Patent Application
20100104513
The invention relates generally to the field of tumor visualization. More particularly, the invention relates to the evaluation and selection of dyes for tumor visualization.
In operative procedures to remove tumors, the surgeon's ultimate goal consists of removing all of the cancerous tissue while sparing as much of the normal tissue as possible. A surgeon must make a visual assessment of the outer boundary of the tumor and then try to completely resect the tumor. A successful resection of the whole tumor generally results in a greater 5-year survival rate for patients than a partial resection. Various imaging techniques may be used preoperatively or intraoperatively in order to determine the extent of the tumor. However, these images may fail to identify the outer layer of the tumor. Thus, after resection of the tumor some tumor cells may remain. The continued presence of such tumor cells may be problematic to the extent that residual tumor cells can lead to a local recurrence and, thus, properly identifying and removing the tumor boundary is a key focus in surgery to remove a tumor.
As one might expect, factors that impact the likelihood of local recurrence include the skill of the surgeon performing the tumor resection and the information available to the surgeon. In particular, as suggested above, one reason why surgical treatment may fail in the early stages of cancer is because the entire tumor may not be removed (i.e., lack of clear margins). At present, the surgeon typically relies on visual inspection and palpitation during tumor resection. However it is often difficult to distinguish cancer tissue from normal tissue by sight and/or by touch.
Therefore, information that may be used to delineate the tumor boundary intra-operatively may improve the effectiveness of resection procedures and thereby diminish the probability of local tumor recurrence. Given the importance of correctly identifying the boundaries of tumors, there is a need to develop tools to help recognize and highlight the tumor boundary in a variety of clinical contexts.
The present disclosure relates to the automatic identification of tumor boundaries with in image or images and the quantification of characteristics of these boundaries. In one embodiment, user input is provided to locate a dye-stained tumor in an image and, based upon this input, automated routines are employed to identify the boundary of the tumor. Characteristics of the boundary (such as measures related to average intensity, variance, contrast, or breaks in the boundary) may then be automatically measured and quantified and used as a basis for comparing the performance of the dye to other dyes or for comparing the performance of the same dye in different clinical contexts. In some embodiments, an intensity level standardization may be performed to standardize the intensity levels in each image so that the comparison of boundary characteristics between images is more meaningful.
In one embodiment, a method is provided that includes the act of accessing an image of a subject. The subject is administered an agent labeled with a dye prior to generation of the image. A tumor labeled with the dye is selected from the image. A first routine is employed to detect some or all of the boundary of the tumor. A second routine is employed to measure one or more characteristics of the boundary.
In another embodiment, a method for selecting dyes is provided that includes the act of accessing a plurality of images of tumors. The tumors are each stained with a respective image-enhancing dye of a plurality of dyes prior to imaging. The plurality of images are processed to identify the respective tumor boundaries within each image. One or more routines are employed to calculate one or more quantitative characteristics of each tumor boundary. One or more of the plurality of dyes are selected based on the one or more quantitative characteristics.
In another embodiment, a method for processing infrared image data to identify a tumor's boundary is provided. The method includes the act of administering an agent labeled with a fluorescent dye to a subject. An infrared image of the subject is generated and a tumor is selected from the image. A first computer-implemented algorithm is executed to identify the tumor's boundary. A second computer-implemented algorithm is executed to generate one or more quantitative characteristics of the tumor boundary. The one or more quantitative characteristics are reviewed to assess the performance of the fluorescent dye.
In another embodiment, a method is provided that includes the act of receiving an input indicative of the location of a dye-enhanced tumor in an image. A first routine configured to determine the boundary of the tumor in the image is executed. A second routine configured to calculate one or more quantitative characteristics of the boundary of the tumor is executed. The one or more quantitative characteristics are stored or displayed.
In yet another embodiment, a system is provided. The system includes a display capable of displaying an image of a dye-enhanced tumor and an input device configured to receive an operator input indicative of the location of the dye-enhanced tumor in the image. the system also includes a storage or memory device storing routines for determine the boundary of the dye-enhanced tumor and for calculating one or more quantitative characteristics of the boundary. In addition, the system includes a processor configured to receiving the operator input, to execute the routines stored in the storage or memory device in view of the operator input, and to display the one or more quantitative characteristics on the display.
These and other features, aspects, and advantages of the present invention will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:
As used herein, the term dye or dyes includes (but is not limited to) organic or inorganic fluorophores, fluorescent nanoparticles, fluorescent beads as well as their derivatives and conjugates to other molecules/vectors. Further, a vector is a vehicle that is used to transport the dye to one or more desired locations and may be targeted actively or passively. The use of dyes such as these to aid in visualizing certain medical phenomena is established. For example, certain dyes may be utilized to differentially highlight certain tissue types or structures, such as tumors. Such dyes may take advantage of particular properties of the tissues being highlighted.
Various approaches exist for developing agent, such as dyes, to highlight tumor tissue. For example, one approach, known as active targeting, targets tumor specific molecular targets, e.g. receptors, proteases, etc. (active targeting). Another approach, known as passive targeting, targets tumor morphology, e.g., leaky vasculature. Agents, i.e., dyes, developed using these types of approaches may be used to differentially highlight tumor structures. Such dyes may then be utilized in invasive procedures to allow a surgeon to visualize the extent of the tumor and to better facilitate removal of all tumor cells.
However, different types of tumors, subjects, or procedures may benefit from different dyes, i.e., different circumstances may call for different dyes. The number of potential suitable dyes, however, is vast and present techniques utilize subjective assessment which is qualitative in nature to screen candidate dyes or use manual procedures to highlight areas of interest before quantification. The latter approach is also subjective as a person visually identifies area of interest for quantification. In addition, manual identification is also laborious and time consuming. Such subjective assessments are generally unsuitable for screening large numbers of candidate dyes and, further, do not facilitate making meaningful comparisons between the candidates dyes.
In addressing this issue, therefore, it may be desirable to provide a more quantitative assessment and to utilize automation where possible. With this in mind, reference is now made to
In one embodiment, an operator may visually inspect the image 22 to determine (block 24) if the image 22 depicts a tumor that is suitably or sufficiently labeled with dye. In such an embodiment, the operator may consider factors such as whether the dye highlights only the boundary of the tumor (i.e., the tumor margin), whether the dye extends beyond the tumor or tumor boundary to an unacceptable degree, as well as, other aspects of proper labeling. If the operator decides the depicted tumor is not suitably labeled, the operator may access a different image 22. If the operator decides that the depicted tumor is suitably labeled, the operator may proceed to process the image 22.
Once a suitable image 22 is identified, the operator may select (block 26) the dye-labeled tumor 28 in the displayed image 22. For example, the operator may employ a mouse, touchpad, touchscreen, or other suitable point-and-select interface to select the tumor 28, such as by “clicking” on the perceived center of the tumor using a mouse or other suitable selection input device. In other embodiments, selection of the tumor 28 may be automated or semi-automated, such as by employing thresholding or other algorithms that identify concentrations of the dye over a certain limit within the image 22. In such embodiments, a tumor 28 may be tentatively identified based on the thresholding algorithms alone or potential tumors may be identified on the image 22 by the algorithm for further review and selection by an operator.
Once a tumor 28 is identified, one or more automated routines may be employed to detect (block 30) the boundary 32 of the tumor 28. The routine 18 may detect the entire boundary 32 of the tumor 28 or only a portion of the boundary 32, depending on the extent the dye highlights the boundary 32 of the tumor 28. In one embodiment, this routine, as well as others discussed herein, is implemented using the IDL language and can be distributed using the IDL virtual machine.
In one embodiment, another automated routine may be employed to measure (block 34) one or more quantitative characteristics 36 of the boundary 32. Examples of such boundary characteristics, as discussed in greater detail below, include average intensity, pixel intensity variance, number and relative length of boundary discontinuities, brightness ratio, average contrast, clearance rate, and so forth. The characteristics 36 of the boundary 32 may be reviewed or evaluated by an operator to evaluate or compare the efficacy of the dye in staining the tumor 28. In addition, the characteristics 36 may be stored for later review or comparison. As will be appreciated, some of the steps depicted in the flow chart of
With the foregoing general discussion the following example is provided by way of illustration. Turning now to
In certain embodiments, the image 22 may be processed prior to tumor selection and/or identification of the tumor boundary. For example, in one embodiment, the image 22 may be enhanced, such as by implementation of anisotropic smoothing and/or other pre-processing filters. In addition, in certain embodiments the image 22 may undergo contrast stretching and/or multi-stage binarization.
Once the tumor 28 is selected a computer-executed algorithm may automatically identify the tumor boundary 32. In one embodiment, the tumor boundary 32 may be identified utilizing an intensity threshold. Pixels having an intensity greater than a set or threshold value may be determined to correspond to tumor tissue. In turn, those pixels determined to correspond to tumor tissue that have intensity values greater than a neighboring pixel in at least one direction may be determined to correspond to the boundary 32 of the tumor 28. That is, those pixels which are stained (e.g., fluorescing) but which are adjacent to at least one other pixel that is not stained (e.g., non-fluorescing) above a certain threshold may be identified as corresponding to the boundary 32 of the tumor 28.
In one embodiment, upon determination of the tumor boundary 32, the circle 38 used to highlight the region having the tumor 28 may be warped to highlight the identified tumor boundary 32, as depicted in the inset to
Turning now to the screenshot depicted in
In one embodiment, the algorithm employed may generate quantitative boundary characteristics 36 of one or more aspects of the tumor boundary 32. For example, in one embodiment, a quantitative descriptor of the average brightness of the tumor boundary 32 may be measured by averaging the intensity values of those pixels determined to correspond to the tumor boundary 32. Similarly, other measures of central tendency such as median and mode values, may be calculated based on the intensity values of those pixels determined to correspond to the tumor boundary 32. These descriptors may then be stored or displayed for evaluation by a reviewer.
Other types of quantitative boundary characteristics 36 may also be calculated. For example, a quantitative descriptor of the variation of brightness of the tumor boundary 32 (e.g., the standard deviation of the pixel intensities for those pixels corresponding to the tumor boundary 32) may also be calculated. In addition, in some embodiments the quantitative boundary characteristics 36 may include the number of discontinuities or breaks 54 in the tumor boundary 32, as well as, the length of each discontinuity 54. For example, the length of each discontinuity 54 may be described by equation (1) as follows:
where Ldisc, refers to the length of the discontinuity.
A further descriptor which may be quantified in certain embodiments is the squared average contrast. The squared average contrast may be described by equation (2) as follows:
where C refers to the squared average contrast, Imargin refers to the average pixel intensity in the tumor boundary 32, and Ibackground refers to the average pixel intensity in the background region surrounding the tumor boundary 32. In the depicted embodiment, the thickness of the background region used in quantifying and generating characteristics 36 such as the squared average contrast may be adjusted by the operator, such as via slider 58 of the user interface screen. Adjusting the amount or thickness of the region designated as background may vary the sensitivity and/or accuracy of the generated quantitative boundary characteristics 36. In implementations where different dyes are ranked with respect to each other, it may be useful to keep the thickness of background region constant. In one embodiment, the background region thickness is set to a default of forty-one pixels.
Yet another boundary characteristic 36 that may be quantified in certain embodiments may be rotational contrast, i.e., the ratio of the rotational average of the tumor boundary pixel intensity to the rotational average of the background pixel intensities surrounding the tumor boundary 32. In such an embodiment, the rotational average may be considered the average of the average brightness along the radius around 360 degrees. The rotational contrast may be described by equation (3) as follows:
Wherein Crotational refers to the rotational contrast, Irot
With the foregoing in mind, it should be appreciated that quantitative boundary characteristics 36 may be generated in a variety of contexts for different dyes, tumor types, points in time, lab animal types, and so forth. These quantitative descriptors may be used to select or grade dyes based on their suitability in different clinical contexts or to select dyes for further testing.
For example, in one embodiment, an operator may process a plurality of images as described herein. In such an embodiment, the operator may access (block 20) a plurality of images 22, such as IR images, of tumors suitably stained with one or more fluorescent or other suitable dyes. The operator may exclude (block 24) those images which exhibit poor or unsuitable staining characteristics from further consideration. In one embodiment, the operator may process the remaining images to select (block 26) the respective tumors 28 within each image 22. One or more automated routines may be executed to identify (block 30) the boundaries of each selected tumor 28. As will be appreciated, the identification of tumor boundaries may occur in a batch processing of the images 22 or may be performed on each image 22 separately as the tumor 28 is selected. The identification of tumor boundaries may be performed contemporaneous with or subsequent to the execution of other routines to enhance the tumor boundaries, such as routines for implementing one or more anisotropic smoothing operations, contrast stretching, multi-stage binarization, and so forth.
One or more automated routines may be implemented to determine (block 34) characteristics 36, such as quantitative measures, of each tumor boundary 32. In certain embodiments, the quantitative descriptors may be standardized (block 80) or normalized for each tumor boundary 32. For example, such standardization processes may account for variations in brightness and/or other image property differences. In one such embodiment, the operator may select a dark area in the respective image 22. The routine calculating the boundary characteristics 36 may in turn use the intensity of the selected dark region (or an average of the intensity in the selected dark region) to normalize or otherwise adjust for differences in brightness between images 22. In this way, differences in image brightness may be normalized by establishing a base darkness level for each image which may be used to scale other intensity levels in the respective image 22.
In this manner, comparable quantitative boundary characteristics 36 may be generated for the respective tumor boundaries 32 observed in each processed image 22. The boundary characteristics 36 may then be ranked (block 82), either automatically or by a reviewer, by one or more of the characteristics, allowing a reviewer to select (block 86) which dyes 84 performed best in different medical contexts, such as in different animal models, on different tumor types, based on clearance rate, and so forth. Selected dyes may then undergo further testing and/or may be selected for use in invasive procedures, such as in surgical procedures for tumor removal. In this way, a reviewer may select dyes based on quantitative measurements, as opposed to a subjective visual assessment. As will be appreciated, the order in which different steps illustrated in
Referring now to
Accessed or processed image data, as well as the boundary characteristics described herein, may be displayed on a display 108 for review by an operator. In addition, the processor-based system 98 may include one or more input devices 110, such as a keyboard, mouse, touchscreen, touchpad, and so forth, allowing an operator to access image data, select images for processing, select tumors, within images, review results, and so forth. In this manner, an operator may review the outputs of the disclosed techniques and provide inputs to further operation of the disclosed techniques.
The identification of tumor boundaries and quantification of dyes used to highlight the tumor boundaries, as described herein, provides a useful tool to the medical and scientific community. For instance, with the methods outlined above a number of dyes can be analyzed and the data obtained stored to allow comparisons between the dyes to determine the best dyes in general and for specific tumor types. In addition, the efficacy of a dye can be shown over multiple tumor types. Possessing quantitative measurements introduces reliability and reproducibility in assessing the dyes, removing the subjectivity normally involved.
Another benefit of the methods is the automatic detection and marking of the tumor boundary, once the operator selects an area of interest, provides an invaluable tool in a dynamic environment such as a surgical setting. Applying these methods to imaging systems used in open surgery would improve the ability of the surgeon to remove the complete tumor while sparing as much of the normal tissue in the patient as possible.
Technical effects of the invention include the automated or semi-automated identification of tumor boundaries and the quantification of dye efficacy in staining the boundaries. Such measures may allow the analysis and comparison of multiple dyes in a quantitative, objective manner.
While only certain features of the invention have been illustrated and described herein, many modifications and changes will occur to those skilled in the art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes as fall within the true spirit of the invention.
