Definitions

• This invention relates to systems that identify tissue type in a patient by the use of combinations of optical and electrical measurements on tissue surfaces. The measurements are compared with data gathered from prior patient studies, and the patient's tissue is then categorized.
• tissue type based upon responses to incident light and/or elec- trical stimulation is well known. This has led to diagnostic techniques and apparatus for identifying tissue types such as cancerous or pre-cancerous.
• diagnostic techniques and apparatus for identifying tissue types run the gamut from microscopic examination of tissue smears by trained cell pathologists, to the study of the fluorescence, electrical and other physical properties of tissues.
• Much research has been devoted to the identification and comparison of optical and electrical characteristics of healthy and damaged tissue in the hope that it could lead to new diagnostic techniques. The research is driven by the fact that none of the present methods for the detection of cervical cancers are sufficiently accurate, and the risks of incorrect diagnosis are severe.
• Pap Papanicolaou
• the sensitivity limitations of the Pap smear have been well documented, and include an overall false negative rate variously reported as between 20 - 40%, and between 6 - 55%. False negative rates for pre-can- cerous lesions have been assessed as 28%, and between 20-50%.
• the estimated specificity for the test has been profoundly affected by the widespread introduction in the USA of the Bethesda -cytology classification system.
• ASCUS Atypical Squamous Cells of Undetermined Significance
• the present invention is a novel system designed for the detection of cervical precancer and cancer.
• the system is a portable optoelectronic instrument capable of giving the physician operator an instantaneous result without requiring tissue sampling for cytologic analysis.
• the device inter- rogates the cervical tissue using a combination of low level electrical impulses and light pulses at various frequencies.
• the measured response, or tissue signature is compared algorithmically in real time to that stored in a databank of cervical tissue types.
• LSIL low grade squamous intraepithelial lesion
• HSIL/IC high grade squamous intraepithelial lesion/invasive cancer
• the present invention provides an instrument preferably capable of providing both opti- cal and electrical data almost simultaneously from very small sections of tissue surface. Although there is no evidence that the optical properties of tissue are affected by electrical stimulation, or vice-versa, it has been unexpectedly determined that properly combining the data from both types of tests on the same small region of tissue, on the order of a few millimeters diameter, e.g. 3-10 mm, provides a statistically significant increase in the pre- dictability of success of tissue diagnosis.
• the present invention concerns the sequencing of the optical and electrical tests on tissue selected by contact with the probe.
• Selection by contact refers to the ability of the probe to determine the properties of a particular small tissue segment that is contacted by the probe tip and possibly a small area of adjacent tissue.
• the currents caused to flow by the probe do not necessarily flow as surface currents, but may penetrate more deeply into the surface and thus more than superficial cells may also be responsible for the probe test results and are considered to be selected by contact.
• the invention resides in the relationship between the optical and electrical measurements and the statistical analysis of parameters defined as linear combinations of both types of data.
• each measurement involves a complex, sequence of events, including (1) optical and electrical tissue stimulation and detection, and filtering and digitization of the tissue responses;
• Figure 1 is a schematic view of the appar- atus of the present invention.
• Figure 2 is a cross section view of the probe of the present invention.
• Figure 3 is a cross section drawing of the rear of the probe of the present invention.
• Figure 4 is a system block diagram of the components of the system of the present invention.
• Figure 5 is a cross section drawing of the tip of the probe of the present invention.
• Figure 6 is a top view of the probe tip in a preferred embodiment having a photodiode at the probe tip.
• Figure 7 is a perspective view of the probe tip of Figure 6.
• Figures 8a-8c are timing diagrams for the optical and electrical measurements made during a measurement cycle.
• Figure 9 is a timing diagram for a single electrical measurement voltage relaxation curve indicating the points in time at which measurements of the voltage amplitude are made.
• Fig. 10 is a block diagram of the synchronous detection system employed in the present invention.
• the present invention provides a method and apparatus for tissue type recognition which is useable at a variety of locations within or about a living being and which can quickly pro- prise an objective identification of the tissue types, including the presence of pre-cancerous and cancerous activity.
• the probes of this invention are designed to distinguish between tissue types when held directly against tissues in the body that are accessible without damage to the tissue. This is primarily the external covering and lining tissues that are collectively termed "epithelial tissues.”
• Epithelial membranes form the covering and lining of the major organs of the body. These epithelial layers are highly structured arrangements of cells. Under them is connec- tive tissue which is more loosely structured and which includes other components such as blood and lymphatic vessels. In turn, under these are other organ structures.
• the epithelial layer functions primarily to protect the underlying connective tissue from wear and damage. This is best exemplified by the skin but can also be seen in the lining of the intestinal, respiratory and urogenital tracts. Epithelial tissues can also have secretory and absorptive functions; for example, the lining of the respiratory tract secretes a mucous to prevent the tissue drying out, and the small intestine has the specialized function of absorbing nutrients from digested food.
• the covering or lining of many organs is easily reached from outside without puncture or tissue damage. Access can be either directly, such as to the skin, oral mucosa and the eye, or indirectly via an instrument such as a speculum to the vagina and cervix or an endoscope to the sinuses, trachea, bronchi, oesophagus, stomach, intestine, uterus and bladder.
• an instrument such as a speculum to the vagina and cervix or an endoscope to the sinuses, trachea, bronchi, oesophagus, stomach, intestine, uterus and bladder.
• the apparatus of the present invention is shown in Figure 1. It comprises a pen-shaped probe 3. / connected via a flexible probe cable 5., a probe console 7 approximately the size and shape of a laptop computer, and a removable mains pack or battery pack 9.
• the probe 3 is a hand held device about 27 cm in length at the proximal end of which (the handle) the probe diameter is approximately 2.5 cm. The device tapers towards its tip and the distal end is approximately 5 mm in diameter.
• the probe is soak-sterilizable in 2% glutaraldehyde solution.
• the connection of a serial cable is possible for purposes of data transfer and storage.
• An earphone may be used to give an audible diagnosis. Diagnostic information is presented on a Liquid Crystal Display (LCD) 11.
• a simple keyboard 13 wraps around the LCD.
• the probe is shown in longitudinal cross section in Figure 2, the rear of the probe is shown in cross section in Figure 3, and the probe tip is shown in cross section in Figure 5.
• the probe has located within an external tube 15 a central optical fiber 17 which conducts electromagnetic radiation to a photo- detector diode in the handle and which is positioned in the center of a bundle of optical fibers 19 extending from LEDs in the handle to the tip of the probe.
• Three gold electrodes 21. 23. and 25 are positioned adjacent and abutting against the internal surface of the external tube 15_.
• the probe cable 5 consists of 16 individual coaxial con- ductors with a single overall braided shield, enclosed in a medically rated silicone outer jacket. Both ends of the cable have round plastic 16 pin male connectors. In another embodiment, only 4 conductors are used and digital signals are employed.
• the electrodes 21, 23, and 25. and optical fibers 17 and 19 come into direct contact with the cervix tissue for stimulation and detection of the tissue characteristics.
• the probe tip is polished and smoothed and has contoured edges.
• An epoxy resin electrically insulates and seals the tip section.
• the hand-held probe which comes into contact with the cervix, continuously inter- rogates the cervical tissue by repetitively pulsing it with low levels of optical and electrical energy.
• Real-time interpretation of the cervix tissue response is achieved by a statistical classification algorithm in software resident in the probe console.
• the measured tissue response is then compared to a catalogue of tissue signatures and the operator informed of the result. Tissue will be classified as normal, low grade abnormality, or high grade abnormality/invasive cancer. An operator error may also be flagged.
• FIG. 4 A block diagram of the relationship between the components of the probe system is given in Figure 4, which is divided into sec- tions representing the probe, the console user interface, and the console signal processing section. All of the functional blocks that appear in the probe section of the system block diagram are implemented within the probe handle.
• LEDs 27 mounted in the handle of the probe are used as the light source to measure the level of backscattered light returning from the cervix.
• the LEDs can be positioned at the tip of the probe without the use of fibers to conduct the light to the tip.
• the LEDs are excited in turn by selecting the appropriate device via the LED wavelength switch 29.
• Light from the selected LED is carried by optical fibers 19 to the tissue under examination.
• the LEDs operate at three distinct wavelengths, red and green in the visible spectrum and infrared, to provide the light to the fibers 19.
• the resultant back- scattered light is directed via optical fiber 17 to a photodiode 31 that produces a photo- current, which is locally converted into a voltage by the preamplifier 33.
• the photodiode 31 can also be positioned, in an alternative embodiment, at the tip of the probe without the use of fibers to conduct the light from the tip to the photodiode.
• the resulting raw optical signal is received by a programmable gain synchronous detector 3_5, which under the control of a microprocessor 36.
• labeled "main routine” in the figure provides output to ' the tissue type classifier 37 in which diagnosis is accomplished together with information derived from the concurrent electrical testing information.
• the electrodes 21, 23., and 25 interface with electrode configuration switches 39, elec- trode excitation switches 49 and 51 and an electrode preamplifier 43.- This provides signals via an anti-aliasing filter 45 to the tissue type classifier 37. and operator error detector 47 to the microprocessor 3_6.
• the electrodes can thus be selected to be anodes, cathodes or high impedance (no connection) through the switch 39., which is controlled by the microprocessor 36.
• the electrode preamplifier 43. is also located in the probe handle.
• the electrode preamplifier is connected in a differential configuration to reduce the effects of common mode noise sources.
• a voltage of 1.25 volts is applied to the electrodes to charge the cervical tissue.
• the electrode supply (electrode source) 53. provides the voltage for charging the cervical tissue. This supply has suitable over-voltage and over- current protection for the safety of the patient.
• Figure 4 comprises the analog signal conditioning and the tissue classifier.
• the analog signal conditioning is responsible for the conversion of the probe signals into signals suitable to interface to the microprocessor's analog-to- digital and digital-to-analog convertors.
• the tissue classifier resides in software running on the microprocessor.
• Probe dependent calibration data is stored in a non-volatile probe memory 54, which interfaces with a calibration routine 55 stored in the microprocessor 3_6. This enables the system console to customize its response to a new probe. Encoded operational coefficients in the probe memory 54 refer to embedded particular operational characteristics and instructions that are read by the console to achieve a calibrated response from the probe. The importance of probe calibration is so the algorithm for tissue classification need not be hardware specific. Since this calibration data is stored within the probe, the console and probe do not have to be matched. Having probe specific information stored within the probe, as opposed to the console, has the advantage of an easier validation and makes it less complicated to upgrade the system via the probe than would be the case if it was required to upgrade the console.
• probe storage may include algorithm coefficients or other modular algorithm components or firmware units.
• This may be a naturally occurring mucus covering the tissue or an artificially applied conductive fluid or gel.
• the electrodes are held against the tissue so that only a thin layer of electrolyte remains between the two.
• the impedance of this thin layer is relatively low through to the tissue but relatively high between the electrodes so electric current is directed through the epithelium.
• the epithelium presents a layer of moderate impedance and beyond it is connective tissue of substantially lower impedance. The electrical measurement is thus dominated by the properties of the epithelial covering.
• the impedance of the epithelial layer at low frequency depends on its particular characteristics.
• Various mechanisms have been proposed to account for differences between tissue impedances. For an epithelial layer, this includes its thickness, the tightness of the intercellular junctions, the strength of the basement membrane, the cellular arrangement, the extracellular space (between the cells) and the composition of the extracellular fluid.
• the cell membranes capacitively couple to the intracellular spaces and so the internal composition of the cells also becomes important.
• the light which reaches the optical detector from the source must first scatter through the tissue under the tip of the probe.
• the path of scatter depends on the wavelength and affects the intensity of the light. It will be influenced by many characteristics of the epithe- lium and underlying connective tissue including the cellular arrangement, the size and shape of cell components such as the nucleus and mitochondria, the vascularization and the fluid levels in the tissue. Along this path some of the light will be absorbed by various cell components such as chromatin, hemoglobin and the opacity of the tissue. The amount of absorption is dependent on the wavelength of light. Differences between the absorption at different wavelengths can be very informative in differentiating between tissue types. During each measurement cycle, the LEDs are activated in sequence.
• the detector photodiode is used for the detection and measurement of backscattered light across the spectral range encompassed by the three LEDs. Significant background noise is encountered due to ambient light and examination lighting, and the signal to noise ratio is boosted by means of a variable gain amplifier system. Ambient light compensation is achieved by performing a set of ambient measurements immediately pre- and post- LED activation. The backscattered optical sig- nal is recovered and then filtered and digitized.
• the electrical measurements are stimulated by the delivery of 1.25 volt electrical pulses of 250 ms duration. Following removal of the applied electrical potential, the residual charge dissipates within the tissue with a decay constant dependent on tissue capacitance, the electrode/tissue interface and electronic and ionic conductance. This "relaxation curve" is characteristic of the underlying tissue type ( Figure 9) . The shape of the electrical relax- ation curve is also highly dependent on hardware-specific features including the electrode material composition, surface composition and position. The measured tissue response is fil- tered, digitized at 9 ms intervals, and thereafter processed in the probe console.
• the electrical pulses are applied across these with periodic reversal of polarity to minimize electrochemical degradation.
• the three electrode configuration is the more general form of the device, the typical measurement cycle is described below with reference to that form.
• the electrical pulses are delivered across varying combinations of these electrodes. In each case, one electrode is active while the remaining two act as a reference. Electrical pulse delivery and the corresponding relaxation curve measurements are continually cycled through the three possible electrode combinations. This feature allows the detection of conditions which result in an asymmetrical charge imbalance between electrodes, such as partial contact. In addition, electrode cycling minimizes electrochemical degradation.
• Each tissue observation incorporates several relaxation curves recorded for each of the three electrode configurations. After each series of measurements an electrode discharge cycle is implemented.
• Figures 8a-c show a typical three electrode measurement cycle, which takes 71.43 ms, i.e. 14 cycles per second, and is divided into nine intervals.
• first (“calibration") interval (0 - 4 ms) an internal calibration of the instrument takes place.
• Calibration of the console is adjustment of the console's electronics so its performance and behavioral characteristics are consistent be- tween consoles.
• Calibration of the electrical offsets is to eliminate probe variation due to different probes and temperature variation as well as to reduce the need for factory calibration. This calibration step enables less costly and lower power circuitry to be used.
• Electrode circuitry calibration is carried out by applying a test signal to the probe, and then measuring this value with an analog to digital converter and adjusting the offset using a digital to analog converter until the correct value is obtained.
• the calibration is carried out under microprocessor control. This method is a successive approximation type of search which reduces the calibration time from 2 n iterations to n+1 iterations. This is depicted schematically as three disconnected terminals in a circle in Fig. 8a.
• a current (the inrush current) is injected respectively from one of the three possible probe tip electrode configurations (in which one electrode is at anode potential and two are cathodes) .
• the temperature of one of the three LEDs is determined at the same time.
• the forward bias voltage of a semiconductor diode is temperature dependent. This temperature dependence is due to the variation in the semiconductor bandgap.
• the optical output of an LED is also temperature dependent. To make, accurate measurements of the backscattered light from the cervix, the output of the light source needs to be either constant or known.
• the optical output of an LED can be determined by the LED's temperature and drive current if that LED has been characterized.
• the need for determining the temperature of the LED light source is critical as the short term environmental temperature changes are likely to exceed 20° C.
• the optical output of uncompensated LEDs are likely to vary by more than 20% under these conditions leading to a very inaccurate measure of backscattered light.
• the temperature of an LED junction can be determined by measuring its bandgap related potential, that is, the forward bias with a known current thus avoiding the need for a separate temperature sensor for each LED.
• the present invention's novel approach of leaving the LED's optical output unregulated and compensating the detector' s gain is superior to prior techniques for compensating LED output, e.g.
• An advantage of this alternative embodiment is that the intensity of the light may be measured and cor- rected using signals that are detected while the LEDs are being pulsed rather than using data from a separate measurement as is done for the bandgap. Where instantaneous correction is not desired during each pulsing sequence the average intensity could be corrected using accumulated data.
• 5 th and 7 th (“optical measurement") intervals (10.5-18.5 ms, 25-32.5 ms, 39-47 ms) , one of the three LEDs whose temperature has been determined emits light the backscattering of which is simultaneously detected.
• the surface under examination is discharged, the data analysis algorithm is executed, and the user interface is updated.
• each of the three current measure- ment intervals four square current pulses of approximately 250 ms duration are employed, separated by 1.8 ms. Three measurements are made of the decay amplitude of each of the first and fourth current pulses during the time prior to the second pulse or prior to the end of the current measurement interval. Thus a series of 18 electrical measurements of pulse decay are made in each 71.43 ms cycle. A set of parameters is generated to parameterize the shape of the inrush current and voltage decay curves in each interval such as with a multiple exponential best fit.
• Alternative shape parameterizations include transforming the data with ordinate and abscissa operators such that they become piecewise straight line segments. Such operators include taking logs so as to produce log/log displays, using inverse time as the abscissa or any other transformations that provide good fit to the data. Parameters associated with the transformed functions can then be associated with the degree of tissue abnormality. Typical operations that can be applied to the curves and variables that can be extracted for use as discriminants are as follows: 1. The slope and intercept of the log voltage/inverse time plot of the curves.
• a total of 21 tissue classification parameters (18 electrical and 3 optical) are extracted from the digitized optical and electrical data, in addition to various parameters extracted for the detection of poor contact conditions. Some of the electrical parameters are functions derived from various portions of the measured relaxation curves. These parameters are then passed to the processor chip for classification. With 21 parameters processed per observation, the total rate of parameter processing is 294 per second. Assuming that 1000 observations are processed per patient, the total number of parameters under consideration is approximately 20,000.
• the apparatus of the present invention categorizes biological tissue by having a probe tip able to select a tissue surface area by contact and applying a group of sequential current pulses from the probe tip to each of a succession of selected tissue surface areas.
• the sequential pulses occur within groups that occur at a rate fast enough so that they are applied to substantially the same tissue surface area.
• a circuit then derives values for a group of parameters that indicate the response to the group of sequential current pulses applied to each selected tissue surface area.
• a memory stores a catalog of tissue types that are associated with respective subsets of groups of parameter values.
• the processor compares the group of parameter values that indicate the response of the selected tissue sur- face area with the stored subsets of groups of parameter values to categorize the tissue surface area.
• the parameters in the parameter group are not necessarily associated on a one-to-one basis with the sequential current pulses in the current pulse group.
• the successive groups of sequential current pulses may be separated in time from each other by a time interval substantially greater than the time interval between the sequential current pulses within an individual group.
• multiple measurements of the tissue potential are taken during decay of the potential following application of the current pulse.
• the system permits at least two parameter values to be derived during the potential decay following each current pulse for which a tissue response is desired thereby allowing a more sophisticated param- eterization of the current decay than a simple exponential.
• Enough measurements are made during the current decay so that each of the parameters may be derived from several of the multiple measurements taken during the decay of the current pulse for which a tissue response is desired. These multiple parameters are then available so that the processor can categorize any tissue surface in accordance with at least two parameter values derived during the poten- tial decay following each of at least two current pulses. In general these two current pulses are separated by at least one other current pulse which is not used by the processor to categorize the tissue.
• the aforementioned pulses are applied by three electrodes. This is done so that non-overlapping current pulses flow between different groups of electrodes and corresponding current pulse applications and measurement cycles occur for different groups of electrodes.
• 9886 corresponding parameter values derived following the current pulses for different groups of electrodes are combined for the categorization of the tissue surface area by the processor.
• Other electrode configurations for example, two or four, would require modifications to the sequence as described.
• the optical and electrical measurements on the same tissue be in- terspersed and that the charge dissipation in the tissue volume underneath a selected tissue surface is not complete by the time the next sequential current pulse is applied.
• This more complex probing by electrical pulses creates a more subtle response to the probing and allows greater discrimination of tissue characteristics.
• the pulses are preferably separated in time from each other by a time interval substantially greater than the time interval between sequential current pulses within an individual group so that the cate- gorizations of successive selected tissue surface areas are substantially independent of each other. Averaging of test results.
• the timing of the various events aids the diagnostic abilities of the invention.
• the first pulse provides the response of the tissue to a current pulse after the tissue has had an opportunity to discharge from the previous measurement interval. Indeed the first pulse of the first measurement interval has had the longest time to recover and perhaps to recover completely.
• the different timing of the pulse recovery times permits tissue at different depths below the surface to influence the measured parameter values.
• the cumulative effect of these different recovery times is determined in the present invention by averaging the responses. Thus some information is lost, but a wider range of effects influence the final result. In an alternative embodiment this averaging is not performed and the greater information content is utilized. Allowing tissue charge from prior tests to dissipate.
• the timing of the various electrical measurements into intervals separated by optical measurement intervals allows a short recovery time after each current measurement interval. Furthermore the lengthy discharge interval permits a more total recovery of the tissue so that individual cycles can maintain independence from one another.
• the three electrical probe tip elements are made active cathodes and kept at low impedance. This is quite contrary to the normal construction of measuring electrodes where the impedance is kept high so that the current characteristics of the object being measured are effectively isolated from the current flow in the measuring instrument. Essentially the benefit of isolation is traded off for the rapidity of recovery of the tissue for the next measurement cycle.
• the order of performance of the optical and the electrical tests also has the beneficial effect of reducing the overall observation time required for each measurement.
• the inactive period between electrical measurements is used for the optical measurements and vice versa.
• the measurement of LED bandgap potential and the subsequent compensation for temperature variation characterized by the bandgap potential requires little computational bandwidth and does not interfere with the rapidity of measurement necessary to characterize each electrical decay curve.
• only eight readings are shown as being taken of the current flow into the electrodes (the inrush current) during the early part of the 250 ms applied pulse. When it is desired to make additional use of the inrush current readings, it will be appropriate to take current readings throughout the 250 ms pulse.
• Figure 9 depicts an individual voltage relaxation curve. As indicated an initial offset voltage is determined by eight consecutive observations sampled at 9 ⁇ s intervals. The height of the square wave pulse is similarly measured by eight consecutive observa- tions sampled at 9 ⁇ s intervals. During the voltage decay samples are taken at 9 ⁇ s intervals, but not all are recorded. Fig. 9 also shows the corresponding current relaxation curve. In this example, only eight readings are shown as being taken of the current flow into the electrodes (the inrush current) during the early part of the 250 ms applied pulse. When it is desired to make additional use of the inrush current readings, it will be appropriate to take current readings throughout the 250 ms pulse. Distinct values from two different decay curves.
• the use of the first and fourth electrical measurement in each set of four as distinct variables without averaging allows recovery of the maximum amount of information from the electrical measurements. This information is utilized in the statistical analysis of the electrical and optical data.
• the optical system measures the bandgap potential of a first LED by applying a small current to the LED and measuring the potential across it. This provides a readout of the temperature of the LED and per- mits correction for temperature variation to be made.
• a 6th signaling category indicates whether the device is working within specifications.
• the method of signaling a diagnosis is via four approaches, namely, the display on the console, LED indicators on the rear of the probe, audible tones via head- phones and a summary printout of the diagnosis.
• the console display mimics that of the LED output with the addition of labeling. In this way the console will serve as an alternative display of diagnosis and as a reference to the meaning of the LED configuration on the rear of the probe.
• the audible signal also follows the same pattern as the LED output, however, using tones, for example, the tones will shift to a higher pitch for a more significant classification.
• the printout summarizes the diagnosis.
• the algorithm first checks for poor contact, and if detected, the operator is signaled via the probe handle lights and the console, and no diagnosis is attempted. As the process is repeated at a rate of 14 times per second, the operator receives instantaneous feedback on the probe position and may adjust device positioning accordingly.
• the poor contact check includes the following conditions: (1) the probe being at an angle to the cervix; (2) the probe partially or fully lifting off the cervix, or lift-off; (3) the probe moving too quickly across the cervix for accurate measurements to be performed, or slip; and (4) the probe is positioned over a junction between tissue types.
• the angle and junction conditions are detected through an imbalance in the electrical parameters, while the lift-off condition is detected by means of out of range electrical and optical readings.
• HSIL High Grade Squamous Intraepithelial Lesions
• IC Intraepithelial Cancer
• LSIL Low Grade Squamous Intraepithelial Lesions
• An initial validity check on the data is performed to ensure that the multivariate data distribution is within the limits of all valid classifier data. If the result signals one out of range, then no diagnosis will be made and the operator is signaled.
• a most probable tissue type is then selected.
• a further validity check is performed to ensure that the multivariate data distribution is within the limits of all valid classifier data for the selected tissue type. Again, if the reading proves to be an outlier, no diagnosis is performed and the operator is signaled. The probability estimate (certainty of assignment to a particular tissue type) is then assessed against a pre-defined decision threshold. If the probability estimate is below the threshold, no diagnosis is performed. Again, because the measurements occur at the rate of 14 per second, the operator receives instant feedback.
• HSIL High Grade Abnormality
• LSIL Low Grade Abnormality
• the pre-determined decision thresholds define Receiver Operating Char- acteristic (ROC) curves.
• the ROC curve is a graphical description of test performance representing the relationship between the true positive fraction (sensitivity) and the false positive fraction (1 - specificity) .
• An increase in the decision threshold will cause an overall increase in device specificity at the expense of sensitivity, and vice versa.
• the first level decision threshold concerns the probability estimate used for the classification of tissue into one of 17 types.
• the second level decision threshold concerns the grouping of tissue types into categories, whereby the grouping can be adjusted, depending on the desired outcome of the screening test. Appropriate adjustment of the decision threshold allows the configuration of an optimal trade off between sensitivity and specificity, with a particular focus on the cut-off between low grade changes and minor atypia. Safety and Reliability of the System
• a number of features have been developed to ensure the safety of the patient and long term reliability of the probe system. These include calibration procedures, temperature compensation and electrical safety precautions. It is necessary to calibrate each probe during manufacture in order to ensure that optical and electrical output signals are the same for each device. Optical calibration is performed in a turbid solution of stable optical characteristics with an optical spectral distribution chosen to simulate that of cervical tissue, and electrical calibration is performed using a stable electrolyte solution. An optical calibration check is also performed at the beginning of each clinical session. The operating temperature of the probe is 5 to 50 degrees Celsius. Temperature compensation of the LEDs is necessary since the optical measurements are extremely sensitive to the ambient temperature. Stability of opera- tion across the required temperature range is achieved by continuous automatic measurement of the temperature and compensatory adjustment.
• the database includes a number of data subsets for each tissue type and for Poor Contact condi- tions, including contact problems induced by excess cervical mucus or blood.
• Data collection for algorithm development proceeds by means of a data collection system incorporating a link from the console to a coir puter for the download of digitized data, and a video mixer, recorder and printer. Probing is performed, followed by formal colposcopy with aqueous acetic acid -staining of the cervix, and the session is recorded on video. A colpo- photograph is taken after acetic acid staining, and the colposcopist marks the diagnosis of all tissue types present on the photograph. Patient history and current status information, including Pap smear, colposcopy and biopsy results are recorded on a clinical record form and subsequently entered into the probe database.
• the data are analyzed in the laboratory by viewing the probe session video concurrently with a display of optical and electrical parameters.
• Colposcopy and biopsy results from participating clinics are subject to a uniform review process in order to reconcile colposcopic and histological diagnoses. Briefly, video images taken during the colposcopy session and histology results, if available, are reviewed by an independent colposcopist. Referral to a second colposcopist is performed in cases of an initial abnormal diagnosis and in cases of doubt. Where a reference diagnosis cannot be established, data are excluded from the algorithm database.
• the 17 tissue classification categories are used in the establishment of the Reference Diagnosis. Tissue type classification is based on the colposcopic classification of Coppleson, Pixley and Reid (Coppleson M, et al.
• the present invention is designed to be a screening, rather than a diagnostic tool. Therefore, the tissue types are grouped into categories which are of use for the clinician when making the referral decision. These categories are: Probe Cancer or High Grade Abnormality; Probe Low Grade Abnormality; and Probe Normal. Note that the tissue types identified as original squamous epithelium with HPV stigmata (tissue types 7 to 9) may potentially be grouped into either of the output categories of Probe Low Grade Abnormality or Probe Normal, depending upon the desired screening result. The two options effectively correspond to alternative operating points on the device receiver operating characteristic.
• the programmable gain synchronous detector 35 receives the raw optical signal and provides output to the tissue type classifier 3_7.
• Synchronous detection is a demodulation process in which the original signal is recovered from a noisy transmission path by multiplying the modulated signal by the output of a synchronous oscillator locked to the carrier. This technique traditionally is used in the communi- cations field for demodulation of amplitude modulated signals.
• Many sources of interference are present when making measurements of the backscattered light from the cervix. These sources of interference are both electrical and optical in nature. Notably, the luminous intensity of colposcope light is far greater than of the light source used by the probe. Until synchronous detection was employed, the signal chain would often saturate.
• Fig. 10 is a block diagram of the syn- chronous detection system employed in the present invention.
• the synchronous oscillator 81 provides both the drive for a typical LED 83 and the synchronizing signal for the detector.
• the oscillator' s frequency is 4 kHz and is thus away from the frequencies of the most common noise sources.
• the photodiode 85 is used to detect the backscattered light from the cervix.
• the photodiode is located in the probe along with a low gain transimpedance amplifier. The gain of this stage is kept low to avoid saturation by ambient light sources.
• the return signal is accompanied by two common noise sources.
• the first is lighting ripple from an illumination source like a colposcope
• the second is random thermal noise.
• the high pass filter 87 is used to remove the steady-state light. It is also effective for reducing the low frequency ripple component from the colposcope illumination, thus avoiding saturation of the following signal processing stages.
• the programmable gain amplifier 89 is used to normalize variations in the LED's optical output. 00/19886
• the multiplier 91 correlates the real signal component of the photodiode signal while randomizing the noise component.
• the low pass filter 92 takes the multiplied signal and pro- vides an average. This helps separate the correlated signal from the uncorrelated signal (noise) .
• the low-pass filter also sets the bandwidth of the signal processing chain. The lower the cut-off frequency the more narrow the bandwidth and hence the greater the rejection. However, if the bandwidth is made too narrow then the system will take a long time to respond. Rather than the traditional integrator or first-order low-pass filter, a high order Bessel filter has been used in the probe's synchronous detector, thus giving excellent out-of-band noise rejection as well as good transient performance.

Landscapes

• Investigating Or Analysing Materials By Optical Means (AREA)
• Investigating, Analyzing Materials By Fluorescence Or Luminescence (AREA)

Applications Claiming Priority (1)

Publications (1)

Family

ID=22268008

Family Applications (1)

Country Status (4)

Families Citing this family (2)

• 1998
  • 1998-10-02 EP EP98949755A patent/EP1115326A1/en not_active Withdrawn
  • 1998-10-02 JP JP2000573249A patent/JP2002526136A/ja not_active Withdrawn
  • 1998-10-02 CA CA002312743A patent/CA2312743A1/en not_active Abandoned
• 2002
  • 2002-02-07 HK HK02100445.9A patent/HK1039443A1/zh unknown

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events