Diagnosis and monitoring of red blood cell diseases (such as sickle cell disease “SCD”) is expensive, difficult, and requires skilled personnel. There is the need for someone to develop a point of care technology that is fast, easy to use, and very affordable. Current solutions include hemoglobin (“HB”) electrophoresis, high-performance liquid chromatography (“HPLC”), microscopy-based processes, and a SICKLEDEX® test available from Streck, La Vista, Nebr. HB electrophoresis can differentiate between sickle cell trait and disease; however it is expensive and requires a skilled operator. The same can be said for HPLC. Microscopy-based tests and SICKLEDEX®, whilst affordable, cannot differentiate the various genotypes of sickle cell disease. In addition, none of the above technologies are platform technologies and they are not useful in patient monitoring.
Wide-field digital interferometry (“WFDI”) is a technique that provides quantitative measurements of optical path delays (“OPDs”) associated with optically transparent samples. The process works by recording the pattern of interference between the interaction of light with a sample (in this case the red blood cells, “RBCs”) and a mutually coherent reference wave. The process provides a quantitative phase and amplitude profile of the sample.
By way of background, U.S. Pat. No. 8,508,746, patented Aug. 13, 2013, to Duke University, is incorporated by reference herein in its entirety.
Some embodiments of the present disclosure are directed to an interferometry system including an interferometric chamber (“InCh”) as an alternative approach for recording the dynamics of transparent biological samples. In some embodiments, the system is configured to perform common-path interferometry wherein the beam is split by the InCh itself at the desired angle. As a result, no special optical elements are required in the path of the beam and interferometric alignment can be performed once, e.g., during the fabrication of the chamber, and not each time before the measurement, further simplifying the process. The system is effective for identifying hemolytic anemias, e.g., sickle cell disease and malaria, within a patient sample.
The following presents a simplified summary of one or more aspects in order to provide a basic understanding of such aspects. This summary is not an extensive overview of all contemplated aspects and is intended to neither identify key or critical elements of all aspects nor delineate the scope of any or all aspects. Its purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed description that is presented later.
In an aspect of the disclosure, an imaging system for imaging a fluid sample is disclosed. The imaging system includes a light source configured to generate a beam of light, an angled element disposed along an optical path of the beam of light, a sample cartridge holder configured to receive a sample cartridge and configured to hold the sample cartridge in a first position in which an imaging region of the sample cartridge is disposed along the optical path, and a sensor configured to capture the beam of light after the beam of light passes through the angled element and the imaging region of the sample cartridge. The imaging region of the sample cartridge is configured to receive the sample fluid.
In another aspect of the disclosure, a sample cartridge is disclosed. The sample cartridge includes a cover plate comprising a sample fluid inlet and a fluidics layer. The fluidics layer includes an opening configured to receive a whole blood sample from the sample fluid inlet and an imaging region configured to receive the whole blood sample from the opening through a fluid channel. The sample fluid inlet, the opening, the fluid channel, and the imaging region are configured to promote a directional flow of the whole blood sample through the imaging region.
The appended drawings illustrate only some implementations and are therefore not to be considered limiting of scope.
Referring to
The sample holder 120 is configured to hold a sample therein and may include a front cover slip 122 and a back cover slip 124. The front cover slip 122 is configured to allow a first portion of the illumination beam 104 to transmit therethrough where it may interact with the sample in the sample holder 120 as it propagates to or from the back cover slip 124. At least part of the first portion of the illumination beam 104 exits the sample holder 120 through the front cover slip 122 after interacting with the sample. This part of the first portion of the illumination beam 104 propagates through the system 100 as a sample beam 130 toward a sensor of a digital camera 110 where the sample pattern of interference is captured. The front cover slip is further configured to reflect a second portion of the illumination beam 104 at an angle relative to the optical axis 112. The second portion of the illumination beam 104 does not interact with the sample and is propagated through the system 100 as a reference beam 126 toward the sensor of the digital camera 110 where the reference pattern of interference is captured. Thus, the sample pattern may be compared to the reference pattern to obtain a quantitative phase and amplitude profile of the sample. Additional details regarding the system 100 are available in U.S. Pat. No. 8,508,746, patented Aug. 13, 2013, to Duke University, incorporated by reference herein in its entirety.
Referring now to
In some embodiments, the optical system includes one or more interferometers. In some embodiments, the interferometers are configured for common-path interferometry. In some embodiments, the system includes at least one common-path interferometer. In some embodiments, the interferometer includes one or more cameras, light sources, beam splitters, light receiving modules, imaging modules, etc. In some embodiments, the system is a fully standalone device with a single board computer, a case enclosing the interferometers, one or more displays (e.g., touch screens), sensors (e.g., flow sensors), etc. In some embodiments, the system includes a non-transitory computer storage media coupled with a computing device and encoded with one or more computer programs, e.g., an artificial intelligence (“AI”) algorithm that automates the system, simplifies the diagnosis and interpretation of results, displays the results and/or a graphical user interface (“GUI”) on the screens (e.g., without the need to connect the system to additional peripherals), etc. In some embodiments, the interferometer includes a sample staging module, as will be discussed in greater detail below. In some embodiments, the system overlaps reference and sample beams, so that the same vibrations occur for both beam paths. This overlap of reference and sample beams advantageously reduces measurement errors in the phase profile associated with instability in the interferometric system, including differential vibrations or air perturbations in the interferometer arms. Thus, the setup may be used in ambient conditions and in very low-resource settings, where vibration-isolating optical tables are inaccessible.
Referring now to
The position of the sample staging module 300 may be adjustable. For example, the stage 302 may be moved in the positive y-direction as indicated by the arrows 310a, 310b. In some embodiments, the stage 302 may be moved in the x-, y-, and/or z-directions such that the sample cartridge inserted therein can be imaged by an interferometry system (not shown) in which the staging module 300 is included. The stage may be motorized and may be adjusted manually by a user or automatically by a motorized system. For example, the stage may be adjusted based on initial imagery collected during the interferometry process or based on results of a calibration process.
In addition to an adjustable stage, the position of the angled element 308 may be adjustable. For example, the element 308 may be moved toward the sample cartridge 306 (e.g., in the negative y-direction) as represented by arrows 312a, 312b. By moving the stage in the positive y-direction and/or moving the angled element 308 in the negative y-direction, the sample cartridge slot 306 (and thus, the sample cartridge contained therein) may be substantially sandwiched between the angled element 308 and the reflective element 304 such that air gaps between the components are minimized or eliminated. In some embodiments, the position adjustments of the angled element and/or the stage are completed during an initial calibration step that does not need to be completed each time the interferometry system is used. By eliminating the need to re-calibrate the system each time it is used, the sample staging module 300 disclosed herein may substantially improve throughput of samples through an interferometry system.
The sample cartridge 306 is configured to reversibly accept sample fluid cartridges. In some embodiments, the sample cartridges 306 configured to insert into the cartridge slot are disposable and may be a one-time use item.
An example embodiment of a sample cartridge slot frame 408 is illustrated in
As discussed above with respect to
Referring to
Referring now to
The cartridge 510 may include one or more fluidic connectors that may be attached to the cartridge. In some embodiments, an additional slab of PDMS (not shown) is attached on or just over the inlet 512 and outlet 516 to the cartridge. In some embodiments, the syringe pump may be a Fusion 4000 pump available from Chemyx Inc., Stafford, Tex. or any syringe pump configured to control fluid flow at two different flow rates. Alternatively, a sample loading system may include two different syringe pumps, e.g., two Fusion 200 pumps also available from Chemyx Inc. The system 500 further includes syringes, tubing to connect syringe to device depending on the type of syringe, and tubing (e.g., from Fluigent, North Chelmsford, Mass. or McMaster-Carr, Santa Fe Springs, Calif.), metal connectors to connect tubing to system, etc.
As discussed above, in some embodiments, an interferometry system may be operated without first separating the cells in a sample. In some embodiments without cell pre-sorting, the sample fluid cartridges, such as cartridge 400 shown in
In some exemplary embodiments, a drop of blood sample is diluted in prefilled tubes (e.g., Eppendorf tubes with prefilled PBS) prior to being loaded into a cartridge for imaging. In some embodiments, the undiluted sample is loaded into the cartridge and the cartridge is inserted into the sample cartridge slot. In some embodiments, the undiluted sample is then imaged. The cartridge is then removed, and the system is ready to receive the next cartridge for imaging. This exemplary embodiment is advantageous in that it uses less time than traditional interferometry processes, is simple to use, and does not require access to expensive lab equipment. Additional fluidic components can be eliminated or reduced, and the user does not need to replace fluids and waste product containers. Thus, the need to prime channels or clean tubing, e.g., via alcohol or deionized “DI” water rinse, may also be avoided. The need for additional fluidic control programming in the interferometry system, such as sensing and indicating to an operator that the waste container is full, a reagent is running low, a pump requires priming, etc., may also be reduced.
In some sample preparation methods, the target components within the sample fluid are sorted prior to being imaged. In some embodiments, the components are sorted within the system itself. There are several approaches to conducting cell separation, many involving external forces, such as electric field, acoustics, centrifugal, etc. In some embodiments, hydrodynamic separation based on size is performed in order to keep the device simple, e.g., according to Yamada, et al. (Anal. Chem 2004). Such a method may be used to separate and collect cells for imaging.
In some exemplary embodiments, 1-2 drops of blood are prepared. A pump is provided with at least two different channels: one for blood sample, and one for PBS. In some embodiments, the blood samples were provided to prefilled Eppendorf tube and mix (dilution of 1:10 in PBS). The samples were loaded into a syringe, e.g., 100 μL syringe, and syringe placed into the syringe pump. In some embodiments, the fluidic system is primed. In some embodiments, the fluidic system is connected to a cartridge. The syringe pump/fluidic system then conducts the cell separation of the blood sample. Cells are then flowed through the cartridge where they are imaged by the interferometer as discussed above. In some embodiments, the sample is flowed through the system for about 1-5 mins. The fluidic system may then be unplugged. This exemplary embodiment is advantageous in that the system can be fully automated. Further, because the cells are separated, there is reduced need for algorithms to identify cells in the sample, making looking at cells easier.
In one exemplary embodiment of the present disclosure, the following materials were provided:
Cartridge frames were 3D printed. Glass slides and glass cover slips were coated using fluorinated solution (Note: once glass is prepared, it can be re-used provided the hydrophobic properties are retained). In this example, the glass surfaces were plasma O2-treated at 100 W for 30 seconds, then immediately immersed in liquid silane solution (e.g., 5% v/v fluorinated silane in EtOH) for one hour at room temperature. The glass pieces were then rinsed with anhydrous EtOH, followed by DI water, followed by EtOH (3×). Finally, the samples were dried with compressed nitrogen and were heated in an oven at 60° C. overnight at ATM pressure.
A PDMS solution was then prepared. A 10:1 v/v ratio of monomer to crosslinker from PDMS kit was used. The solution was mixed very thoroughly (e.g., for approximately two minutes), and was degassed using the vacuum/desiccator chamber for approximately 20 minutes.
Referring now to
The glass cover slip was taken from the top. A scalpel was used to disconnect parts of PDMS where needed. The cartridge frame was taken out, now with the PDMS bottom portion attached thereto.
Third-party lithography services were used to create SU-8 molds (e.g., molds fabricated using SU-8 epoxy-based photoresist) on silicon wafers, e.g., following the steps shown in
The PDMS was then peeled from the master wafer. Holes were punched on the inlet and outlet of the top PDMS chamber, followed by O2 plasma treatment of the PDMS pieces. The top PDMS piece was aligned to the bottom and pressed down gently to ensure good contact. The cartridge was left overnight at 60° C. to secure the PDMS bonding.
Referring to
The cover plate 900 includes fluid inlets 902 configured to receive a fluid (e.g., a sample fluid) into the assembled sample cartridge. In some embodiments, the sample fluid may be a diluted sample including whole blood, PBS, and an anticoagulant mix. The ratio of whole blood to PBS and anticoagulant may have an optimal range depending on the design and size of fluidics features within the sample cartridge. For example, as discussed above, it may be advantageous to keep the fluidics features, such as fluid channels and mixing regions, as shallow as possible so that cells are near a focal plane of the interferometry system. Diluting the whole blood collected from a patient to a desired ratio of whole blood to PBS and anticoagulant may help prevent blockages within the fluidics features. In some embodiments, a small jar or container having a desired amount of PBS and anticoagulant pre-loaded therein may include an indicator mark to show a user how much sample should be added to the jar to achieve the desired amount of sample dilution.
Sample fluid or diluted sample fluid may be introduced to one or more fluid inlets 902. In some embodiments, a sample holder 904 (e.g., an open-ended container such as a cylinder, tube, bowl, or other container) may surround one or more fluid inlets 902 and is configured to hold the volume of fluid sample. The sample holder 904 is configured to hold the sample fluid such that gravity pulls the fluid down and creates a hydraulic pressure in the fluid that helps to push the sample fluid from the sample holder into the inlets 902. The pressurized fluid then flows through the sample cartridge and imaging is performed on the flowing sample, as will be discussed in further detail herein. Using gravity to pressurize fluid and promote flow of the sample fluid through the cartridge may eliminate the need to include pumps in some embodiments.
From each of the one or more fluid inlets 902, pressurized sample fluid (e.g., pressurized using passive gravity-based hydraulic pressure and/or using active pumping) flows into one or more fluid channels 1006 via openings 1008 disposed within the fluidics layer 1000. The openings 1008 are in fluid communication with the inlets 902 on cover plate 900. The fluid channels 1006 may fluidly communicate with a common mixing channel 1010 that directs sample fluid to an imaging region 1012. Referring to
In some embodiments, a portion of the fluid is pushed through a plurality of fine channels 1014 toward one or more side chambers 1016. While not required, the fine channels 1014 may act as a filter by preventing large particles or cells within the sample fluid from reaching the side chambers 1016. For example, the fine channels 1014 may have a height and a width of approximately 12 mm such that particles or cells having a dimension larger than 12 mm (e.g., white blood cells “WBCs”) may be prevented from reaching the side chambers 1016. This filtering mechanism may assist with separation of cells for easier identification and differentiation during interferometry image processing. In embodiments that do not include a filtering mechanism, differentiation of cells may be accomplished using computer vision.
Referring to
By performing imaging on the sample as it is circulating, more of the sample cells can be viewed and more data can be collected for analysis and diagnosis. This may be particularly advantageous when searching for abnormal cells that make up a relatively low percentage of a patient's cells. For example, in patients with malaria, very few red blood cells may carry the parasite. Thus, many red blood cells must be imaged to detect the parasite. Imaging the flowing sample instead of imaging a smear or other static sample may significantly decrease the amount of time required to image a large number of cells. Furthermore, because all cells in a flowing sample may be imaged as they pass through an optical path of the interferometry system, a lower volume of fluid may be collected from the patient compared to current procedures using a static sample. For example, only a finger prick and between 1-24, of blood from the patient may be needed in the flowing system facilitated by the disclosed sample cartridge. By comparison, the current imaging process which may require multiple sample smears on multiple slides for imaging, typically requires collection of a much larger volume of blood from the patient using a much larger needle and often a tourniquet.
Moreover, because traditional systems do not make use of computer vision and machine learning programs to identify and differentiate different types of healthy and unhealthy components within a sample (e.g., red blood cells, white blood cells, platelets), the components of a sample must be sorted prior to imaging. The sorting process is generally time-intensive and requires access to lab equipment and supplies (e.g., large needles to collect samples, ethylenediaminetetraacetic acid “EDTA” tubes, Ficoll, a conical tube, a centrifuge, PBS, syringes, male Luer fluid connectors, silicone tubing, waste containers, etc.). Additionally, the separating process must be performed by a trained professional. Thus, the current imaging process is not suited for use in low-resource areas where lab equipment and trained professionals are scarce. Furthermore, because the sorting process takes generally at least 40 minutes just to prepare a sample for imaging, throughput using this method is very low.
With the disclosed sample cartridges and computer vision-assisted image processing, the entire process of pre-sorting cells may be eliminated. This drastically simplifies complexity of the imaging process, reduces the number of required supplies, reduces cost associated with imaging, reduces amount of time needed to obtain imaging results, and does not require a trained professional to perform various steps associated with pre-sorting a sample. The computer vision portion of image processing may use the interferometry images collected on the flowing sample to obtain information about components (e.g., red blood cells, white blood cells, platelets, etc.) within the sample. For example, information about red blood cell shape, membrane flexibility, sickle features, percentage of sickling, and other parameters may be collected. The images may also be used to identify other morphological changes to cells that are indicative of different types of diseases, such as sickle cell anemia or malaria.
Direct current voltage may be applied to the sample cartridge via the electrodes using a voltage source (not shown) to perform electrophoresis on the sample fluid within the sample cartridge. In some embodiments, the voltage applied may be between approximately 0.2V and approximately 5V. The voltage applied may be determined as a function of the pH level and contents of the sample being directed through fluid channels in the fluidics layer. For example, whole blood having red blood cells that have been broken apart (e.g., by a lysing reagent prior to entering the sample cartridge) such that hemoglobin contained therein is released from the RBCs and may be imaged. In some embodiments, the fluidic layer 1400 in sample cartridge may include one or more filter membranes (not shown) and/or fine channels (e.g., similar to fine channels 1014 in fluidics layer 1000) to prevent lysed RBC fragments and other debris from entering an imaging region while allowing hemoglobin to pass through. When voltage is applied to the fluidics layer 1400, the hemoglobin may separate into bands within the imaging region 1412. The bands, rather than individual cells or components, are imaged using a high-resolution interferometry system. The high-resolution imaging system may capture data (e.g., hemoglobin separation under voltage charged conditions) that can be analyzed for making diagnostic predictions.
While the example described above includes the step of lysing RBCs in a sample prior to introducing the sample to the cartridge, this step is not required. In alternative embodiments, a cellulose acetate (CA) membrane embedded with ammonium chloride (NH4Cl) and potassium bicarbonate (KHCO3) in part of the cartridge 1300 that is exposed to the sample. This membrane may break down RBCs after the unprocessed sample is introduced to the cartridge, thereby eliminating the need for a professional to perform a separate sample lysing step.
The various sample cartridges described above provide cost, time, resource, and complexity savings with respect to current sample handling techniques. Additional advantages can be realized when the sample cartridges are used with an interferometry system having a cartridge holding slot. Such a system is illustrated in
As discussed above, the image analysis and diagnostic aspects of the present system may be further refined and improved by implementing machine learning in the software used in identifying target objects, such as a sickled red blood cell in the case of sickle cell disease diagnostics. The image analysis and recognition software used in the diagnostics may be trained using training sets prior to installation on the imaging system, and improved software may be uploaded to the imaging system as further refinements are made in the machine learning training sets and resulting software.
In particular, the convolutional neural network aspects of the image recognition and diagnostic software may be trained to efficiently and automatically recognize sickled RBCs or other diseases using training sets including interferometric images of sickled and healthy RBCs. For instance, the training inputs may include interferometric images of whole blood samples (e.g., processed with a saline solution for dilution) and known sickle cell disease diagnosis of those processed samples. The outputs from the training process may include, for instance, automated SCD diagnosis, an index of the health of a patient's RBCs, and trained neural network models that may be implemented in the software used with the imaging systems described above.
Referring to
Within training unit 1810, training system 1800 includes an image preprocessing unit 1812 for pre-processing interferometric images 1804. In image preprocessing unit 1812, a variety of processes may be implemented such as Fourier transform and/or inverse Fourier transform to extract phase images 1814 from interferometric images 1804, phase unwrapping, and image flattening.
Phase images 1814 may be processed in a variety of ways within training unit 1810. For example, each one of phase images 1814, with each image possibly containing multiple types of cells, may be directly processed by a cell type object detection block 1820, which draws a bounding box around each cell found in the image, classifying each cell into a cell type, such as a white blood cell (WBC), RBC, or a platelet. Alternatively, each one of phase images 1814 may be processed by a cell type instance segmentation block 1822, which labels each of the pixels within the phase image with the cell to which the pixel belongs, along with the cell type corresponding to that pixel. In an alternative process, a short video of phase images 1814 (e.g., of a few frames of the phase images or longer time frames) is created in a video creator block 1830. The short video may then be processed by cell type object detection block 1820 and/or cell type instance segmentation block 1822. As a further alternative, phase images 1814 may be processed by a simple segmentor 1860, which extracts images of cells from their background using, for example, threshold segmentation. Then, the images of cells so segmented may be processed by a cell type classifier block 1862, which identifies images of RBCs.
The results of cell type object detection block 1820, cell type instance segmentation block 1822, and cell type classifier 1862 may then be processed by a sickled/not sickled classifier block 1870 for determining whether each RBC identified is sickled or not. Alternatively or in addition, a RBC health regressor block 1872 may be used to determine the relative health (i.e., analyses beyond sickle cell disease) of the identified RBC. For example, an identified RBC may not be sickled yet be affected by another condition. Thus, in some embodiments, RBC health regressor block 1872 may be used to determine the relative health of an RBC that is not necessarily identified as sickled. Additional analyses of the identified RBCs may be performed, such as analyzing the cell membrane flexibility of the identified RBC based on a review of the short video of phase images as generated with video creator block 1830. The “ground truth” to be used as the training basis for the analysis performed by sickled/not sickled classifier block 1870 and RBC health regressor 1872 may be obtained from, for example, manual analysis and diagnosis of sickle cell disease patients' blood by a trained cytologist. Additionally, purposely distressed RBC samples (e.g., RBCs treated with various levels of a stressing agent such as sodium metabisulfite) can also allow training of regression models that provide a quantitative index of the health of the analyzed RBCs at RBC health regressor 1872. Finally, the results of the SCD diagnosis and health of the analyzed RBCs may be provided to the user in a step 1880.
It is noted that similar training methods may be used to refine the machine learning processes for the diagnosis of other conditions, such as malaria. By providing training unit 1810 with image parameters specific to the diagnosis of other diseases such as malaria and other blood infecting pathogens. Training unit 1810 may be further modified to perform analysis of other components of whole blood samples, such as WBC counts and platelet health, by providing different classifier and segmentation blocks specific to those blood components. Similarly, training unit 1810 may be configured for learning to diagnose diseases that may be detectable by analysis other samples, such as urine or saliva.
An alternative method for training the machine learning algorithms used for SCD diagnosis is illustrated in
Continuing to refer to
Following steps 1930 and 1946, process 1900 proceeds to a step 1950 in which the analyzed phase images are processed to determine the sample classification by genotype. For example, SCD may present in a variety of different forms (e.g., hemoglobin SS, hemoglobin SC, hemoglobin SB+, etc.) and each form may be differentiated and classified using the imaging system and machine learning processes described herein. In some embodiments, differentiating between certain sickle cell genotypes may alternatively or additionally require the use of additional sample processing (e.g., electrophoresis) and/or imaging techniques (e.g., imaging and analysis of hemoglobin bands in a sample after additional sample processing). Finally, the analysis results are saved in database 1910 in a step 1960, and the process is repeated using different phase images that are specifically relevant to the disease being diagnosed.
Methods and systems of the present disclosure are advantageous to provide affordable and easy to use interferometry for diagnosis and monitoring of red blood cell diseases. These systems and methods do not require skilled personnel and are a platform technology, with the potential of being applied to multiple disease states in the blood. Setup is very simple and can be used in very low resource settings. The user experience is also simple and does not involve more than 3 steps. The process is label-free and therefore does not utilize staining. Additionally, no biological reagents are used.
As discussed above, the systems and methods of the present disclosure reduce measurement errors in the phase profile associated with instability in the interferometric system, including differential vibrations or air perturbations in the interferometer arms. Thus, the systems can be used in ambient conditions in very low-resource settings, where vibration-isolating optical tables are inaccessible.
Although the invention has been described and illustrated with respect to exemplary embodiments thereof, it should be understood by those skilled in the art that the foregoing and various other changes, omissions and additions may be made therein and thereto, without parting from the spirit and scope of the present invention.
The present application claims the benefit of copending U.S. Provisional Patent Application Ser. No. 63/061,820, filed Aug. 6, 2020 and entitled “Diagnostic Systems and Methods for Hemolytic Anemias and Other Conditions,” which application is incorporated herein by reference in its entirety.
Number | Date | Country | |
---|---|---|---|
63061820 | Aug 2020 | US |