Method and apparatus for combining continuous wave and time domain optical imaging

This is the first application filed for the present invention.

The invention relates to the field of optical imaging of turbid media, such as mammalian tissue. More specifically, the invention relates to optical imaging using both continuous wave and time domain methods.

Imaging of mammalian tissues has been used extensively to obtain information on the internal structures as well as on the biodistribution of molecules. This information can of course be utilized for diagnosis purposes. Several techniques based on different physical principles are currently available to obtain images that encompass a broad range of spatio-temporal resolution. Such techniques include Magnetic Resonance Imaging (MRI), Positron Emission Tomography (PET), single-photon emission computed tomography (SPECT), X-ray, ultrasound and, now emerging, optical imaging.

In optical imaging three approaches have been used to generate the optical data necessary to reconstruct images of volume of interest (VOI) namely Continuous Wave (CW), which involves the measurement of light attenuation; Time Domain (TD), which involves injecting a pulse of light of short duration within the VOI and detecting the light as a function of time as it exits the VOI; and Frequency Domain (FD), which relies on frequency modulation of the light source and analysis of the phase and amplitude of the signal as it exits the VOI.

Continuous wave permits rapid acquisition and is the least expensive of the three approaches but provides a limited amount of information. More specifically, in CW imaging, the scatter coefficient of the VOI must be an assumption in order to obtain absorption coefficient information. CW cannot determine absorption separately from scatter. TD provides more information. In TD, a short laser pulse is injected in the part of the mammal to be imaged and the distribution of the time of flight of the photon exiting the volume of interest is measured. The resulting signal is referred to as a temporal point spread function (TPSF) that can be used to calculate such characteristics as the mean time of flight of photons. In FD, the intensity of the source is modulated with high frequency. As a result, a photon density wave propagates in the tissue and the amplitude and phase shift of the wave relative to the incident wave is measured. In principle, by scanning the tissue with a range of different frequencies, the entire TPSF can be reconstructed. However, in practice, a single frequency is usually employed to estimate the mean time of flight of the photons.

Image reconstruction using optical data belongs to the class of inverse problems. The problem consists of finding the distribution of optical parameters in tissue based on the detected optical signal. While image reconstruction techniques are still the subject of intense research activities, several tested approaches have been used with relative success. Some of these approaches are summarized and described in Boas et al. (IEEE SiG. Proc. Mag., Vol. 18, No. 6, pp. 57-75, 2001) and Hawrysz and Sevick-Muraca (Neoplasia, Vol. 2, No. 5, pp 388-417, 2000).

TD measurements provide detailed information about the absorption and the scatter from within a tissue, however, the method suffers from long acquisition time, expensive hardware and complicated software analysis. Furthermore the acquisition often results in noisy TPSF data from which accurate estimates of spatial optical information are difficult to obtain.

It would therefore be desirable to provide a method that would overcome the limitations of CW and TD, while retaining their advantages.

The inventor has provided a method and apparatus by which the advantages of TD optical imaging can be exploited while avoiding certain disadvantages. In particular long acquisition times can be substantially reduced.

In one aspect of the invention there is provided a method for optical imaging of a volume of interest (VOI) in a turbid medium, the method comprising:

optically scanning the VOI at a plurality of source/detector geometries using continuous wave (CW) to generate a measured attenuation value at one or more wavelength, the measured attenuation value being measured from a continuous light intensity value provided to the source and detected at the detector; optically scanning the VOI using time-domain (TD) to generate data representing at least a portion of a measured temporal point spread function (TPSF) at the one or more wavelength, the TPSF data being measured for each of the plurality of source/detector geometries from a plurality of light pulses from which a statistical average of light received as a function of time is obtained, the TPSF data yielding information about absorption and scatter within the VOI; and combining information provided by CW and TD, whereby to generate an improved image of the VOI with fewer light pulses and a shorter acquisition time by using both time-domain and continuous wave modalities.

In another aspect of the invention there is also provided an optical imaging apparatus for imaging a turbid media object, the apparatus comprising: at least one optical source for providing continuous and pulsed optical energy; at least one optical detector for detecting optical energy and generating time-dependent and continuous data; a source/object optical coupling for coupling the optical source to a desired position on the object; a detector/object optical coupling for coupling the optical detector to a desired position on the object; an acquisition controller connected to the optical source and the optical detector for collecting the time-dependent and continuous data for a plurality of source/detector geometries within a volume of interest in the object; a raw TPSF data compiler for receiving a time-dependent output signal from the detector and generating raw TPSF data output; a TPSF data enhancer for processing the raw TPSF data output and the continuous data to provide enhanced TPSF data, wherein the continuous data corresponds to the raw TPSF data output according to the source/detector geometries.

It will be appreciated that the present invention allows for a number of different possible economies to be achieved by combining CW with TD techniques. Firstly, the estimate of absorption coefficients for a VOI from CW (making an assumption about scatter, such as assuming that scatter is constant for example) can be used as initial values in processing the TD data to yield more accurate absorption coefficients (along with scatter coefficients). Secondly, the process of determining the optical properties of the VOI can involve simultaneously comparing CW model data with CW measured data and comparing TD model data with TD measured data. Thirdly, processing time and acquisition time can be reduced by using an image generated from CW data alone as a less-accurate image, whose purpose is to help in defining a subset of source/detector configurations or geometries to be used in collecting TD data for specific portions of the VOI requiring better detail.

Further features and advantages of the present invention will become apparent from the following detailed description, taken in combination with the appended drawings, in which:

FIG. 1 is a high-level flow chart illustrating the general method according to some embodiments of the present invention;

FIG. 2 is a block diagram of an imaging apparatus according to another embodiment of the invention in which CW data is used to help in determining optical properties of a VOI of an object being imaged using TPSF data and a TD photon migration model;

FIG. 3 a is a plot of amplitude vs time (bin number) for a TPSF and its first derivative; and

FIG. 3 b is a plot of amplitude vs time (bin number) for the first derivative of a TPSF and its second derivative.

It will be noted that throughout the appended drawings, like features are identified by like reference numerals.

In the present description, by CW measurements it is meant optical signals measured while the light source that illuminates the object to be imaged is continuously on, namely a steady state flux measurement.

By Time Domain (TD) measurements, it is meant that the optical signal is measured as a function of time following a brief pulse of the light source. The resulting signal is the so-called temporal point spread function (TPSF), and the TPSF data collected may be either one, two, a few or several points on the TPSF (often referred to as different time gates), a single parameter resulting from a plurality of points, such as a slope or other characteristic of the TPSF curve, or the whole TPSF data. When the TD optical detector is a single-photon counting detector, the whole TPSF can be efficiently collected as a result of the measurement technique. When a gated ICCD is used, it is more practical to select time gates. Due to the nature of the detection system, the TPSF data requires a large number of acquisitions (and in comparison to CW, much more time) to achieve relatively stable measured values. In many cases, the TPSF data will be inherently noisy.

Optical image reconstruction can be based on complex photon propagation models such as Boltzmann transport equation (Arridge, Inv. Prob., Vol. 15, pp. R41-R93, 1999) or on simpler model photon diffusion (migration) equations which incorporate an optical diffusion coefficient, D, which is a function of absorption coefficient μ a and the modified scatter coefficient μ s ′. Based on these equation and optical signal measurements, μ a and μ s ′ can be estimated and therefore provide a spatial distribution of these optical properties within the object (Hawrysz and Sevick-Muraca, Neoplasia, Vol. 2, No. 5, pp 388-417, 2000).

Photon diffusion equation can be solved using both CW and TD measurements to predict the photon fluence based on known spatial distribution of μ a and μ s ′ with the difference that there is no time dependence for the CW case. However, if these properties are not known a priori they must be calculated or estimated. TD measurements with appropriate processing of the data can yield μ's information, but as mentioned previously requires longer acquisition times. CW measurements are faster and more economical but cannot distinguish between μ a and μ s ′.

In one aspect of the invention, there is provided a method that advantageously combines CW and TD measurements without some of their disadvantages to produce optical images. Thus and referring to FIG. 1 , CW and TD data are acquired at 10 and 12 over a given part of an object such as a mammal comprising a volume of interest (VOI) to obtain an attenuation value and a TPSF, and are combined at 14 to produce an image of the VOI. The combination of the data from the two measurement modes permits a more accurate image reconstruction and can also result in substantially reduced acquisition times.

In one embodiment of the invention the optical properties μ a and μ s ′ are estimated using both CW and TD measurements obtained independently, namely without necessarily having the same source/detector geometries. The resulting values of μ a and μ s ′ obtained with each modality are compared and are reciprocally adjusted until the difference is less than a predetermined value. In other words, the values of μ a and μ s ′ are iteratively adjusted for one modality by taking into account the values obtained with the other modality. This approach advantageously reduces the acquisition times for TD measurements by providing additional information with CW allowing acquisition of TPSF with lower S/N ratios. Alternatively CW and TD data can also be combined to estimate μ a and μ s ′ as opposed estimating μ a and μ s ′ with each modality.

It will be appreciated that iteration may be saved if the estimate of μ a by way of CW imaging is used as a starting point for determining μ a and μ s ′ by way of TD imaging. Furthermore, the process of imaging and calculating the values for μ a (i.e. by making an assumption about μ s ′, for example by holding μ s ′ constant) by CW can be carried out in parallel with the acquisition of TPSF data. Preferably, the CW imaging will be completed before the TD data acquisition is completed and determination of μ a , μ s ′ values begin using the TPSF data. In this way, the μ a data determined by CW can be used as an initial best guess in the TD imaging. Of course, the collection of CW data for the various source/detector geometries as is needed by the CW data processing unit may require little time in comparison with the CW data processing time. Provided that TD data acquisition can continue in parallel with the processing of the CW data, parallel imaging can be provided.

The above embodiments of the method of the invention can be carried out using a system as described in FIG. 2 in which a combination of CW and TD modalities is described that synergistically integrates the two approaches to provide faster and more accurate estimates of the optical properties of an object. In this regard, a continuous 40 and a pulsed 42 laser source are provided to inject light within an object 41 . Also provided are one or more detectors suitable for CW 44 and TD 46 detection. The optical coupling of the sources and detectors at 48 and 50 respectively with the object to provide desired source/detector configurations can be realized using known technologies such as optical fibers, free-space optics and the like.

The controller 56 has a control output signal to coupling device 48 for selecting the source position. This is preferably achieved by way of an optical switch in the case of a fiber optic coupling, and in the case of free-space optics by way of a galvo-mirror inserted within the optical system. The detector coupling device 50 preferably requires no switching control. This is achieved in the case that the detector has a channel for each detector position, as is practical in the case of a gated ICCD. However, when using free-space optics, the detector position is controlled by galvo-mirror or other means to achieve movement of the detection spot over the object.

The time resolved detector 46 is linked to a raw TPSF data compiler 52 that can generate TPSF's. Compiler 52 may also comprise a normalizer unit wherein TPSF data can be normalized by the CW data (attenuation) to produce dimensionless data to avoid, for example, experiment-to-theory unit mismatch thereby enabling quick processing of the data. An estimator 54 is provided to estimate values for μ a and μ s ′ from the measured optical data and the calculated optical data for the given source/detector geometry. The calculated CW optical data is provided by a calculator 62 , while the calculated TD optical data is provided by a calculator 64 . In the embodiment of FIG. 2 , the TPSF data acquired and enhanced may be the full TPSF, a TPSF characteristic parameter, or a plurality of time-gate TPSF points.

The modes (CW or TD) and the sequences of data acquisition are controlled by an acquisition controller 56 . In particular the acquisition controller can advantageously dynamically control the acquisition of data based on the information received by the detectors. Once the acquisition is completed the image data is stored in raw image data store 58 .

It will be appreciated that the system can be modified and still achieve similar results. For example, a single laser source that can provide a continuous as well as pulsed output can be used. Also a single detector can be used to detect continuous or time resolved signals. Other modifications as would be obvious to one skilled in the art are also considered to be within the scope of the invention.

The apparatus of FIG. 2 can be used in a variety of ways. Firstly, the estimator 54 can process CW data for a VOI. This is done by making an assumption about μ s ′ value for the whole VOI. While this assumption may be wrong, it allows for an estimation of μ a that is close in some cases, and at least a useful first approximation in other cases. The resulting 3D map of μ a and μ s ′ stored in store 58 would have the assumed μ s ′ values (which could be constant values), and a first approximation of μ a . In estimating both μ a and μ s ′ using the TD optical data, the assumed μ s ′ value and the first approximation of the μ a values is are fed to calculator 64 . The estimator 54 will compare the resulting calculated TPSF optical data to the measured TPSF data to determine in a first pass an adjustment to both μ a and μ s ′. This may be done for more than one TPSF time gate in the case that time gate values are used. The mode of operation allows for the possibility of acquiring cW data and processing it to obtain the first approximation of μ a for the whole VOI or object, while acquisition of raw data on the TD side continues during CW calculation time (but not during CW data acquisition unless parallel optical data acquisition hardware is provided).

Secondly, the estimator 54 can compare the CW and TD calculated and measured optical data together and decide on a best value for μ a and μ s ′. In this case, the calculator 62 uses the μ a and μ s ′ data in which there is no assumption about μ s ′ values. Since the CW data is less noisy, a better estimate of μ a and μ s ′ is obtained using both the CW and TD data than with the TD data alone. While this second approach may be done wholly simultaneously, namely CW and TD acquisition and data processing together, it may also be done following a first CW pass as in the first approach.

Thirdly, it will be appreciated that the apparatus of FIG. 2 can be used to obtain a CW image prior to imaging by TD. In this case, the CW image may be used to select a specific portion of the object 41 that would require better resolution by TD imaging. In this case, an input to the acquisition controller 56 (not shown) is provided to select the specific VOI to be imaged by TD. The CW-based image data in store 58 would then be enhanced by the TD data. Again, for the specific VOI, the CW data could be re-measured or retrieved from a data store (not shown) and used in conjunction with the TD data to enhance image reconstruction. By reducing the volume to image by TD, image acquisition and reconstruction data processing time is saved.

For example CW can be used to find a lesion (VOI) within a breast, while TD measurements can be used to reveal finer details about the spatial distribution of optical properties within the lesion that may be helpful in diagnosing the breast lesion. This strategy of data acquisition may also be used for example to find the biodistribution of a compound by CW, while TD measurements of the regions comprising the compound may serve to calculate the concentration of the compound.

The TD and CW data can be acquired either sequentially or simultaneously. Sequential acquisition may be achieved for example by using a time-gated intensified CCD camera. This type of camera provides the capability for measuring optical signals both in the CW and TD modes. The CW mode data can be acquired by simply turning the time gate off. Alternatively TD and CW data can be acquired using optic components that are exclusively designed for one mode.

Advantageously, the CW and TD data can be acquired using a gated ICCD camera. The advent of the gated intensified charged coupled device (ICCD) has permitted that significant number of time-domain spatial measurements be performed simultaneously. Moreover, when the gate is turned off the camera works in a CW mode, thus obviating the need for a separate CW and TD hardware. An image sequence of CW mode, TD mode data acquisition can thus be obtained

Simultaneous acquisition may be achieved by acquiring the TD and CW data at different wavelengths thereby permitting spectral discrimination and simultaneous acquisition. In this case the acquisition of the optical signal is performed at a plurality of wavelengths and the CW and TD data are combined to reconstruct an image.

In a preferred embodiment certain specific time points or intervals of time (time gates) of the TPSF at which TD data will be acquired are selected. The determination of the time gates is preferably based on the following approach:

A noiseless TPSF is a continuous and infinitely differentiable function defined on R + . Critical points of a function can be identified via zeros of this function's derivative. Due to the fact that a TPSF has a characteristic shape, the i th derivative will yield a set of points where the derivative became zero, which corresponds to extreme values of the underlying function. The first subset s 1 ={s 1,1 } corresponding to the 1 st derivative describes the unique TPSF maximum. The second subset s 2 ={s 2,1 ; s 2,2 } corresponding to the 2 nd derivative reveals two points where the curvature directionality changed. Derivatives of the higher order yield information on more subtle details of the TPSF shape. The subset s i can be defined as s i ={s i,1 ; . . . ; s i,i } for the case of continuous TPSF.

Given a TPSF f(x) represented with a set of M values (M>>N), confined within an interval [a o ,b o ] one can perform recursive numerical “differentiation” of the TPSF and of its' “derivatives” by computing the quotient:

Q i = f ⁡ ( x i + 1 ) - f ⁡ ( x i - 1 ) x i + 1 - x i - 1 , x i ≥ a 0 , x M ≤ b 0

which would approximate the value of f′(x 1 ).

When plotted, characteristic sequence of functions will emerge as shown in FIGS. 3A and 3B . In this example, one observes that: s 1 ={31.5}, s 2 ={20.5; 40.5}, s 3 ={14.5; 25.5; 48.5}.

In order to obtain an adequate TPSF representation with N points or time gates, one would want to reflect the TPSF shape as fully as possible. The sequence of numbers {s 1 ; s 2 ; . . . } allows for identification of the most relevant points in the order of diminishing significance as the subset number increases. Elements within a given subset s i have equal significance.

One can replace s i with:

s i ′≡{( s i,1 −0.5 r,s i,1 +0.5 r ); . . . ; ( s 1,i −0.5 r,s i,i +0.5 r )}

where r is the width of a time gate. The set s i ′ is a collection of intervals rather than points.

In the case where several partial TPSF's are to be acquired, when for example different source-detector configurations are used, the procedure outlined above can yield S i ={s 1 i ; s 2 i ; . . . }; i=1, . . . , K for a given TPSF where K is chosen a priori. In order to reflect the most relevant features of the set of TPSFs, the union of the nth derivative zeros is obtained. For example S 1 =s 1 1 ∪ s 1 2 ∪ . . . ∪s 1 k , S 2 =s 2 1 ∪ s 2 2 ∪ . . . ∪s 2 k and so on. The set {S 1 ; S 2 , . . . } provides a sequence of numbers that should be used to represent the set of K TPSFs in an optimal manner.

As above, s i might be replaced with s i ′ in order to analyze time gates instead of time points.

In one embodiment of the invention the steps to select the appropriate time points or time gates are as follows:

Acquire TPSFs with high temporal resolution for all source-detector pairs that are capable of functioning simultaneously. At least one typical source position should be used. Robustness of the subsequent analysis increases, as more typical source positions are included. Approximate all measured TPSFs with curves based on a model of photon propagation. Perform analysis as described above to determine the N time gates to be used during acquisition. Acquire data with the N points of the set {S 1 ; S 2 , . . . }.

Preferably N is an input parameter. In practice it will be determined via estimation of the resulting data set size that can be used to reconstruct images within desired time frame. One might want to estimate N as well in case data set size is not an issue.

Time window [a 0 ,b 0 ]⊂R + used for initial data acquisition is assumed an input parameter. However, it should be chosen wide enough based on prior knowledge in order to maximize chances of capturing most significant TPSF features.

Temporal sampling, i.e. the distance between x i and x i+1 is an input parameter.

Temporal resolution r 0 , i.e. the minimal distance between x i and x i+1 that yields “uncorrelated” values f(x 1 ) and f(x i+1 ), comes primarily from the limitations of the data acquisition hardware and is, therefore, an input parameter.

It is preferable to perform initial data acquisition with the steps of not more than 0.5×r 0 and time gate width of not more than r 0 . Eventual time gate width r≧r 0 might be chosen to minimize data set size while preserving the distinct TPSF features revealed by the analysis. For example, distinct points s i,j and s i′,j′ would fall within the same time gate if |s i,j −s i′,j′ |<r.

The use of variable time gate widths may yield superior results. One would replace s i ′ with s i ″≡{(s i,1 −0.5r i,1 ,s i,1 +0.5r i,1 ); . . . ; (s i,i −0.5r i,i ,s i,i +0.5r i,i )} to analyze this case.

The selected time gates may be used to reconstruct the TPSF. Constraints may be used in the reconstruction. For example, the integrated TPSF may be constrained to a value substantially identical to a measured attenuation value obtained by CW measurements. Further, the number of time points or time gates that is necessary to acquire in order to reconstruct an acceptable image is reduced because of the additional information conveyed by the CW measurements.

With partial TPSF acquisition, gated ICCD camera may allow acquisition of CW and TD data according to predetermined sequences as would be obvious to one skilled in the art. Zoom lenses may also be used to obtain CW data in full field mode and TD data by zooming in on a desired VOI.

The embodiment(s) of the invention described above is(are) intended to be exemplary only. The scope of the invention is therefore intended to be limited solely by the scope of the appended claims.