the object to be imaged 6 are acquired at different angles. As such, the radiation source 2 and the sensor 3 can be mounted on an arm 8 movable relative to the support 5, and can for example turn 360 ° as shown by the arrow 9 of Figure 1. The rotation in question is for example performed step by step, including instants to which, the remaining fixed arm, a projection of the volume to be imaged is acquired, and moments to which the arm is moved to a subsequent acquisition position. During the rotation of the arm, the emission of radiation by the radiation source 2 can be interrupted, so as to reduce the radiation to which the object to be imaged is subjected 6. The rotation of the arm 9 can be controlled by the computerized system 7 In addition, its synchronization can be predetermined, or variable. For example,. lqsque_llo.bjet_at._J.ma_ger 6. _is _a_pe _. animal tit of laboratory, can be provided to synchronize the ^Sequence "displacement-acquisition" based on a physiological signal of the small animal to be imaged such that the respiration or the cardiac cycle. For example, there is a sensor 10 adapted to trigger the acquisition at each expiration of the small animal 6, so that all projections of the small animal 6 represent it in a given position corresponding to an end of expiration position, this which makes it possible to obtain a net three - dimensional image of the object 6.

As explained in more detail later, the raw data from the sensor 3 are not necessarily directly exploitable to reconstruct a three-dimensional image of good quality. Therefore, a series of one or more pretreatments can be applied to them in the computerized system 7.

To construct the three-dimensional image, it is possible, for example, to use a Feldkamp-type algorithm, in which the intensity in a coordinate voxel 11 (x, y, z) is equal to a weighted sum of the projections obtained according to each incidence, identified by its angle φ with respect to a reference position, such as for example the position shown in FIG. 2 according to a formula of the type:

I (x, y, z) = Σφ w (x, y, φ) .Pφ [u (x, y, φ), v (x, y, φ)],

in which :

w (x, y, φ) is a weighting factor applied to each projection P (u, v) and

P (u, v) is the intensity calculated during the pretreatments for the pixel 12 of coordinate (u, v) for the projection obtained according to the incidence characterized by the angle, φ. the angle φ.

This calculation is performed in each voxel of the three-dimensional image. These voxels may, for example, be 2.10 ⁸ voxels of 0.1 mm × 0.1 mm × 0.1 mm.

5 cover the volume to be imaged including a small whole laboratory animal.

A first embodiment of a three-dimensional imaging installation is shown in FIG. 3. The computerized system 7 comprises three groups of

10 machines 13, 14, 15 connected together on a network 24 supporting a data transfer rate sufficient to adapt to the data transfer rate at the output of the sensor 3. For example, the network supports transfer rates of the order of 100 megabits per second (Mb / s), or

Preferably at least 1 Gb / s, using Gigabit Ethernet technology. In the example presented, all. machines are star-connected to a Gigabit Ethernet switch 25, such as for example a CATALYST 3750 marketed by the company Cisco.

The first group 13 of machines comprises several machines 13a, 13b, 13c working in series, and each applying pre-processing to the raw data from the sensor 3. The second group 14 of machines comprises a plurality of machines, for example 8 14a machines, ...

14g, 14h, working in parallel, to perform retroprojections according to the Feldkamp algorithm described above. For the realization of this step, it is possible, for example, to predict that, from the data relating to the totality of each projection, each machine 14a,

30 14h, said processing machine, reconstructs a distinct portion of the volume to be imaged, said processing volume. The partial volumes reconstructed by each processing machine are then transmitted to the third group of machines 15, which may comprise one or more visualization machines, which assemble the volumes, partial to complete.

For each incidence, the data are subjected to the following treatment:

The raw data is transmitted from the sensor 3 to the first machine 13a of the first group 13 at a rate, depending on the selected sensor, and in the present embodiment equal to 16 Megabytes per second.

(Mo / s), corresponding to two possible projections per second. The first machine 13a of the first group 13 applies to the raw data a first pre-treatment to correct the physical phenomena due to the acquisition. This pre-treatment is for example configured according to a preliminary calibration step of the sensor. According to this preprocessing, the data is transmitted, at a rate of 32 MB / s due to the passing of integers lβbits to 32-bit floats, to the second machine 13b of the first group 13. This may, for example , apply a filtering of the data, for example by transformation in the Fourier space of the data acquired. The pre-processed data is transmitted to the third machine 13c of the first group 13 which is in charge of the data transmission to the processing machines of the second group 14.

The flow of data in the first group of machines is buffered if necessary in the different machines, to take into account possible variations of the flow.

Multicast data is implemented from the transmitting machine _. 13c to all of the machines 14a, ... 14h processing the second group 14. Each machine 14a to 14h treatment is connected to the output of the switch. Each processing machine 14a to 14h specifies that it is ready to participate in the reconstruction by sending on the network a request to be one of the receivers of the data broadcast. Alternatively, the transmission is carried out "Broadcast" to all machines in a closed network.

In order to obtain a fast transfer of data to each of the processing machines, it was necessary to develop a new library of functions allowing a simultaneous, or almost simultaneous, transfer. and reliable data to all processing machines. Indeed, the process multicast MPI library used by default by the processing machines, really does not make a ^transfer. simultaneous data to all processing machines. Indeed, multicast MPI operates as follows: in a first step, the data is transferred from the transmitting machine 13c to a first processing machine 14a. During a second step, the data are transmitted from each of the machines comprising the data (13c, 14a) to a machine that does not include them (for example 14b and 14c), and so on until the all machines that process data have retrieved them in memory. Such methods lead to congestion of the network incompatible with the objectives of speed of reconstruction fixed by the invention.

Data is duplicated within the network, at the switch, to all processing machines to receive data. This allows a true simultaneity of data transfer to all the processing machines 14a to 14h.

With the new library developed, a secure transfer protocol is realized because the three-dimensional reconstruction must be performed on the basis of all the processed data.

For this purpose, there is provided a secure protocol operating in the following manner. The pre-processed data, corresponding to the data, are given _. cut out, in data blocks each having X numbered datagrams. It is based here on the UDP protocol (acronym for the English expression "user datagram protocol" or "user datagram protocol"). The transmitting machine 13c transmits a data block consisting of X datagrams. The UDP protocol ensures that a transmitted datagram, if received, is received in full. On the other hand, it does not ensure that all the datagrams sent are received. Each processing machine 14a to 14h ensures the completeness of the data received. If the data block has been received in full, it returns a positive acknowledgment to the transmitting machine 13c, operating as a control transmitter machine. If the processing machine ^• notices that a number of datagrams has not been correctly received, it sends a request retransmission of missing datagrams to the sending machine 13c.

If the transmitting machine 13c receives positive acknowledgments from all the processing machines, it can start a subsequent transmission step, to issue the next data block to all of the processing machines. If certain datagrams have not been correctly received by certain processing machines, the transmitting machine 13c lists all the datagrams to be reissued, and re-transmits all these datagrams towards the set of processing machines 14. This step is possibly repeated a or several times until a positive acknowledgment is received from each of the processing machines. It can further be provided that, if the transmitting machine finds that too many datagrams are to be re-transmitted, it re-transmits the entire current data block to all of the processing machines.

A detailed example of communication is described below. Each block of data. a projection, is divided into X datagrams of 1500 bytes each, presenting the following structure:

- IP header defined by the standard, 20 byte content, - UDP header defined by the standard, 8 byte content,

- number of the projection in progress, coded on 2 bytes,

- UDP packet number for the projection, coded on 2 bytes, and

- useful data on 1468 bytes.

Each receiving machine has a virtual processor constantly listening, the network, apart from the few moments where the processor in question implements system functions. As a result, almost all the datagrams for the given projection will be received by each receiving machine.

Following the issuance of the last datagram for the current data block, the sending machine goes into receive mode. ^'Each preceptor machine after receipt of the last datagram or after receiving the numbered X datagram after a predetermined period of time, checks the number of datagrams received. It sends a TCP acknowledgment message of the data block to the sending machine. This acknowledgment can be either:

- an acknowledgment of the complete projection, in the case where it has received all the datagrams for the current data block, - a request for the retransmission of the complete data block, if it finds that a large number of datagrams has not been received,

- a request for retransmission of the only missing packets. In this case, the receiving machine transmits, to the attention of the transmitting machine. following the. TCP message acknowledgment of receipt, a second TCP message for example in the form of a bit table indicating for each packet whether it is necessary (bit to zero or if it is not necessary (bit to 1) to resend the packet in question On the basis of the reception of all the acknowledgment messages from all the receiving machines, the machine transmits, for the data block in question, whether retransmit, retransmits using the protocol described above all the missing packets for • the data block in question Redundant packets are set aside by each receiving machine Each receiving machine having requested a retransmission checks again if the missing packets are received, and the procedure described above may possibly be implemented again, until each receiving machine has issued a positive acknowledgment for the data block considered, for the attention of the transmitting machine.

The transmitting machine then proceeds to the transmission of the next data block.

It will be noted that the secure multicast broadcasting protocol that has just been described is suitable for its implementation in the context of a tomographic reconstruction method, but can be used for any type of multicast or broadcast data broadcasting method. for a set of machines of a restricted network.

It is furthermore possible to provide security that, if the machine emits notes that it has too often to retransmit a large amount of data, it signals to the user that it is necessary to intervene physically on one or more installations of the machine. network to replace a faulty switch or receiver.

The data is written directly into the memory of each. ma.chine_ of. _tra ± t.ement, .for .. that the reconstruction can be done quickly.

Each processing machine having received all the data corresponding to a projection, it applies the Feldkamp algorithm to the projection

5 received to retro-project it in a processing volume assigned to this processing machine, and corresponding to a predetermined fraction of the volume to be total image. For each voxel, each processing machine adds to a partial result obtained from the

10 previous projections the intensity value calculated for this projection.

Finally, when all the projections have been processed by the second group of machines, each of them transmits the processing volume it has processed 15 to the third machine group 15, for assembly and display, storage, or other . From the three-dimensional reconstructed image, the machine ^'display may show sections taken in the reconstructed image, of surface renderings, or other, ....

20 With the plant described _t in relation to Figure 3, the data flow are controlled optimally from the sensor to the display of the three-dimensional reconstructed image.

For a great flexibility of the process, it is possible

For example, using a set of machines 13a, 13b, 13c,

14a, ... 14h, 15 all identical, such as for example 11

2 Gigahertz (GHz) dual processor PCs running Linux operating system.

Having a group of machines connected to a network allowing suitable data transfer rates, this group of machines can be configured as follows:

Each machine is assigned a function as part of the reconstruction process, such as a function

3.5- of _M_Las-ion ,. _a_function ... jde treatment ,, a ... -function of pre-processing, a collection and assembly function, or other. For each machine, it also designates the identity of the previous machine (s) and / or the next machine (s) in the processing chain. . Each machine then executes the same program, which uses five logical functions that are specific to the function of the reconstruction process to be performed by the machine. These logical functions are:

- starttreatFN, which is traversed only once when a new treatment is initialized,

- stoptreatFN, which is executed once to complete a treatment,

- startblockFN which is called before the processing of a block of data,

stopblockFN, which is invoked to finish the processing of a data block, and

- treatblockFN, which is the function of processing a block of data.

The logical function library is available locally on each machine, or via the network. Also, it is expected that this method of configuring a set of machines connected in a network to an installation for three-dimensional reconstruction according to the method described above is implemented by software, in the form of a program that can be executed on at least one of the machines of the installation. This logical construction makes it easy to configure a new installation, or to reconfigure an existing installation for a new sensor, or whatever. In particular, nothing precludes that a preprocessing is then easily executable by several machines connected in parallel to the place of one and / or on the other machines 13a and 13b of the first group 13.

According to a second embodiment, shown in FIG. 4, it is intended to further reduce the number of machines intended to perform the pre-treatments in favor of an increase in the number of processing machines. With respect to the embodiment of FIG. 3, the machine group 13 consists of the only transmitting machine 13c.

An acquisition machine 16 includes a real-time processor card VME (acronym of the English expression "versa module eurocard") 17 which includes a daughter card input and output MFCC (acronym for the English expression "mult, ifunction Computing core ") 18 which comprises a processor 19 and an FPGA (acronym for the English expression" field programmable gâte area ") 20. This is plugged into one of the two PMC slots of the VME card thanks to its PCI interface. The VME card 17 also comprises a processor 21 and a random access memory 22 to the PCI standard and a connection member 23 to the machine 13c. ,

The FPGA 20 establishes the circuitry necessary to ^sequences acquisition and initialization of the sensor operating parameters.

The MFCC card 18 is connected directly to the output of the sensor 3 so that the data coming out of the sensor 3 is rearranged in flight via the FPGA, and the processor 19 applies a corrective preprocessing to the data before they are written in the random access memory. 22 of the VME 17 card, at a bit rate of at least 32 MB / s. Then, the processor 21 performs a last corrective treatment, such as a filtering, on the data written in RAM, and writes the pre-processed data, via the PVIC interface 23 in the random access memory of the machine 13c, via a parallel data transmission link If the machine 13c is temporarily overloaded, it is furthermore possible to store the data pre-processed in the random access memory 22, or in a suitable buffer at the input of the machine 13c. If necessary, it is also possible to make a copy - of the data backup in non-volatile memory, if one wishes to preserve _, these data for further processing. This copy can be made online while simultaneously processing real-time.

The structure that has just been described is also extremely flexible, to apply a modified preprocessing to the raw data at the output of the sensor, to replace the sensor, or else, since it is then sufficient to reprogram the FPGA 20 and / or the processor. 19 and / or the processor 21 in a suitable manner, all .de software.

Claims

1. A three-dimensional imaging installation of a volume to be imaged, comprising at least: at least one transmitting machine (13c), adapted to transmit on a network supporting transfer rates at least equal to 100 Megabits per second, destined to a plurality of data processing machines transmitted corresponding to a projection of said acquired volume according to an incidence,

^{• •} - said processing machines (14a, ..., 14h), each adapted to receive the received data relating to the transmitted data, said received data being obtained from the transmitted data, by duplication within the network to each processing machine, said apparatus being adapted to ensure that for each processing machine, the received data correspond to the ^'data output, | each processing machine (14a, ..., 14h) being adapted to process said received data to reconstruct a three-dimensional image of the volume to be imaged.

2. A three-dimensional imaging installation according to claim 1 wherein the transmitting machine is adapted to transmit the data transmitted in the form of data blocks each comprising a plurality of datagrams, in which each processing machine is adapted to transmit to the machine. transmitter, a request for retransmission of missing datagrams for the data block, in which the transmitting machine is adapted to list all the missing datagrams for the data block and to issue all the missing datagrams for the data block intended for all the processing machines.

3. Installation according to claim 1 or 2, wherein each of said processing machines (14a, ..., 14h) is adapted to perform a processing of said received data to reconstruct a three-dimensional image of a respective processing volume corresponding to a fraction of the volume to be imaged.

The plant of claim 3, further comprising a collection machine (15), for connection to each processing machine through the array, said collection machine being adapted to receive said three-dimensional images of respective treatment volumes of each processing machine, and to assemble them to form a three-dimensional image of the volume to be imaged.

5. Installation according to any one of claims 1 to 4 further comprising at least one pre-processing machine (13a, 13b) intended to be connected to at least one transmitting machine through the network, said pre-processing machine being adapted to apply a preprocessing function to a projection to generate data to be transmitted, and to transmit said data to be transmitted to said transmitting machine.

The apparatus of claim 5, wherein said pre-processing machine (13a, 13b) is adapted to receive a projection from a sensor at a sensor rate, and wherein the network rate rate is selected from function of said sensor flow rate.

7. Installation according to claim 5 or claim 6 further comprising a radiation source (2) adapted to emit radiation in the direction of the volume to be imaged, and a sensor (3) adapted to detect said radiation after its passage through the volume to be imaged, and to generate said projection.

8. Installation according to one of claims 1 to 7, wherein said machines are PC central units.

9 ¹ . Installation according to one of claims 1 to 4 comprising an acquisition machine (16) comprising a computer (17) and a connection member to a sensor (3) comprising an output delivering to said calculator raw data representative of a projection detected by said sensor, said computer comprising: an input / output daughter card (18) which comprises a preprocessing processor (19) programmed to apply preprocessing to said raw data output of said connector member, to generate pre-processed data,

- ^a RAM (22), and a bus connecting said processor and said pre-processing memory (22), said calculator being adapted to write in said memory said pre-processed data from the processor.

10. Installation according to claim 9, wherein the computer further comprises: an external connecting member, and a processor (21) adapted to read in said random access memory (22) said pre-processed data, apply them another preprocessing function for generating data to be transmitted, and writing the data to be transmitted in RAM from a transmitting machine via said connection member to the outside.

The apparatus of claim 9 or claim 10, wherein said preprocessing processor (21) is a reconfigurable processor adapted to be configured based on the sensor.

12. Method of configuring a group of. machinery in a three-dimensional imaging installation of a volume to be imaged, comprising at least:

-. at least one transmitting machine (13c), adapted to transmit on a network supporting transfer rates at least equal to 100 Megabits per second, to a plurality of processing machines, data transmitted corresponding to a projection of said acquired volume according to an incidence, said processing machines (14a, ..., 14h), each adapted to receive received data relating to the data transmitted, said received data being obtained, from the transmitted data, by duplication within the network to destination of each processing machine, said installation being adapted to ensure that, for each processing machine, the received data correspond to the transmitted data, each processing machine (14a, ..., 14h) being adapted to process said data received to reconstruct a three-dimensional image of the volume to be imaged, said configuration method comprising a configuration step during of which assigns each machine of said machine group with a selected function from a group of functions including ^at least a transmitting function, a processing function and a receiving function.

13. Configuration method according to the claim

12 in which the transmission function comprises a function of transmitting data transmitted in the form of data blocks each comprising a plurality of datagrams, in which the reception function comprises a transmission function, to the transmitting machine, a request to retrieve the missing datagrams for the data block ,. _and. in which the transmission function comprises also a listing function of all missing datagrams and a function of issuing all missing datagrams to all the processing machines.

The configuration method of claim 12 or 13, wherein the function group further comprises a pre-processing function, and a collection and assembly function.

15. Computer program comprising program codes for implementing the configuration method according to one of claims 12 to 14, when executed on at least one programmable machine.

16. A method for three-dimensional imaging of a volume to be imaged (4), comprising at least: (a) a transmission step during which at least one ^issuing machine (13c) transmits over a network permitting rate, of transfer at least equal to 100 megabits per second (Mb / s), to a plurality of processing machines, data transmitted corresponding to a projection of said volume, acquired according to an incidence,

(c) a control step in which it is ascertained, for each processing machine, that the data received corresponds to the data transmitted, and (d) a processing step during which each processing machine processes said data. data received to reconstruct a three-dimensional image of the volume to be imaged.

The three-dimensional imaging method according to claim 16, wherein, in step (a), the The transmitting machine transmits the transmitted data in the form of data blocks each comprising a plurality of datagrams, and wherein in step (c), (cl) each processing machine transmits to the transmitting machine a request for retransmission of missing datagrams for the data block,

(c2) the sending machine lists all the missing datagrams for the data block, and (c3) the sending machine issues all missing datagrams for the data block to all the processing machines.

18. A three-dimensional imaging method according to claim 16 or 17 wherein step (a) is repeated with other data corresponding to a projection of said volume, step (a) is repeated only after one _. positive result of said control step.

The three-dimensional imaging method according to claim 18, wherein steps (a) to (d) are repeated for a plurality of data corresponding to a plurality of projections of said volume, each acquired at an incidence, said incidences being distinct two together.

20. Three-dimensional imaging method according to one of claims 16 to 19, wherein, during the receiving step, in each processing machine

(14a, ..., 14h), receiving received data corresponding to at least a portion of said transmitted data, and wherein, during the control step, each processing machine (14a, ..., 14h ) ensures that the data received corresponds to the data transmitted in the following manner: each processing machine (14a, .., 14h) estimates a reception quality of the received data, and transmits towards a control transmitter machine ( 13c) a acknowledgment of reception quality of said received data, and said control transmitter machine waits for said acknowledgment of receipt of each of the processing machines. 21. The three-dimensional imaging method according to claim 20, wherein said data comprises a predetermined number of data datagrams, wherein, during the checking step, each processing machine (14a, ..., 14h ) transmits said acknowledgment of receipt only after receipt of said number. predetermined datagrams.

22. Three-dimensional imaging method according to claim 21, wherein in each machine of

Processing (14a, ..., 14h), said transmitted data correspond to said first received data and non-received data, wherein, during the checking step, said processing machine issues a reissue request

^'20 non-received data, and wherein, during a step of retransmitting at least said non-received data is re-transmitted on said network.

23. Three-dimensional imaging method according to one of claims 16 to 22, further comprising, before

The broadcast step, a communication establishment step (z) in which each processing machine (14a, ..., 14h) transmits a request to the network to receive the broadcast of said transmitted data, said data transmitted being duplicated within the

30, during the broadcast step, to the processing machines having issued a request to be a receiver of the broadcast of the transmitted data.

24.The three-dimensional imaging method according to one of claims 16 to 23, further comprising a step

35 (e) __ acquisition, prior to the dissemination stage, at during which a sensor (3) acquires raw data corresponding to said projection.

25. Three-dimensional imaging method according to claim 24, wherein the acquisition step is triggered according to a signal emitted by the volume to be imaged (4).

26. A three-dimensional imaging method according to claim 24 or claim 25, further comprising a pre-processing step (f) between said acquisition step and said transmitting step, in which said data issued by applying to the raw data a corrective preprocessing taking into account the physical characteristics of the acquisition.

27. A three-dimensional imaging method according to one of claims 16 to 26 wherein, during the processing step, each processing machine (14a, ..., 14h) processes a respective processing volume corresponding to a fraction of the volume to be imaged.

The three-dimensional imaging method according to claim 27, wherein during the processing step, each processing machine (14a, ..., 14h) retro-projects said received data into said processing volume to obtain a a three-dimensional retro-projected image corresponding to said projection, and adds said three-dimensional image to an earlier partial result obtained from previous projections of the volume to be imaged.

A three-dimensional imaging method according to claim 27 or claim 28, further comprising an assembly step (g) in which a collection machine (15) receives from said processing machines said respective processing volumes and assembles said processing volumes for reconstructing a three-dimensional image of the volume to be imaged.

30. Computer program including codes of program for the implementation of the three-dimensional imaging method according to one of claims 16 to 29, when it is performed on at least one programmable machine.