Aspects of the present disclosure relate to a Raman spectroscopy system and method for enhanced accuracy identification of a type or nature of a sample or a tissue to which excitation energy (e.g., collimated illumination) is directed, such as an abnormal or apparently abnormal growth (e.g., as cancer). In particular, but not exclusively, aspects of the present disclosure enable real-time enhanced accuracy diagnosis of abnormal tissues such as gastrointestinal growths, in vivo and ex vivo.
It is highly desirable to accurately and rapidly characterize and identify the nature of tissues such as neoplastic gastrointestinal growths by way of minimally invasive techniques such as endoscopy. For instance, colorectal cancer (CRC) is a common disease with a high mortality rate when discovered in the late stage.
The identification and eradication of neoplastic polyps is one of the most important measures to reduce colorectal mortality and morbidity.
Differentiation between hyperplastic polyps that pose little or no risk of malignant transformation and adenomas with prominent malignant latency remains a clinical challenge using conventional colonoscopy techniques.
With over 1.2 million new cancer cases and 608,700 deaths estimated to occur annually, CRC is a major problem in the modern world. Early identification of precancerous polyps (i.e., adenoma) in the curable stages together with appropriate therapeutic interventions, such as polypectomies or endoscopic mucosal resections (EMRs), remain the most important measures to reduce colorectal mortality and morbidity. The existing colonoscopic approaches suffer from a number of fundamental clinical limitations. This is because conventional colonoscopy entirely relies on the visualization of gross mucosal features of polyps, such as pit patterns, vascular patterns, etc. and provides little or no bio-molecular information about the tissue. Current standards of care therefore recommend resection of all suspicious polypoid lesions or abnormal growths identified during colorectal examinations. This approach is labor intensive; results in high cost histopathological assessments; and results in unnecessary risk to the patient since one-third to one-half of all polypoid lesions turn out to be hyperplastic. Although the absolute risk of polypectomy is considered relatively small, it still remains the most common cause of complication, such as bleeding, perforation, etc. during colonoscopy. Taking into account the existing clinical challenges and recent introduction of widespread colorectal population-based screening programs, the need for advanced endoscopic approaches has never been greater. Recent research has thus been directed towards development of more sophisticated molecular imaging and spectroscopy techniques for improving in vivo diagnosis and analysis.
Evidence has confirmed that the application of ex vivo Raman spectroscopy on colon tissue specimens provides encouraging accuracies, ranging from 89% to 99% for discrimination between different pathological types (i.e., normal, hyperplastic polyps, adenoma and adenocarcinoma). This still requires the removal of the polyps in order for the Raman spectroscopy to identify the nature of the pathological types and thus has the same problems as the conventional endoscopic techniques mentioned above. In addition, there have been many barriers to the translation of Raman spectroscopy into in vivo clinical diagnostics.
These include technical limitations such as inherently weak tissue Raman scattering, lengthy acquisition times (>5 s) and a fundamental necessity for developing long (>1.9 m) miniaturized fiber-optic probe designs with low fused-silica interference, high collection efficiency and depth resolving capability. The probe needs to be made with low fused-silica as fused-silica has a strong Raman signal and fluorescence background which will interfere with the weak tissue Raman signal. To date these technical limitations remain unsolved.
Recent technological advances, including the development of rapid Raman spectroscopy techniques and miniaturized fiber-optic Raman probes with confocal capability, have enabled real-time histopathological assessments in vivo, (i.e., optical biopsy), during ongoing endoscopy. Raman spectroscopy studies of colorectal polyps have tended to be limited to the so-called fingerprint (FP) spectral range (e.g., 800-1800 cm−1). Some attentions have been directed towards the use of a high-wavenumber (HW) regime (e.g., 2800-3600 cm−1) since this spectral range exhibits stronger tissue Raman signals as well as less interferences from the silica background from fiber-optic Raman probes. At present, however, there are no techniques which enable identification of cancerous cells in vivo that have sufficient accuracy to make this type of technique a viable option. A need exists for Raman spectroscopic techniques that enable higher accuracy identification of cancerous cells in vivo and ex vivo.
An object of embodiments in accordance with the present disclosure is to overcome at least some of the problems associated with the prior art and current endoscopic techniques for characterizing, identifying, and/or diagnosing the type and/or nature of tissue(s) such as abnormal growths, for instance, as cancers and the like in essentially any part of the body. A further object is to provide an enhanced accuracy system and method based on Raman spectroscopy for rapidly characterizing, identifying, and/or diagnosing the type or nature of abnormal tissues, such as polyps and pre-cancer, in vivo during endoscopic investigations (e.g., gastrointestinal endoscopic investigations such as colonoscopy).
Various embodiments in accordance with the present disclosure describe a system and method for Fiber-optic Raman spectroscopy, which provides a label-free vibrational spectroscopic technique enabling enhanced accuracy optical biopsy at the bio-molecular level in vivo. Multiple embodiments enable a combination of simultaneous measurements from both FP and HW spectral ranges during ongoing endoscopy. The rationale for combining the FP and HW spectral ranges for in vivo and ex vivo Raman measurements are diverse:
In accordance with an embodiment of the present disclosure, a combined FP and HW fiber-optic confocal Raman spectroscopy technique (e.g., involving the simultaneous acquisition of FP and HW spectra) can improve the real-time diagnosis of cancer, pre-cancer, and/or other abnormal growths in vivo during examinations of the body. The combined FP and HW technique can also be used ex vivo on tissue samples, to more accurately identify the types of abnormal growth present in the samples from any part of the body.
In accordance with an aspect of the present disclosure, a Raman spectroscopy apparatus includes: a first illumination source configured for directing illumination into a tissue; a Raman spectrograph configured for simultaneously detecting FP and HW Raman spectra from illumination scattered by the tissue; and a computerized control and analysis module comprising at least one processing unit and a memory storing program instructions executable by the at least one processing unit for analyzing discrete spectral sub-intervals (e.g., approximately 3-15 or about 5-10 discrete spectral sub-intervals, where a given or each spectral sub-interval can have a spectral width of approximately 10-30 cm−1 or about 20 cm−1) of the detected Raman spectra in FP and HW wavelength ranges to identify a match with one or more reference markers in one or both wavelength ranges.
In multiple embodiments, the Raman spectrograph has a single broadband diffraction grating. The first illumination source includes a source of collimated illumination for generating an excitation energy to apply to the tissue, and the apparatus further includes a probe for transmitting the collimated illumination to the tissue and returning the detected Raman spectra from the tissue to the Raman spectrograph.
The one or more reference markers can include or be specific peaks in the detected Raman spectra. The computerized control and analysis module can include program instructions executable by the at least one processing unit for diagnosing an abnormal growth based upon the match.
The probe can include or be a confocal fiber-optic probe. The apparatus can further include an endoscope having an elongate shaft having an instrument channel within which the probe is carried.
The computerized control and analysis module can include program instructions executable by the at least one processing unit for dynamically adjusting a power of the collimated illumination, and/or dynamically adjusting a time to which the tissue is exposed to the collimated illumination.
The apparatus can include a calibration apparatus configured for standardizing the probe or the entire Raman apparatus with respect to at least one calibration reference.
The apparatus can include an additional illumination source configured for outputting additional illumination into the tissue; and a hot mirror filter configured for compensating for illumination interference between the illumination output by the first illumination source and the additional illumination output by the additional illumination source.
In accordance with an aspect of the present disclosure, a method performed by a Raman spectroscopy apparatus includes: directing illumination output by a first illumination source into a tissue; simultaneously detecting by way of a probe FP and HW Raman spectra from illumination scattered by the tissue; and analyzing discrete spectral sub-intervals in the detected Raman spectra (e.g., approximately 3-15 or about 5-10 discrete spectral sub-intervals, where a given or each spectral sub-interval can have a spectral width of approximately 10-30 cm−1 or about 20 cm−1) in both FP and HW wavelength ranges to identify a match with one or more reference markers in one or both wavelength ranges.
Simultaneously detecting FP and HW Raman spectra can include diffracting illumination in both FP and HW wavelength ranges using a single broadband diffraction grating.
The method can include diagnosing the nature of an abnormal growth based upon the match. The one or more reference markers can include or be specific peaks in the detected Raman spectra.
The method can further include dynamically adjusting the power of the illumination, and/or dynamically adjusting a time to which the tissue is exposed to the illumination.
The method can include performing a calibration or standardization procedure to standardize the probe or the entire Raman apparatus with respect to at least one calibration reference prior to illuminating the tissue.
The method can further include directing additional illumination into the tissue using an additional illumination source while directing the illumination output by first illumination source into the tissue; and compensating for illumination interference between the illumination output by the first illumination source and the additional illumination output by the additional illumination source using a hot mirror filter.
Reference will now be made, by way of example, to the accompanying drawings to provide a better understanding of embodiments in accordance with the present disclosure. The drawings should not be interpreted to be limitative and dimensions may not be to scale.
Raman spectroscopy represents a unique optical vibrational technique based on the fundamental principle of inelastic light scattering. When incident laser light induces a polarization change in molecules, a small proportion of incident photons (˜1 in 108) are inelastically scattered with frequency shifts corresponding to the specific Raman active vibrational modes of the molecules in the sample. Different molecules and different bonds vibrate at different frequencies. Raman spectroscopy is therefore capable of harvesting a wealth of specific bio-molecular information from a huge number of inter- and/or intracellular components, such as proteins, lipids and deoxyribonucleic acids (DNA), water, etc. in tissue.
The fiber-optic confocal Raman spectroscope 100 includes a collimated illumination source such as a near infrared (NIR) diode laser 102; a high-throughput reflective imaging spectrograph 104; a NIR-optimized charge-coupled device (CCD) camera 106; and a probe 108 optically coupled to both the laser 102 and the spectrograph 104 by way of an optical fiber 110. The probe 108 can be carried by an elongate shaft 105 of an endoscope (e.g., within an instrument channel provided by the elongate shaft 105). The fiber-optic Raman spectroscope 100 can further include a band pass filter 112 for background rejection of laser light illumination, and a long pass filter 114 for passing tissue Raman signals while eliminating scattered laser light and fiber background interference. A computer or microcontroller system 124 can provide an automated/computerized control and analysis module for controlling aspects of Raman spectroscope operation and performing Raman spectral analyses.
In a representative implementation, the near infrared diode laser 102 can have a maximum output of 300 mW and a wavelength of 785 nm, such as would be consistent with the device produced by, for example, B&W Tek Inc. The NIR laser 102 generates an excitation energy at the tip 116 of the probe 108 which can cause vibration in any species “illuminated” by the probe 108 and thereby give rise to a Raman spectrum. Other types of collimated illumination can be used to replace the diode laser 102. The spectrograph 104 can be equipped with thermo electric-cooling, for instance, to about −70° C. The spectrograph 104 can be consistent with a device such as the Acton LS785 f/2, produced by Princeton Instruments Inc. The camera 106 can be consistent with a Pixies 400BR eXcelon as produced by Princeton Instruments Inc. In such a representative implementation, the spectroscope 100 can acquire in vivo Raman spectra in the spectral range of 400-3600 cm−1 with a resolution of about 11 cm−1. The illustrated devices are presented as representative examples and are not intended to be limitative.
The atomic emission lines of mercury-argon spectral calibration lamps can be used for wavelength calibration. The lamps may be those consistent with the HG-1 and AR-1 produced by Ocean Optics, Inc., Dunedin, Fla. All wavelength-calibrated spectra are corrected for the wavelength dependence of the system, using a tungsten calibration lamp such as a RS-10 as produced by EG&G Gamma Scientific, San Diego, Calif.
In some embodiments, in order to measure the FP and HW spectra, the tissue Raman signals can be measured either by successively switching different laser excitation frequencies, or by using a dual-transmission grating to cover the entire spectral range i.e., (i.e., −150 to 1950 cm−1; 1750 to 3600 cm−1) in high resolution, as disclosed in International Patent Application No. PCT/SG2014/000063.
The system 200 includes a near-infrared (NIR) diode laser 202 (λex=785 nm), a high-throughput reflective spectrograph 204 equipped with a thermoelectric-cooled, NIR-optimized charge-coupled device (CCD) camera 206 and a specially designed 1.8-mm (outer diameter) fiber-optic confocal Raman probe 208. The system 200 further includes a computer/microcontroller system 124 configured for providing an automated/computerized control and analysis module for controlling aspects of Raman spectroscope operation and performing Raman spectral analyses. More particularly, the computer/microcontroller system 124 can include one or more processing units configured for executing memory-resident program instructions for performing particular Raman spectra acquisition and analysis operations, procedures, or processes in accordance with an embodiment of the present disclosure.
A customized gold-coated broadband reflective grating (e.g., 830 g/mm, having a diffraction efficiency of >90% at ˜800 nm) is incorporated or integrated into the fiber-optic confocal Raman system 200 to cover the entire spectral range (i.e., 400-3600 cm−1) with a spectral resolution of ˜11 cm−1. The fiber-optic confocal Raman endoscopic probe 208 is used for both laser light delivery and in vivo tissue Raman signal collection. The confocal Raman endoscopic probe 208 has previously been described in International Patent Application No. PCT/SG2014/000063, and includes a plurality of 200 μm filter-coated beveled collection fibers (NA=0.22) surrounding a central light delivery fiber (200 μm in diameter, NA=0.22). A miniature 1.0 mm sapphire ball lens (NA=1.78) is coupled to the fiber tip of the confocal probe 208 to tightly focus the excitation light onto tissue, enabling the effective Raman spectrum collection from the epithelial lining (tissue depth<200 μm). The fiber-optic confocal Raman probe 208 can be inserted into the instrument channel of medical endoscopes and placed in gentle contact with the epithelium for in vivo tissue characterization and diagnosis using a broadband confocal Raman endoscopy technique in accordance with an embodiment of the present disclosure.
Intense Raman peaks are also seen in the HW region such as:
This broadband technique in accordance with an embodiment of the present disclosure allows either the FP or HW range or both FP and HW spectral regions to be used simultaneously for tissue analysis, identification, characterization, and/or diagnosis, and is therefore particularly useful in endoscopically accessible organs that exhibit intense auto-fluorescence, which rapidly saturates the CCD. The computer/microcontroller system 124 can be configured for selectively using the FP spectral region, the HW spectral region, or both the FP and HW regions together for tissue analysis, identification, characterization, and/or diagnosis, for instance, in response to user input directed to a graphical user interface (GUI).
One embodiment includes the use of the HW spectral region for diagnosis if the FP region is exceeding the dynamic range of the CCD. In another embodiment, both FP and HW spectral regions are used for tissue characterization, identification, and/or diagnosis. The broadband fiber-optic confocal Raman endoscopy platform allows switching between different spectral regions (e.g., either HW, FP, or simultaneous FP and HW) according to the CCD saturation level and/or tissue type measured.
A tissue characterization, identification, or diagnosis technique or method in accordance with an embodiment of the present disclosure utilizes the complementary diagnostic information from the broadband FP and HW spectra.
In general, tissue biomedical spectra are extremely complex. To convert the subtle molecular differences of Raman spectra between different tissue types into valuable diagnostic information requires sophisticated multivariate statistical analysis techniques, such as principal components analysis (PCA).
This has been widely practiced by utilizing the entire (i.e., continuous) Raman spectra for tissue diagnosis and characterization on either the FP or HW regions, respectively.
PCA reduces the dimension of the Raman spectra by decomposing them into linear combinations of orthogonal components, such as principal components (PCs), such that the spectral variations in the dataset are maximized. Thus, PCA has typically been integrated with effective clustering algorithms such as support vector machines (SVM), logistic regression (LR) and linear discriminant analysis (LDA) for classification of biomedical Raman spectra. PCA is very efficient for data reduction and analysis.
Alternatively, the partial least squares (PLS)-discriminant analysis (DA) has been applied for classification problems by encoding the class membership of zeroes and ones, representing group affinities in an appropriate Y-indicator matrix. PLS-DA employs the fundamental principle of PCA, but further rotates the components, such as latent variables (LVs) by maximizing the covariance between the spectral variation and group affinity so that the LVs explain the diagnostically relevant variations rather than the most prominent variations in the spectral dataset. In most cases, this ensures that the diagnostically significant spectral variations are retained in the first few LVs.
Most multivariate algorithms (e.g., PCA or PLS-DA) are originally not designed to cope with large amounts of irrelevant spectral variables. In other words, some spectral regions in Raman spectra may have a degrading effect on the diagnostic model, for example, due to large variance, interferences, inter-anatomical variability, etc.
In an embodiment, a novel diagnostic technique or procedure is provided that utilizes the complementary information from the FP and HW spectral ranges, e.g., as obtained by FP and HW spectral measurements made using a broadband fiber-optic confocal Raman spectroscopy system 200. Such an embodiment makes use of discrete spectral subintervals (e.g., approximately 3-15 or about 5-10 discrete spectral sub-intervals, where a given or each spectral sub-interval can have a spectral width of approximately 10-30 cm−1 or about 20 cm−1) in the broadband FP and HW Raman spectra rather than the continuous spectral ranges for diagnosis.
An embodiment in accordance with the present disclosure relates to the diagnosis of gastrointestinal abnormal growths, such as colorectal abnormal growths. The raw FP and HW Raman spectra measured from in vivo colorectal tissue represents a combination of weak tissue Raman signals, intense auto-fluorescence background, and noise. In order to view and analyze the Raman signals, the background and noise must be treated or removed. The raw spectra are therefore pre-processed by a first-order smoothing filter to reduce the spectral noise, such as a Savitzky Golay filter having a window width of 3 pixels. In the FP region (800-1800 cm), a fifth-order polynomial is found to be optimal for fitting the auto-fluorescence background in the noise-smoothed spectrum and this polynomial is then subtracted from the calibrated FP spectrum to yield the tissue Raman spectrum alone. In the HW range (2800-3600 cm−1) a first order linear fit is found to be optimal for removing the weaker auto-fluorescence base-line. Such preprocessing of each received signal is completed within about ms, and hence the processed Raman spectra and diagnostic outcomes can be displayed on a display device such as a computer screen in real-time.
Following pre-processing, the Raman spectra are analyzed to determine peaks and/or markers. These peaks or markers are then used to estimate, characterize, or determine the nature of the tissue (the nature of the lesion or abnormal growth) from which the spectra are derived. This can be done by the computer system using an appropriate means of comparison with known control spectra, reference markers, and the like. The term reference markers as used herein is intended to include one or more individual peaks in the Raman spectrum, or indeed the whole Raman spectrum. If one or more Raman reference markers are indicative of a certain nature of abnormal cell growth, a comparison between an acquired spectrum and a reference marker can be made using a lookup table or the like. It will be appreciated that many different types of comparison techniques or methods can be employed. Once the comparison has been made and a best match has been determined, the type of abnormal cell growth that is present can be indicated to the user of the system 200 by any appropriate means. This can include a visual representation on a computer screen and/or an audible message.
In a trial of an embodiment of the system 200, a total of 50 consecutive symptomatic patients were colonoscopically examined. The patients had presented for examinations for surveillance or screening of various colorectal indications such as anemia, bleeding, etc. Prior to colonoscopy, the patients were administered polyethylene glycol (PEG) electrolytes bowel preparation. Sedation was performed using intravenous administered propofol. The endoscopists cleaned the colon during inspection and prior to the confocal Raman scans, the polypoid and flat colorectal lesions were further flushed with a physiological saline solution to further reduce confounding factors (i.e., residual stool and fluid, etc.). During a typical examination the endoscope was directed to the distal colon and a Raman scan (n˜15 spectra) was performed on suspicious lesions during withdrawal of the probe from the body. Each tissue Raman measurement was acquired within a period of about 0.1 to 0.5 second. This permitted a rapid survey of colorectal polyps. Any Raman spectra that were acquired in non-contact with colonic polyps (˜10%) were automatically discarded by on-line clinical software using principal component analysis (PCA) methods associated with Hotelling's T2 and Q-residual statistics. The PCA methods are the subject of previous International Patent Application No. PCT/SG2014/000063, and work as follows:
After the Raman scan had been completed and the results saved, each tissue specimen was removed; fixed in formalin; sectioned; stained with hematoxylin and eosin (H&E); and sent for histopathological examination. Colorectal tissues were classified into the following three clinically important categories:
The simultaneous FP and HW fiber-optic confocal Raman technique in accordance with an embodiment of the present disclosure was compared with the histopathology assessments to determine the ability to differentiate neoplastic from non-neoplastic colorectal lesions in vivo.
The comparison included the use of statistical analysis of the results to validate whether the simultaneous FP and HW fiber-optic confocal Raman spectroscopy technique was sufficiently accurate to replace currently used techniques. Cohen's κ statistics were calculated to assess the agreement for the histopathological characterization. Analysis of variance (ANOVA) with Fisher's post hoc least significant differences (LSD) test was used to test differences in means between groups. Multivariate statistical analysis was used to extract the significant Raman spectral features for clinical diagnostics. A probabilistic partial least squares (PLS) discriminant analysis (DA) was applied for tissue diagnosis. A “leave-one patient-out” cross-validation was used to assess and optimize the PLS-DA model complexity reducing the risk of over fitting. The receiver operating characteristic (ROC) curves were generated and the area under the curves (AUCs) were calculated to evaluate the capability of the FP and HW fiber-optic confocal Raman spectroscopy technique to differentiate neoplastic from non-neoplastic polyps in vivo.
In an experiment, which is presented by way of example only, to show the results of use of an embodiment in accordance with the present disclosure on a particular group of test subjects or patients, fifty patients (27 male and 23 female) with a mean/range age of 52/(23-83) where enrolled for fiber-optic confocal Raman examination. Thirteen patients presented with adenomas (eleven tubular and two tubulovillous adenomas) harboring low-grade dysplasia. Three patients were associated with advanced stage colorectal adenocarcinoma. A Cohen's kappa of 0.89 demonstrated a high level of agreement between the pathological findings for the three tissue groupings. A total of 1731 in vivo colorectal Raman spectra were successfully acquired from 126 lesions or abnormal growths. Of these lesions, 1397 were benign, 235 were adenoma and 99 were adenocarcinoma. This was confirmed by histopathology examinations.
The detailed distribution of the patients and lesions including pathological subtypes and anatomical locations, such as ascending, transverse, descending, sigmoid, rectum, are summarized in Table 1.
Intense Raman peaks are also seen in the HW region such as:
The in vivo Raman spectra were correlated with representative histopathology slides of the different pathology categories.
Using the complementary bio-molecular information from both the FP and HW spectral ranges, fiber-optic Raman spectroscopy for in vivo diagnosis is applied.
It is known that a large group of people harbor small colorectal hyperplastic polyps or flat polypoid lesions that significantly add to the cost of histopathological assessments. The most significant motivation for developing and adopting advanced endoscopic modalities for colorectal examinations is the capability of discriminating benign hyperplastic polyps from adenoma.
Embodiments in accordance with the present disclosure demonstrate that for the first time high quality in vivo confocal Raman spectra covering the FP and HW spectral ranges can be measured (e.g., simultaneously) from colorectal polyps and analyzed in real-time and this can be used to offer an improved manner of identifying neoplasia as is shown in
The above mentioned peak intensities relate to vibrations of specific bonds within molecules. Each type of bond has a different peak intensity and identification of a peak intensity that can be associated with a specific bond can reveal bio-molecular information. The fact that a lesion or abnormal growth includes a specific bond or bonds and gives rise to a specific associated peak intensity can be used to differentiate different types of abnormal growth. For example, if an abnormal growth which is benign exhibits a specific peak intensity, later detection of that peak intensity can lead to the determination that in the later situation the lesion or abnormal growth is also likely to be benign.
In accordance with embodiments of the present disclosure, the link between peak intensity and bio markers can be exploited to identify the type of abnormal growth that is present. This can occur both in vivo and ex vivo.
In some results, the water content was found to be markedly different in adenoma. The increase in intensity of the broad anti-symmetric OH stretching vibration at 3200 cm−1 (p<0.001) indicates that neoplastic epithelium holds an increased content of bound water which could partially be explained by the expression of aquaporins altering the water permeability thereby inducing hydration of the neoplastic cells. The increase of bound water may be interrelated to the simultaneous decrease of hydrophobic lipids. The significance of this finding was analyzed by calculating the peak intensity ratio (i.e., 12885/13200) associated with lipid and bound water. This peak ratio alone may distinguish adenoma from benign polyps with a sensitivity of 81.3% (191/235) and a specificity of 80.4% (1132/1397). Hence, the lipid content and water perfusion in polyps are very useful biomarkers for colorectal neoplasia in situ. The direct correlation of the epithelial Raman spectral signatures with cell and tissue bio-chemistry can therefore deepen the understanding of colorectal carcinogenesis at the bio-molecular level in situ. Use of the peaks or markers can be employed to diagnose the type of abnormal growth type which is present.
By capitalizing on the broad range of complementary optical biomarkers including proteins, lipids, DNA and the conformations of protein-bound and unbound water, for the first time it has been demonstrated that accurate diagnosis of adenoma can be realized in vivo as shown in
The effectiveness of FP and HW fiber-optic confocal Raman colonoscopy technique for in vivo detection and diagnosis of colorectal neoplasia is clear from the results. It has also been demonstrated that by utilizing both the FP and HW spectral ranges yields an AUC that was superior to using either of the FP or HW ranges alone. These substantial results establish that the FP and HW Raman spectroscopy technique can largely reduce the misclassification rate, confirming the addition of complementary bio-molecular information from the FP and HW ranges for enhancing the colorectal diagnosis in vivo. For instance, the HW spectral range contains information related to local conformation of water as well as CH2 and CH3 stretching moieties that are not contained in the FP range. The combination of FP and HW ranges can also be used for diagnosis purposes ex vivo.
It should be noted that the combination of the FP and HW Raman spectroscope techniques is typically a preferred method of operation, such as by way of simultaneous FP and HW spectral measurements. However, the use of either FP or HW in vivo yield results which are capable of identifying different types of abnormal growth. As a result, embodiments in accordance with the present disclosure can use FP alone, HW alone, or a combination of FP and HW (e.g., simultaneously).
It should also be noted that the use of discrete spectral sub-intervals (which may also be referred to as predetermined values and/or reference markers) in the broadband FP and HW spectra provides an improved diagnosis technique or method.
The use of a single gold-coated broadband reflective grating (e.g., for obtaining FP and HW spectra simultaneously) is an important development in the equipment for enhanced accuracy target tissue characterization and diagnosis, e.g., for characterizing/diagnosing potentially abnormal or abnormal growths in vivo with increased accuracy in real time during an endoscopic procedure. The single gold-coated broadband reflective grating means that there is no need to switch between diffraction gratings during in vivo measurements, as had been the case in the prior art. This clearly has many advantages. A system or device in accordance with an embodiment of the present disclosure is thus able to make all necessary measurements without any switching processes being required. The device is more compact and cost-effective as it has only one grating rather than multiple gratings. Although the singe grating design is a bit compromised in spectral resolution compared to a dual-grating design, this compact system would not affect or significantly affect diagnostic purposes or outcomes as the spectral resolution of the compact system matches the tissue Raman spectral bandwidth that is usually in the range of ˜10 cm−1.
Fiber-optic confocal Raman spectroscopy provides objective uninterrupted real-time computerized diagnosis, which is straightforward to operate and requires no additional endoscopic training or administration of contrast agents. By enabling functional and bio-molecular/bio-chemical assessment of the intestinal epithelium in vivo, the introduction of FP and HW fiber-optic confocal Raman spectroscopy will have a major impact on gastrointestinal endoscopy practice, such as colonoscopic practice. This new bio-molecular endoscopic approach enables objective and immediate decision-making during clinical colorectal examinations. There can be two key roles for fiber-optic confocal Raman spectroscopy in colorectal examinations as follows:
Embodiments in accordance with the present disclosure open up the possibility for use in endoscopic and laparoscopic surgery of CRC thereby offering the gastroenterologist an objective tool for real-time assessment and definition of resection margins as well as the follow up evaluation of post-treatment efficacy or recurrence at the molecular level. This can aid in complete tumor excision and subsequent margin assessment to reduce the risk of recurrence. Thus, FP and HW fiber-optic confocal Raman spectroscopy in accordance with embodiments of the present disclosure can be used in the field of gastrointestinal endoscopy, as well as other body sites, in both screening settings and therapeutic colorectal applications by enabling real-time, in vivo objective tissue assessments.
Another clinical example utilized an embodiment in accordance with the present disclosure for real-time in vivo diagnosis of esophageal squamous cell carcinoma (ESCC) during endoscopy by way of simultaneously acquiring both fingerprint (FP) (i.e., 800-1800 cm−1) and high-wavenumber (HW) (i.e., 2800-3600 cm−1) Raman spectra from esophageal tissue in vivo. In this set of experiments, a total of 1172 high-quality in vivo FP/HW tissue Raman spectra (normal (n=860); ESCC (n=312)) were acquired from 48 esophageal patients undergoing routine endoscopic examination. The total in vivo Raman dataset was split into two parts: i.e., 80% of the total dataset for training (938 in vivo FP/HW Raman spectra [normal (n=736); ESCC (n=202)] from 34 esophageal patients); while the remaining 20% of the total dataset for predictive testing (234 in vivo FP/HW Raman spectra [normal (n=124); ESCC (n=110)] from 14 esophageal patients).
Intense Raman peaks are also observed in the HW regionError! Reference source not found. Error! Reference source not found. as follows:
To elucidate the diagnostically important Raman-active components,
Capitalizing on the complementary biochemical/biomolecular information identified in both the FP and HW spectral ranges, PLS-DA and LOPCV were implemented on the training dataset (80% of the total dataset) to develop robust diagnostic model for enhancing in vivo ESCC diagnosis. A Cohen's kappa of 0.91 demonstrated a high level of agreement between the independent pathologists for the esophageal tissue groupingsError! Reference source not found.
In the light of these promising diagnostic results, simultaneous FP/HW Raman spectroscopy and diagnostic algorithms developed were applied for predictive diagnosis of the independent testing dataset (20% of the total dataset). The predictive accuracy of 93.2% [i.e., sensitivity-92.7% (102/110) and specificity-93.6% (116/124)] can be achieved with integrated FP/HW Raman spectroscopy, substantiating the advantages of integrated FP/HW Raman spectroscopy over either FP (predictive accuracy-91.0%; sensitivity-90.9% (100/110), and specificity-91.9% (113/124)) or HW (predictive accuracy-80.3%; sensitivity-76.4% (84/110), and specificity-83.9% (104/124)) Raman technique alone for in vivo ESCC.
In a further embodiment, the identification of other types of cancer were considered, for example, using the system 200 of
To select the complementary spectral intervals from the FP and HW spectral regions, variable/feature selection techniques are incorporated into the broadband Raman endoscopy technique. The benefits of variable/feature selection include:
In an embodiment, an interval PLS-DA algorithm is used, but in principle other, multiple, or all feature/variable selection techniques or methods can be applied to select the complementary spectral regions for any clustering algorithm, such as PCA-LDA, SVM, LR etc. The feature/variable selection techniques could be genetic algorithms (GA), swarm intelligence, selectivity ratios etc. Briefly, interval PLS-DA used herein performs a sequential, exhaustive search for the best combination of intervals in the Raman spectra. Accordingly, interval PLS-DA creates individual PLS models, each using only a subset or window of variables. If there are 200 intervals for a given spectral data set, 200 PLS-DA models are calculated (i.e., one for each interval).
A leave-one patient out cross-validation is performed for every model and the interval, which provides the highest diagnostic accuracy, is selected. This is the optimum single-interval model. If only one interval is desired, the algorithm stops at this point. If, however, more than one interval is desired (to increase information content and improve predictive performance), additional cycles can be performed. In the second cycle, the first selected interval is used in all models but is combined with each of the other remaining intervals, one at a time, when creating a new set of new PLS models. In this way, regions of complementary diagnostic value from the FP and HW range are extracted while redundant or irrelevant spectral ranges, such as regions of poor predictive power are excluded from the model.
Other spectral ranges (e.g., the so-called Raman silent range between 2000 cm−1 to 2800 cm−1) other than FP and HW ranges can be used in certain applications.
To demonstrate an application of this embodiment, a broadband spectra from a series of cervical patients was acquired. A total of 44 non-pregnant female patients (between 18 and 70 years of age) who underwent a colposcopy procedure due to an abnormal Pap smear were recruited studied. Prior to the in vivo tissue Raman spectral measurements, a 5% acetic acid solution was applied topically on the cervix for 2 min for evaluation of the color whitening in the tissue (the degree of white discoloration in the cervix is related to the grade of pre-cancer).
Confocal broadband Raman and concomitant auto-fluorescence spectra were measured by placing the fiber-optic confocal probe in gentle contact with the tissue.
To demonstrate that the broadband Raman technique integrated with interval selection can improve the diagnosis of cervical pre-cancer a conventional PLS-DA is applied to the continuous spectrum (
It should be noted that thresholds can be imposed on the probabilistic classifications (
In a further clinical example using the broadband confocal Raman endoscope fiber-optic confocal Raman spectra were acquired from a series of gastric patients, focusing on early diagnosis of gastric malignancies.
To demonstrate an embodiment in accordance with the present disclosure, conventional PLS-DA is applied to the continuous spectrum as well as interval PLS-DA.
Yet another clinical example utilized an embodiment in accordance with the present disclosure in which a simultaneous FP and HW fiber-optic Raman endoscopic technique was utilized for in vivo detection of gastric intestinal metaplasia (IM)-precancerous lesions during endoscopic examinations. In this example, the Raman spectroscopy system 200 included a near-infrared (NIR) diode laser (λex=785 nm) (maximum output 300 mW, B&W TEK Inc.), a high-throughput reflective imaging spectrograph (Acton LS-785 f/2, Princeton Instruments Inc.) equipped with a gold-coated 830 gr/mm grating and a thermo electric-cooled, NIR-optimized charge-coupled device (CCD) camera (PIXIS: 400BR-eXcelon, Princeton Instruments Inc.). The system 200 acquired in vivo tissue Raman in the spectral range from 400-3600 cm−1 with a resolution of ˜9 cm−1. A 1.9 m long fiber-optic Raman probe 108 having a 1.8 mm outer diameter was utilized for both laser light delivery and in vivo epithelial tissue Raman signal collection. The Raman endoscopic probe 108 designed for endoscopy includes 18×200 μm bevelled collection fibers (NA=0.22) surrounding a central light delivery fiber (200 μm in diameter, NA=0.22). A 1.0 mm sapphire ball lens (NA=1.78) is coupled to the fiber tip of the probe to tightly focus the excitation light onto the gastric tissue surface, enabling the effective Raman spectrum collection from the epithelial lining. The depth-selective capability of the fiber-optic Raman spectroscopy system 200 ensures shallower tissue interrogation (<200 μm) with microscopic probing volume (<0.02 mm2), thereby reducing the interferences and signal dilution from deeper bulky tissues, while selectively or preferentially interrogating the epithelium associated with neoplastic onset and progression. The atomic emission lines of mercury-argon spectral calibration lamps (HG-1 and AR-1, Ocean Optics, Inc., Dunedin, Fla.) were used for wavelength calibration. All wavelength-calibrated spectra were corrected for the wavelength dependence of the system using a tungsten calibration lamp (RS-10, EG&G Gamma Scientific, San Diego, Calif.). The entire FP/HW fiber-optic Raman endoscopic system 200 is controllable using a foot pedal and an intuitive software framework configured for providing feedback (e.g., auditory and/or visual probabilistic feedback) to the gastroenterologist in real-time.
Raw spectra were preprocessed by a third-order Savitzky-Golay smoothing filter (a window width of 3 pixels) to remove spectral noise. In the FP region (800-1800 cm−1), a fifth-order polynomial was found to be optimal for fitting the AF background in the noise-smoothed spectrum and this polynomial was then subtracted from the measured FP spectrum to yield the FP tissue Raman spectrum alone. In the HW range (2800-3600 cm−1), a first-order polynomial fit was used for removing the AF background. The FP/HW Raman spectra were normalized over the integrated area under the FP and HW ranges to allow a better comparison of the spectral shapes and relative Raman band intensities between normal and IM gastric tissue. All raw spectral data were processed on-line with software developed in the Matlab environment (Mathworks Inc., Natick, Mass.). Principal components analysis (PCA) and linear discriminant analysis (LDA) were implemented to develop robust diagnostic algorithms for the differentiation between normal and IM gastric tissues. Leave-one tissue site-out cross-validation was utilized to evaluate the diagnostic models developed in an unbiased manner. Multiple Raman spectra (10-15) were acquired from each tissue site within 1 second, and the majority voting strategy was applied for final classification. The diagnostic outcomes were displayable on a display device such as a computer screen in real-time. Receiver operating characteristic (ROC) curves were also generated by successively changing the thresholds to determine correct and incorrect classifications for all tissues. All spectra preprocessing and multivariate statistical analysis were performed online using scripts written in the Matlab programming environment.
A total of 1412 in vivo Raman spectra (i.e., normal (n=1083 spectra) and IM (n=329 spectra) were successfully acquired from 115 sites (i.e., normal (n=88 sites) and IM (n=27 sites)) as confirmed by consensus histopathology examinations.
In addition, intense Raman peaks in the HW region are also observed at
To develop sophisticated multivariate diagnostic algorithms and compare tissue diagnostic performance among the three different Raman techniques (i.e., FP, HW and the integrated FP/HW), PCA-LDA together with Student's t-test were implemented on the in vivo tissue Raman spectra acquired to evaluate the elusive differences observed in the spectra of different tissue types. The leave-one tissue site-out, cross-validated PCA-LDA diagnostic algorithms were further developed based on the diagnostic significant PCs (p<0.01) as shown in
This embodiment has several major advantages over the prior art Instead of using the entire broadband spectrum for diagnosis this technique or method only uses a subset (e.g., 5-20%) of complementary information greatly simplifying the diagnostic model. Secondly the variable selection provides qualitative insights into the biochemical and bio-molecular foundation of the disease. Thirdly, the accuracy is increased significantly compared to spectral analysis based on the entire spectral range. Moreover, if certain spectral ranges have interferences or confounding factors (e.g., blood) etc., the effect of these can efficiently be reduced.
In an example a patient scheduled for endoscopic screening of suspicious symptoms for esophageal reflux undergoes the following:
Embodiments in accordance with the present disclosure demonstrate that real-time simultaneous fingerprint (FP) and high-wavenumber (HW) fiber-optic confocal Raman spectroscopy can be performed during screening of patients in vivo. Fiber-optic confocal Raman spectroscopy uncovers the bio-molecular and bio-chemical changes, such as protein, DNA, lipid and water occurring in the epithelium during colorectal carcinogenesis. The use of complementary bio-molecular information from the FP and HW range improves the detection of abnormal growths compared to using either the FP or HW ranges alone. The FP and HW fiber-optic confocal Raman spectroscopy technique has a great potential to improve pre-cancer and cancer detection and characterization. Use of a subset of the whole spectrum has a notable advantage in processing and in the identification of abnormal growths in many different parts of the body. The FP and HW ranges have been particularly identified as yielding good results for the types of detection discussed herein. It will be appreciated for other types of detection other ranges may prove more useful.
Embodiments in accordance with the present disclosure are not confined to biomedical Raman spectroscopy, but can also have application in other areas. These includes for example, fluorescence spectroscopy, elastic scattering spectroscopy, surface enhanced Raman spectroscopy, process analytical technology, water and environment monitoring, pharmaceutical process/drug delivery control, food industry, quality control industry, forensics, etc. In such embodiments, excitation energy provided by an illumination source is directed to a target sample (e.g., a chemical, water, or environmental material/substance sample, a pharmaceutical/drug or food sample, or another type of sample), which need not include or be tissue. Furthermore, such embodiments need not involve an endoscope or endoscopy. Reference to the term Raman herein is intended to include other types of spectroscopy including those mentioned above.
Embodiments in accordance with the present disclosure can serve as a diagnostic platform for any organs, such as the lung upper and lower GIs (e.g. esophagus, stomach, colorectum), liver, bladder, head and neck (e.g., nasopharynx, larynx, oral cavity), cervix, skin, bone, or any other place where a conventional endoscope, laparoscope, or arthroscope can be used.
Some of the tissue Raman spectra obtained using certain embodiments in accordance with the present disclosure may suffer from endoscope illumination interferences. This can make diagnosis of abnormal growths difficult to achieve. As a result, a further embodiment offers a solution to eliminate such interferences.
A trimodal endoscope imaging system, according to an embodiment of the present disclosure, used to guide the confocal Raman fiber-probe as described above, includes a 300 W short-arc xenon light source, a gastrointestinal (GI) endoscope, and a video signal processor. The xenon light source is coupled with different sets of filters to provide different illumination light(s) for trimodal endoscopic imaging in tandem (not shown). The filters can include, for instance: WLR (red filter, 585-655 nm; green filter, 500-575 nm; blue filter, 390-495 nm), AFI (blue filter, 390-470 nm; green filter, 540-560 nm for reflectance image normalization), and NBI (green filter, 530-550 nm; blue filter, 390-445 nm).
The light reflected or fluorescence emitted from tissue is detected using two different CCDs mounted at the distal tip of the GI endoscope (not shown): one a conventional CCD for WLR/NBI and one a high-sensitivity CCD for AFI observation. The endoscope short-arc xenon lamp emits broadband continuous light covering UV/VIS/NIR with prominent discrete peaks in the NIR spectral range>700 nm. Since the 785 nm laser excitation Stokes Raman spectroscopy also falls in the NIR range, ambient xenon light can interfere with the tissue Raman signal and fully obscure the in vivo tissue measurements and diagnosis. For this reason, Raman endoscopic diagnosis is conventionally performed with xenon illumination light turned ‘off’ or dimmed, which is highly undesirable in clinical settings.
To remove the ambient xenon illuminations interference, an embodiment is proposed in which the integration of a hot mirror low-pass filter (cutoff at ˜700 nm, ˜95% average transmission in visible range) in front of the xenon light source in the endoscope system enables elimination of ambient interference.
To demonstrate the application of this embodiment, in vivo Raman spectra are measured under different endoscopic illumination conditions.
The Raman spectra was measured with the probe in contact with the tissue.
It should be noted that the filtering of illumination light for guidance using a hot mirror is not limited to the endoscopic application but is more general for filtering any light or imaging modality used to guide NIR tissue spectroscopy (i.e., surface-enhanced Raman spectroscopy (SERS), NIR fluorescence or NIR reflectance) for internal or external organs (e.g., skin, cervix, bladder, gastric, esophagus, nasopharynx, larynx, oral cavity, colon, rectum, lung, etc.).
The concept has therefore been generalized in
It is known that instrument standardization has fallen far behind the pace of advances in clinical Raman spectroscopy. Because tissue Raman spectroscopy is based on inherently weak and highly resolved peaks, the technology is very sensitive to instrumental changes. It is therefore of imperative to develop and employ techniques for testing/calibrating and standardizing Raman endoscopy instrumentation prior to clinical measurements in human patients to ensure that the in vivo diagnosis is reliable and consistent among different Raman systems.
The filter wheel rotation has been synchronized with laser excitation and acquisition in the clinical Raman software. Spectra are acquired from each sample in the filter wheel and stored. The steps performed in the calibration/testing routines comprise measurement of the CCD characteristics (i.e., temperature, dark current), laser excitation power, and fused silica fiber background, a fluorescent material for system response calibration/testing, a material for wavelength calibration/testing and finally a tissue phantom. The signals measured can be used for recalibrating the parameters (e.g., intensity response, wavelength accuracy, background noise, etc.) of the Raman endoscope system.
The routines and standard samples that have been integrated in the filter wheel of the calibration device will now be described along with examples of the signals obtained.
In a first instance a detector signal of the CCD is measured and temperature is logged (2402). This may include measuring multiple spectra (i.e., at 0 . . . 1 sec exposure time) of the signal intensity in the absence of laser excitation. The system subsequently verifies that the dark current is less than a maximum value which was stored earlier in a factory configuration file. This embodiment further includes ensuring that the dark current coefficient of variation, over a series of spectra, is less than a maximum value which was also stored earlier in the configuration file.
The laser excitation power at the tip of the fiber-optic Raman probe is also measured (2402) using an integrated laser-power meter to ensure that this is within the range that is less than the American National Standards Institute (ANSI) maximum permissible skin exposure limit (which is set at 785 nm laser beam). This embodiment includes ensuring that the laser power is less than a maximum value which was stored earlier in the configuration file.
The filter wheel contains an empty slot with dark environment so that a fused silica fiber probe background signal can be measured (2404). Multiple spectra of the fiber probe backgrounds are measured with laser excitation and different exposure times (e.g., 0.1 . . . 1.0 s). These spectra contain information about the condition of the fiber-optic probe. For instance, if the fiber is damaged or the probe tip is contaminated, this could be reflected by an increase in Raman or fluorescence intensity.
An example of background signals for the broad-band Raman endoscopy probe is shown in
A fluorescent glass that exhibits a broad stable fluorescence spectrum (e.g., chromium doped borosilicate glass, green glass, kopp2412 etc.) is also measured and stored for testing and/or calibrating the response and collection efficiency of the Raman endoscopy system (2406). The standard fluorescence glass is factory calibrated beforehand using a National Institute of Standards and Technology (NIST) tungsten lamp according to the procedure detailed in
A material with well-defined narrow Raman peaks is also measured (2408). An example is polystyrene shown in
The filter wheel also contains a layered tissue phantom. The tissue phantom consists of a material with diffuse properties and/or layered tissue phantoms that exhibit well-known Raman peaks which are measured (2410). This embodiment further includes ensuring that the coefficient of variation, over a series of spectra, is less than a maximum value which was stored earlier in the configuration file. An example is given in
After the various measurements have been made, analysis of the signals is carried out (2412). This is to identify problems such as failure detection, calibration problems etc. If problems are detected the system is defined as not passing (2414). Alternatively system is passed if no problems are encountered (2416).
A further embodiment includes a program instruction set or software framework integrated with or as a GUI that is capable of displaying and saving Raman data on a spectrum and lesion basis in response to user selection(s)/input(s) when Raman endoscopic diagnosis is performed. For instance, if a first selection is made or a first button is pushed, Raman data from a first lesion will be saved. If a second selection is made or a second button is pushed, Raman data from a second lesion will be saved etc. Such a system also includes storing patient information, integration time, laser excitation power, time, date, diagnosis, probe background signal, system calibration function, endoscopy video and a logging-file containing measurement details. An integrated foot pedal device can be provided for interfacing and saving data by the clinician. If the footswitch is pushed, the data acquired will be stored as a first lesion. If pushed a second time, the data will be stored as a second lesion etc. The footswitch can also contain a safety pedal for turning on and off the laser excitation.
A further embodiment includes a program instruction set or software framework for displaying clinical Raman diagnosis together with the recorded wide-field video imaging. Endoscopic video imaging (i.e., WLR/NBI/AFI) and video recording has been integrated into the Raman endoscope software. Both the wide-field endoscopic video and the in vivo Raman diagnosis of the tissue imaged can be displayed on a display device such as a computer screen. The endoscopy video is synchronized in time with the measured tissue Raman spectra. Therefore, Raman spectral diagnosis can be displayed simultaneously with video playback. This enables the clinician to trace-back for each patient precise correlation between Raman endoscopic diagnosis and the suspicious tissue site sampled. This integrated imaging and Raman endoscopy software is a substantial improvement in the clinical system enabling review of historical clinical data.
The limited dynamic range of CCDs and the weak tissue Raman signals remain challenging in Raman endoscopic applications since in vivo tissues exhibit varying degree of auto fluorescence. For some tissues (e.g., gastric, lung, dorsal tongue, liver), detector saturation can occur in 0.1 second or even less than 0.05 second. For these tissues, the Raman signal can have prominent noise compared to other tissue types (e.g., esophagus, nasopharynx, larynx, cervix etc.). In general the Raman signal intensity of tissue scales linearly with laser excitation power and exposure time. The signal to noise ratio (S/N) is proportional to the square root of the exposure time. Currently, the user adjusts exposure time or laser excitation powers manually for every spectrum acquired using the Raman endoscopy technique. This can be highly impractical in real-time applications. Hence it is critical to define automated methods to prevent CCD saturation and assure that the Raman spectrum is acquired with optimum S/N ratio within times that are acceptable in clinical conditions. The realization of true real-time diagnosis is required for Raman endoscopy technology to gain widespread acceptance in clinical medicine. The novel method to automatically adjust laser excitation power, exposure time and spectrum accumulations to realize uninterrupted real-time diagnosis during clinical Raman measurements with high S/N ratio is invaluable in this quest. Two automatized techniques or methods are set out by way of representative example.
In a first technique or method as shown with reference to
In a second technique or method as shown with respect to
The above two functions detailed in respect of
A still further embodiment in accordance with the present disclosure includes the capability of switching between different fiber-optic probes and different organs. For each type of fiber-optic probe (e.g., confocal, or volume type probe) a database exists with diagnostic models specific to each organ (i.e., larynx, colon, nasopharynx, gastric, esophagus, oral cavity, skin, cervix, bladder etc.). For instance, if a user chooses nasopharynx as an organ, one or more models belonging to the nasopharynx are loaded into the system. A representative database structure is shown in
It will be appreciated that the various embodiments described above are not intended to be limitative in their interpretation. Instead there may be combinations of one or more embodiments and variations from the specific examples shown in the description. Some of the processes may be implemented in hardware, software or any combination thereof.
Number | Date | Country | Kind |
---|---|---|---|
1411800.4 | Jul 2014 | GB | national |
1411803.8 | Jul 2014 | GB | national |
1411820.2 | Jul 2014 | GB | national |
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/SG2015/050195 | 7/2/2015 | WO | 00 |