Patent Application
20240127430
The following description relates to tools and methods of imaging and measuring a stoma.
A small intestine is composed of three parts, with the duodenum being the first part, the jejunum being the second part, and the ileum which is attached to the colon (large intestine) being the last part. In normal circumstances, 90% of nutrients are absorbed within the first 150 cm of the small intestine, with 9-10 L of endogenous fluid entering the stoma, and 1.5 L of fluid exiting into the colon where 90% of the liquid is absorbed, eventually resulting in fecal excrement containing about 0.1 L of fluid. The fluid entering is primarily composed of bile, saliva, gastric and pancreatic juices, whose absorption in the small intestine is dependent on electrolyte transport and the intercellular space permeability. When an ileostomy is performed, around 200-700 mL of effluent is excreted daily, but this can vary over a wide range. Normal ileostomy effluent is primarily water (90%), with a 120 mL mmol/L sodium concentration. However, complications such as ileostomy diarrhea, cholelithiasis, urolithiasis, and more can affect the composition of the effluent.
An ostomy may be temporary, constructed in advance of a gastrointestinal surgery, or may be permanent. Due to the nature of ostomies' impact on patients' health and lifestyle, it is important to understand the effect of an ostomy on patient's quality of life. The function of the stoma, odor control, and appliance adequacy/fit are all important considerations when determining the patient's quality of life and are determined by preoperative, operative, as well as postoperative conditions including an ostomy pouch used by the patient. Postoperatively, it is important for medical professionals to perform individualized fitting of ostomy pouches, and track changes in stoma size and shape so that a leak and odor proof seal may be created around the pouch and peristomal skin health may be maintained.
Over 750,000 individuals in the United States live with a stoma, and 130,000 individuals undergo a procedure resulting in a stoma each year. Stomal complications are relatively common, with a 34% complication rate over a twenty year period at Cook County Hospital from around 1,600 patients. (J J Park, A Del Pino, C P Orsay, R L Nelson, R K Pearl, J R Cintron, H Abcarian. Stoma complications: the Cook County Hospital experience. Dis Colon Rectum. 1999 December; 42(12):1575-80. doi: 10.1007/BF02236210).
Peristomal skin irritation is also relatively common, occurring in up to 42% of patients.
A common method for intestinal stoma imaging is contrast enhanced-computer tomography (CT). In CT, multiple x-rays from different angles may be taken to produce cross sectional images of a target. However, radiologists often overlook stoma sites when interpreting CT studies. This can often cause complications, as abnormalities are not recognized in the imaging process. The CT is also relatively expensive, as a radiologist is required to look at the CT studies by hand.
Another imaging technique is CT combined with a stomal enema. This technique is achieved by initially using unenhanced acquisition of the target area, followed by a contrast enhanced CT using injection of contrast mediums, such as iopamidol (Bracco) or iopromide (Schering Pharma). This is also considered relatively expensive and is only used for selected patients. MRI imaging may be used, for example, for younger patients who may require continual imaging throughout their lives.
Options for imaging a stoma and peristomal skin are limited. Further, examination of images can require continuous imaging assessment by trained medical personnel and can be problematic for some ostomates, for example, younger ostomates and those who are unable to be in constant communication with medical professionals.
Accordingly, it is desirable to provide improved stoma imaging and measuring tools and methods to better understand, detect, and potentially prevent peristomal and stomal conditions to maintain health thereof and for improved ostomy appliance fittings.
Tools and methods of imaging a stoma are provided according to various embodiments to increase understanding of temporal changes in stoma topography, dimensions, and volume which can guide improved ostomy product development. Post-surgical swelling and tissue remodeling causes changes in stoma size and shape over time. Understanding these changes can lead to treatment and appliance fitting algorithms, improvements in patient education, and novel product designs. The stoma imaging and measurement tools and methods may be configured to enable developers and clinicians to better understand, diagnose, and predict stoma qualities and conditions via computational image analysis.
In one aspect, a stoma imaging method may comprise obtaining a 2D image of a stoma and generating a 3D image of the stoma from the 2D image. In an embodiment, the 2D image may first be converted into a grayscale and plotted into the 3D image. In another embodiment, the 2D image may be converted into a holographic image, which may be converted to the 3D image. In some embodiments, multiple 2D images (e.g. rapidshot) or fast acquisition method may be used.
In an embodiment, the stoma imaging method may further include determining at least one dimension of the stoma from the 2D image, which may be performed using an algorithm including a bounding box method to determine a height and a width of the stoma in real time. The height and/or width measurements may be used to calculate a volume of the stoma and/or to assign a shape classification to the 2D image (See
Further, changes in dimensions, shape, and/or volume of the stoma over time may be collected, analyzed, and used to instruct treatment algorithms, product designs, and/or product recommendations. The collected data may also be used to instruct predictive algorithms, for example, artificial intelligence, to guide users to alter their application procedures, product selection (for example, convexity), and/or product dimensions to prevent peristomal skin complications.
In an embodiment, the stoma imaging method may include generating contours of the stoma using the 2D image. In some embodiments, the method may analyze and determine margins of the stoma and peristomal skin using predefined thresholds, or may use a contralateral skin to guide an algorithm (see
In an embodiment, a topography of a peristomal skin may be imaged, measured and modeled using the foregoing methods described regarding quantifying the stoma.
The step of plotting the grayscale into the 3D image may include determining a depth of each pixel in the grayscale from an intensity value of the pixel, wherein the intensity value may range from black to white. The intensity value may be plotted as a z-axis in the 3D image. In some embodiments, the stoma imaging method may further include a scaling process, wherein the z-axis is scaled based on estimated pixel-based height and width measurements.
In some embodiments, the stoma imaging method may also determine the color of the stoma and/or peristomal skin from the 2D image. The color of the stoma and/or peristomal skin may be measured using the foregoing imaging and analysis methods described regarding quantifying the stoma. The color values may be measured using the HSV (Hue Saturation Value) Color Scale and/or may incorporate a calibrating color scale. In some embodiments, the imaging/thresholding method may also be used to determine the color of individual's skin from the 2D image.
In an embodiment, redness is defined using a specific hue range in the HSV colorspace. In this embodiment, a redness filter identifies ranges of red colors from a specific background. A red hue may have pixel values in the ranges [0,10] and [160-180]. The redness filter may be, for example, a Python OpenCV filter. The redness filter may perform differently based on a use of a flash when taking the stoma image.
In an embodiment, the stoma imaging method may further include a 3D image clean-up step using a background subtraction process. The stoma imaging method may include a stoma or surrounding peristomal skin identification step using a background or contralateral skin image subtraction process.
In another aspect, a stoma imaging method may comprise obtaining a 2D image of a stoma and determining at least one characteristic of the stoma from the 2D image. In an embodiment, the at least one characteristic of the stoma may be determined using an algorithm that includes a bounding box method to identify a height and a width of the stoma in a real time. The stoma imaging method may also generate contours of the stoma using the 2D image and determine the color of the stoma from the 2D image. For example, the stoma imaging method may be configured to determine a redness of the stoma and peristomal skin and/or temporal changes in the pigmentation of the peristomal skin, which may be indicative of adaptation to serial mechanical or chemical injury.
In embodiment, the 2D image may be quantified using the HSV Color Scale, which provides a numerical value that corresponds to color in degrees from 0 to 360.
In an embodiment, the stoma imaging method may include a capture and averaging of multiple images to increase resolution, decrease noise or reduce artifacts.
In another aspect, an imaging method may comprise obtaining an image of an anatomical site, obtaining a mask using a machine learning model trained to identify a stoma in the image, obtaining contours of the stoma based on the mask, and obtaining a contoured image of the anatomical site based on the contours of the stoma. The anatomical site includes a stoma.
In an embodiment, the machine learning model may use quantitative redness analysis to identify the stoma in the image. In some embodiments, the quantitative redness analysis may include analyzing red hue values of the image to identify a range of red objects from a background. The quantitative redness analysis may use a gaussian fit of redness peaks to generate parameters related to the stoma. The parameters may include a shape and perimeter of the stoma.
The foregoing methods and devices may also be used to quantify additional anatomical sites and indications, for example, but not limited to, the quantification of healing and/or closure of wounds and surgical sites, temporal changes of percutaneous interfaces (for example, a gastrostomy tube, tracheotomy, or osseointegrated prosthetic attachment hardware), subcutaneous implants, and/or the change in dimensions of soft tissue, as in tissue expansion for post-mastectomy staged breast reconstruction or transposition flap creation.
The foregoing methods and devices also be used to quantify additional pigmentation or color-changing applications and indications, for example, but not limited to, quantification of erythema and/or inflammation from injury (for example, microdermabrasion or fractional laser resurfacing), development and resolution of skin reactions to topical or systemic drugs, discoloration of scars or keloids, progress of tattoo removal, or treatment of hyperpigmentation, port wine stains, or vitiligo.
Other objects, features, and advantages of the disclosure will be apparent from the following description, taken in conjunction with the accompanying sheets of drawings, wherein like numerals refer to like parts, elements, components, steps, and processes.
While the present disclosure is susceptible of embodiment in various forms, there is shown in the drawings and will hereinafter be described one or more embodiments with the understanding that the present disclosure is to be considered illustrative only and is not intended to limit the disclosure to any specific embodiment described or illustrated.
Imaging of a stoma and peristomal skin can help better understand and detect dermatological complications in peristomal skin and changes in stomal shape, size, and positions to promote improved ostomy appliance fitting and maintain ostomy skin health. Common challenges associated with stomas include skin irritation and leakage, which make self-care post-operation difficult to manage. Improper stoma installation and site selection are considered one of the most common and early complications of stoma surgery. It is more difficult to use stoma appliances when this occurs, due to injury at the attachment site, and leakage is more likely to happen.
Using 3D modeling of a stoma to find cases in which there is improper site selection can assist with providing insights for surgeons, enterostomal nurses, patients, caregivers, and product designers. 3D modelling may also be used to predict vascular compromise arising in the long term from stoma failure, which can cause intestinal and stomal necrosis, among others. Further, 3D modelling may assist in determining whether there have been complications that have latent symptoms in patients.
Automated imaging methods according to various embodiments of the present disclosure may be configured to reduce inconveniences and problems of the known imaging methods, to provide clinical research tools, and to detect abnormalities before an onset of symptoms. Sporadic imaging or scheduled imaging of a stoma may enable medical professionals to recognize latent or subclinical symptoms. Further, the imaging method may be configured for easy data collection, which may be performed by an ostomate. Such ease of data collection can increase data collection across demographics and allow for non-invasive longitudinal studies.
In an embodiment, a stoma imaging method may comprise collecting and analyzing two-dimensional (2D), three-dimensional (3D) and pigmentation information of a stoma and/or the surrounding peristomal skin.
2D Parameters
In an embodiment, a stoma imaging method may be configured to obtain a 2D image of a stoma and peristomal skin area and determine various dimensions and characteristics thereof from the 2D image. For example, the stoma imaging method may be configured to estimate width, height, area, and volume of the stoma and peristomal skin, depth/height of the stoma, color (e.g. redness) of the stoma and peristomal skin, and any changes thereof.
In an embodiment, the stoma imaging method may comprise an algorithm including a bounding box method to find height and width dimensions in real time. The algorithm may be configured to determine dimensions of a target on an object with known dimensions. In an embodiment, the algorithm may be configured to determine various dimensions of a stoma from a 2D stoma image in real time. The bounding box method, for example, may include determining the outer boundaries of a stoma in an image. The outer boundaries of the stoma may be determined using various methods including determining edges in the image based on a color analysis.
In an embodiment, the stoma imaging method may apply the bounding box method from within OpenCV package or other similar codes to determine a shape of the stoma. From the selected shape, distances in pixels may be measured calibrated to the real-world measurements of the target stoma model. The same steps may be applied to several target stoma models to create a mapping. Table 1 shows an example mapping.
In an embodiment, the stoma imaging method may be configured to capture color or pigmentation of a stoma to identify a condition of the stoma health. For example, the stoma imaging method may be configured to determine the redness of the stoma and/or peristomal skin. In such an embodiment, the stoma imaging method may provide an algorithm configured to define a red color as a specific hue range in the HSV colorspace. The algorithm may comprise a Python OpenCV redness filter or other similar codes configured to identify ranges of red from a specific background. The stoma imaging method may be configured to take a 2D image of the stoma using a flash and ambient shadow and lighting effects minimized to optimize the redness determination. In some embodiments, the stoma imaging method may be configured to analyze the redscale of an RGB image or HSV color scale to map a change in the redness of the stoma and peristomal skin area over time. A change in the redness of the stoma and/or peristomal skin may be used to predict or identify stoma or peristomal skin complications.
In an embodiment, the stoma imaging method may comprise the step of generating contours of a stoma. The stoma contours may be generated from a single image. By generating a contour, a qualitative description of the stoma shape, as well as quantitative information concerning the peristomal area and circumference of the stoma may be obtained. The step of generating contours of a stoma may include obtaining an image having a sufficient contrast. The image of a stoma having the sufficient contrast may be taken using a flash.
The stoma contour representation may utilize K-means clustering that generates “cluster centers” based on a set of initial guesses and determines in an N-dimensional space how close certain points are to these initial guesses. After iteratively updating the position the cluster centers are updated to be closest to the “true clusters” in a dataset. The distance metric may include the following exploratory variables: contour length, contour center, redness sum, redness centroid (Gaussian fit), redness background height. A training data may be generated from the stoma models. This analysis may be performed using about 100 images for each stoma taken in the same conditions. Other forms of clustering may be utilized, like balanced iterative reducing and clustering using hierarchies (BIRCH), Clustering Using Representatives (CURE), hierarchical, expectation-maximization (EM), mean shift, density-based spatial clustering of applications with noise (DBSCAN), ordering points to identify the clustering structure (OPTICS), and others. A multiple linear regression algorithm, structure prediction models like graphical models may also be used.
In an embodiment, the stoma imaging method may compare an image of a peristomal skin area against an image of a “normal abdominal skin area” to identify margins of a stoma as shown in
In an embodiment, the stoma imaging method may be configured to track changes in an area and shape of the stoma, which may be indicators of the stoma health and/or normal anatomical post-surgical healing and remodeling. The changes in the area and shape of the stoma may be signs of infection and other stoma conditions, or a normal remodeling process as the stoma matures.
2D to 3D Modeling
In an embodiment, the stoma imaging method may further comprise the step of generating a 3D model from a 2D image of a stoma. The step of generating 3D model may include converting an RGB image into a grayscale via an algorithm. The depth of each pixel in the grayscale may be ascertained from the intensity value of the pixel, which ranges from black to white. The grayscale may then be plotted into a 3D image, with the pixel intensities acting as a z-axis. The 3D modeling step may further include a scaling process, wherein an algorithm may take in estimated pixel-based height and width measurements and scales the z-axis to these numbers. In some embodiments, the 3D modeling step may analyze the redscale of an RGB image in conjunction with the grayscale to generate a 3D image. The stoma imaging method may be configured to provide a 3D image that looks qualitatively similar to the target stoma.
In an embodiment, pictures of a stoma in different lighting conditions and using different stoma color/background color contrasts may be taken to better gauge the effectiveness of the grayscale method in different conditions. To ensure fair assessment of images and to obtain useful data, a testing rig was created to keep the height, image quality, and image stability consistent across images, while changing the parameters of interest. About 50 useful comparison images were generated and analyzed to identify critical parameters to optimize the 3D model quality.
The machine learning process may include using trained machine learning models that identify a stoma in an image. The machine learning models may include deep learning models that utilize neural networks to process the image based on redness values. The machine learning models may be trained, for example, using several hundred images of a stoma at different angles and under different lighting conditions or using a still frames of a video of a stoma. The machine learning models may be trained using approximately 80% of the total images while reserving 20% of them for testing.
Quantitative redness analysis may be used to analyze the redness value of the stoma image in order to generate a mask. The analysis may include looping through values in a redness range (160, 10), checking how the mask sum changes and taking the largest sum. The mask sum is then analyzed to determine the best mask for identifying a stoma in an image. A machine learning model may be used to generate the mask using the quantitative redness analysis. Redness analysis may use a gaussian fit of redness peaks to generate precise parameters related to the stoma (shape and perimeter of the stoma). The redness can be quantized by taking a single value mask (i.e. the range [160,161]) and iteratively applying these masks to the stoma image. By summing the resulting mask values (which become a 0 or 1) the amount of each “color” in the image may be quantified.
A gaussian model may include the following equation (1), with a resulting gaussian fit model with the image of a stoma having a yellow background, as shown in
Gauss(x)=ae−((x−b)
A double-Gaussian distribution can be used when the stoma has certain backgrounds that make it difficult to identify the contours of the stoma. The double-Gaussian distribution includes the following equation (2), with a resulting double gaussian fit model with the image of a stoma having a yellow background, as shown in
DGauss(x)=Gauss(a,b,c)+Gauss(d,e,f)+gx+h (2)
In an embodiment, a contour method may be used to generate a stoma image with contours for identifying the outline of a stoma. An image of a stoma, as shown in
In step 1101, the computing device may obtain a 2D image of an anatomical site. The 2D image may also be a photograph or video of the stoma at a certain distance and under certain lighting conditions. The 2D image may also be, for example, a CT, MRI or other medical image of the stoma.
In step 1102, the computing device may convert the 2D image into grayscale. The grayscale, for example, may be a grayscale image.
In step 1103, the computing device may plot the grayscale into a 3D image. The grayscale may be done based on the intensity of the grayscale at each pixel. The grayscale can be used to generate contours of the stoma in the image.
In step 1201, the computing device may obtain a 2D image of an anatomical site. The 2D image may also be a photograph or video of the stoma at a certain distance and under certain lighting conditions. The 2D image may also be, for example, a CT, MRI or other medical image of the stoma.
In step 1202, the computing device may obtain a redness mask of a stoma in the 2D image. The computing device, for example, may include a machine learning algorithm trained to obtain the redness mask based on a redness range. The machine learning algorithm may use redness analysis to identify a range of red objects from a background.
In step 1203, the computing device may generate a contour of the stoma based on the redness mask. The computing device, for example, may use a shape and contour analysis to find a shape and perimeter of a stoma.
In step 1301, the computing device may obtain an image of an anatomical site. The anatomical site comprises a stoma. The anatomical site, for example, may include a stoma surrounded by red marks resulting from complications, as shown in
In step 1302, the computing device may obtain a mask using a machine learning model trained to identify the stoma in the image. The machine learning model, for example, may be trained to find a range of redness values that can generate a mask for identifying a stoma in an image. The machine learning model, for example, may also be trained to use a grayscale image to generate a mask for identifying a mask for stoma in an image.
In step 1303, the computing device may obtain contours of the stoma based on the mask. The mask, for example, may be used to separate the stoma from the background of the image. For example, in the image, a stoma may be separated from red marks on the anatomical site.
In step 1304, the computing device may obtain a contoured image of the anatomical site based on the contours of the stoma. The contoured image, for example, may be an image of a stoma with an outline of the edges of the stoma, as shown in
The processor 1422 typically controls overall operations of the computing system 1420, such as the operations associated with the display, data acquisition, data communications, and image processing. The processor 1422 may include one or more processors to execute instructions to perform all or some of the steps in the above-described methods. Moreover, the processor 1422 may include one or more modules that facilitate the interaction between the processor 1422 and other components. The processor may be a Central Processing Unit (CPU), a microprocessor, a single chip machine, a GPU, or the like.
The memory 1424 is configured to store various types of data to support the operation of the computing system 1420. Memory 1424 may include predetermine software 1426. Examples of such data comprise instructions for any applications or methods operated on the computing system 1420, video datasets, image data, etc. The memory 1424 may be implemented by using any type of volatile or non-volatile memory devices, or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic or optical disk.
The I/O interface 1428 provides an interface between the processor 1422 and peripheral interface modules, such as a keyboard, a click wheel, buttons, and the like. The buttons may include but are not limited to, a home button, a start scan button, and a stop scan button. The I/O interface 1428 can be coupled with an encoder and decoder.
Communication Unit 1430 provides communication between the processing unit and an external device. The communication can be done through, for example, WIFI or BLUETOOTH hardware and protocols. The communication unit 1430 may communicate with a CT machine, MRI machine, medical device, photo camera, video camera or other imaging system to obtain or capture an image for processing.
User interface 1440 may be a mobile terminal or a display.
In some embodiments, there is also provided a non-transitory computer-readable storage medium comprising a plurality of programs, such as comprised in the memory 1424, executable by the processor 1422 in the computing system 1420, for performing the above-described methods. For example, the non-transitory computer-readable storage medium may be a ROM, a RAM, a CD-ROM, a magnetic tape, a floppy disc, an optical data storage device or the like.
The non-transitory computer-readable storage medium has stored therein a plurality of programs for execution by a computing device having one or more processors, where the plurality of programs when executed by the one or more processors, cause the computing device to perform the above-described method for motion prediction.
In some embodiments, the computing system 1420 may be implemented with one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), graphical processing units (GPUs), controllers, micro-controllers, microprocessors, or other electronic components, for performing the above methods.
It is understood that the relative directions described above, e.g., “upward,” “downward,” “upper,” “lower,” “above,” “below,” are used for illustrative purposes only and may change depending on an orientation of the ostomy pouch and/or the patient. Accordingly, this terminology is non-limiting in nature. In addition, it is understood that one or more various features of an embodiment above may be used in, combined with, or replace other features of a different embodiment described herein.
All patents referred to herein, are hereby incorporated herein in their entirety, by reference, whether or not specifically indicated as such within the text of this disclosure.
In the present disclosure, the words “a” or “an” are to be taken to include both the singular and the plural. Conversely, any reference to plural items shall, where appropriate, include the singular.
From the foregoing it will be observed that numerous modifications and variations can be effectuated without departing from the true spirit and scope of the novel concepts of the present invention. It is to be understood that no limitation with respect to the specific embodiments illustrated is intended or should be inferred. The disclosure is intended to cover by the appended claims all such modifications as fall within the scope of the claims.
This application is based upon and claims priority to U.S. Provisional Application No. 63/157,290 filed on Mar. 5, 2021, the entire contents thereof are incorporated herein by reference in their entireties for all purposes.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/US2022/018907 | 3/4/2022 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 63157290 | Mar 2021 | US |
