Abstract: A suite of tests and parameter monitoring methods for at least five major subsystems that are found in most automated image processing systems, as well as calibration routines for three major system functions. The five areas of subsystem verification include processing quality, illumination quality, image collection quality, autofocus quality, and position quality. An automated biological specimen analysis system uses self measures of system parameters to provide a system status. The system status is reported to the user.

Abstract: A method for identifying the size, shape, and location of objects in a specimen where the image of the specimen is represented by image data and where the image data is processed to provide mask data representing a mask where the mask identifies the size, shape, and location of the object. Generally, the method includes the step of enhancing the image and creating a threshold image where the threshold image includes a threshold intensity value associated with each pixel of the image. The threshold image is combined with the original image to provide a mask image that identifies the size, shape, and location of the objects. The mask image may be further refined to ensure accurate identification of the object. Various other techniques are disclosed within the general method for processing image data.

Abstract: A computer receives image data from a star-shaped optical target in the object plane and calculates angle-dependent boundary sharpness. The horizontal, x-direction, amplitude derivative and the vertical, y-direction, amplitude derivative are computed over a portion of each star pattern image from a Z panning sequence. A microscope slide stage, carrying the target, is moved vertically from a level just below where the target is in focus to a level just above where the target is in focus. For each small increment of vertical motion, Z panning, an image of the star pattern is captured for analysis. Computations are performed on the differentiated images to search for evidence of elongation of the point spread function and variation with stage Z position of the angle of long axis of such an out-of-round point spread function. The presence of a distorted point spread function that varies along the optical axis indicates astigmatism.

Abstract: An automated method for checking cytological system autofocus integrity. The automated method includes the steps of checking focus illumination integrity, checking focus camera Modulation Transfer Function, checking focus camera position integrity, and checking closed loop accuracy. Checking focus illumination integrity includes checking focus illumination system integrity, and checking a focus noise floor level. Checking focus camera position integrity includes checking focus camera longitudinal separation, and checking focus camera lateral separation. Checking focus camera position integrity includes checking focus filter frequency response.

Abstract: A biological specimen classification strategy employs identification and integration of multiple cell patterns. An automated microscope acquires an image of a biological specimen such as a Pap smear and provides an image output to biological classifiers. The classifiers independently detect and classify a number of specimen types and provide classifications to an output field of view integrator. The integrator integrates the classifications. The integrated output then determines whether the classifiers should be reapplied to the image.

Abstract: Coverslip detection locating all four coverslip edges. A field of view processor receives image data from a charge coupled device camera. The charge coupled device camera images a slide and coverslip that is mounted on a movable frame. The slide and coverslip are illuminated from below with a uniform light source. The moveable frame is under computer control and moves in response to the field of view processor. The field of view processor locates the coverslip by first positioning the movable frame to view a portion of the slide on a predetermined potion of the slide within a predetermined area of the slide. The slide is then re-imaged after the movable frame moves the slide toward a chosen edge of direction. Edge type objects are located and followed over multiple fields of view. If the edge object satisfies a set of predetermined criteria the coverslip edge has been found. The edge is extended to find all four corners of the coverslip. The edge objects are processed using morphological operators.

Abstract: A computer aided biological specimen screener is responsive to a priori information about biological specimens. An automated slide classifier scans a slide specimen and generates a set of scores. The classification which results from this set of scores depends on a priori information concerning the patient, such as patient age and prior health history and diagnosis. Using the patient information and an appropriate score, the specimen is classified. A normal threshold used for classification is changed over a wide range depending on the specificity appropriate to a particular patient risk category. The sensitivity of the classification is automatically maximized for a predetermined normal specificity.

Abstract: An automatic screening apparatus that screens all pap smear slides called normal by cytotechnologist. These normal pap smears will contain the mistakes, false negatives, which must be recovered as a measure of the cytotechnologist's performance. Additionally, these mistakes must be discovered to improve the quality of the laboratory services. The automated screening machine after rescreening the normals, rank orders of the pap smears with those "normals" most likely to be missed ranked at the top. Another human screener re-examines those pap smears with high ranking to determine if the first human screener misclassified the pap smear. Mistakes are recovered and the performance of the first screener is assessed. Proficiency testing is accomplished with 100% quality control.

Abstract: A network system for review and analysis of computer encoded microscope slides and specimens which were originally computer encoded from a microscope (attached via an encoder device to a local computer site), during an initial examination. The encoding includes parameters of viewing locations and events of interest on the slide, with such information being stored on a networked file server. The encoding also includes information regarding the manner in which the initial examination was conducted, for quality control purposes. The computer encoded information is retrievable at all remote locations of the network (either local or connected via modem) for supervisor review or for pathologist analysis. The network is further constituted by microscope sites having similar computer encoding devices attached thereto, which function, in this aspect, as computer terminals of the network. For enhanced analysis, the computer terminals have direct access to patient background information, e.g.

Abstract: A computerized slide encoder for use with microscope analysis and pathological studies is provided. The slide encoder is attached to the movable microscope stage, whereby x-y direction plane movement and location, is correlated to examination of an identified slide, with information marking and location being directly correspondingly written, at pre-defined times, on computer storage media, during the examination. Subsequent use of the computer-stored information, coupled with the slide encoder, in a slide re-examination, permits independent retrieval of such information and location on the slide. The computer storage media includes digitized vocal transcription of history of the slide made during the original examination. An embodiment of the slide encoder includes a computer mouse pen affixed to the stage of the microscope and a rolling surface, for the mouse pen, being affixed to the movable slide holder.

Abstract: Image enhancement for a digital image, such as an image of a biological specimen mounted on a microscope slide or a cytological specimen is herein described. A digital representation of the image is dilated to produce a dilated image and eroded to provide an eroded image. The dilated image, eroded image and the digital representation of the image are then processed to produce an enhanced image.

Abstract: Method and apparatus for identifying normal biomedical specimens. Image data is gathered representing an image of the specimen. The image data is processed to measure features of the objects to determine whether the object is normal. The feature measurements for all objects in the specimen are then combined to determine whether the specimen as a whole is normal.

Abstract: Reference information for a biological slide is obtained. The reference information normalizes the measured object features. Calibrated feature measurement, not based on absolute measurements, self adjusts to match the situation of each slide, where each slide is characterized by the reference information. The reference may be different from slide to slide because of the preparation variations. The calibrated features will not carry the inter-slide variations. In addition, the reference information provides a good indication of the slide condition such as dark stained, air dried, etc. which can be used as slide features for the specimen classification. No alteration of the current practice of specimen preparation is required. The many slide context dependent features improve the classification accuracy of the objects in a specimen.

Abstract: A method for high speed morphological processing in a computerized image processing system starts by acquiring a binary image and selecting feature and height values corresponding to a desired morphological processing of an image. The image is scanned in a raster scanning sequence so as to produce a scanned image including a plurality of pixels. A sequential labeling transformation is performed on each of the plurality of pixels of the scanned image so as to produce a transformed image. A thresholding operation is performed on the transformed image. The resultant image is translated.

Abstract: A sensor for locating the position of and measuring the velocity of an impacting object. The sensor employs a net attached at each edge to sensor modules mounted on a frame. Each sensor module includes an output drum, a take up drum, a spring and a motor. The net is attached to the sensor modules by cords wound on a spring loaded output drum. The drum is coaxial with a motor that generates a voltage proportional to the angular velocity of the drum. The diode isolated motor partially dissipates the energy of impact to heat on the return stroke in response to the restorative force of the spring. An object impacting a net draws the cord from the output drum, generating a voltage from an attached motor. A maximum voltage or an integrated voltage generated by the motors determines the velocity of impact. Each particular net has an associated set of axes. A pair of motors, their position known relative to the frame, are placed along these axes attached to one or more strings by a cord.

Abstract: An automated method for checking cytological system autofocus integrity. The automated method includes the steps of checking focus illumination integrity, checking focus camera Modulation Transfer Function, checking focus camera position integrity, and checking closed loop accuracy. Checking focus illumination integrity includes checking focus illumination system integrity, and checking a focus noise floor level. Checking focus camera position integrity includes checking focus camera longitudinal separation, and checking focus camera lateral separation. Checking focus camera position integrity includes checking focus filter frequency response.

Abstract: An automated method for checking cytological system autofocus integrity. The automated method includes the steps of checking focus illumination integrity, checking focus camera Modulation Transfer Function, checking focus camera position integrity, and checking closed loop accuracy. Checking focus illumination integrity includes checking focus illumination system integrity, and checking a focus noise floor level. Checking focus camera position integrity includes checking focus camera longitudinal separation, and checking focus camera lateral separation. Checking focus camera position integrity includes checking focus filter frequency response.

Abstract: Field of views of a slide are examined to assess the likelihood of existence of detectable single cells, groups, and thick groups of cells to locate objects of interest by an automated microscope. The FOV features consists of features selected from the distribution profiles of size, shape, layout arrangement, texture and density of all objects within a FOV which are compared against pre-determined criteria.Each field of view is assigned a likelihood value based on FOV features. Areas that are blank, or contain air bubbles, or are too dense for analysis are identified and excluded for further analysis. Each FOV is ranked according to its likelihood of containing SIL (Squamous Intraepithelial Lesion) cells or cell groups of interest. These results, such as SIL, single cell ranking, group ranking are used to arrange the further examination of FOVs in a priority order.

Abstract: A classifier measures robustness responsive to object presentation effects and decision boundary effects. A cytological image analysis computer obtains objects of interest and classifies them responsive to a decision tree classifier. The robustness of classification is calculated dynamically as objects are classified responsive to a segmentation robustness and a classification decisiveness measure. The results of the decisiveness measure and the segmentation robustness data are combined to provide enhanced overall classification reliability.

Abstract: A robust biological specimen classifier. An automated microscopy system obtains features from an image of a biological specimen slide. A computer system in the automated microscopy system computes feature variations. Clusters are created that comprise samples of similar characteristics. A cluster membership generator generates membership values for each cluster. Classifiers specialized to certain feature sets are used to provide independent outputs. These outputs are offset and biased by the output of the membership generator. The output of the adjusted classification values are summed to create a slide score output.