Abstract: In a preferred embodiment a radiologist traces the pulmonary artery and pulmonary veins visible in a set of CT images and identifies the arteries and veins. The radiologist's identification of the pulmonary arteries and pulmonary veins is then received by an image analyzer and combined with the analyzer's identification of the pulmonary arteries to form a combined identification; and the analyzer then reviews this combined identification of the pulmonary arteries to detect any pulmonary embolisms. The radiologist's identification of any pulmonary embolisms is compared with the analyzer's identification of any pulmonary embolisms to determine if there are any embolisms identified by the analyzer that were not identified by the radiologist.

Abstract: Detection inputs are classified using two thresholds. In a preferred embodiment, a multiplicity of inputs are scored and the scored inputs are searched to locate an input that has a score greater than a first threshold. If such input is found, every input having a score in excess of a second threshold lower than the first threshold is identified as belonging to a first class of interest. If no input is found having a score in excess of the first threshold, no inputs are identified as belonging to the first class.

Abstract: Detection inputs are classified using two thresholds. In a preferred embodiment, a multiplicity of inputs are scored and the scored inputs are searched to locate an input that has a score greater than a first threshold. If such input is found, every input having a score in excess of a second threshold lower than the first threshold is identified as belonging to a first class of interest. If no input is found having a score in excess of the first threshold, no inputs are identified as belonging to the first class.

Abstract: An intuitive user interface is provided that allows for selection of an abnormality level in features in a medical image on the basis of graphical depictions of the abnormalities themselves. In one embodiment, the user interface displays realistic images of actual abnormalities in a series ranging from least severe to most severe and the medical practitioner selects the abnormality level by mouse-clicking on one of the images. The user interface then displays an annotated map of suspected abnormalities with annotations only on those suspected abnormalities that are at least as severe as the selected abnormality. In alternative embodiments, the user interface displays stylized images of the abnormalities plotted by the display using mathematical functions. In yet another alternative, the images are hand-drawn images.

Abstract: In a preferred embodiment a radiologist traces the pulmonary artery and pulmonary veins visible in a set of CT images and identifies the arteries and veins. The radiologist's identification of the pulmonary arteries and pulmonary veins is then received by an image analyzer and combined with the analyzer's identification of the pulmonary arteries to form a combined identification; and the analyzer then reviews this combined identification of the pulmonary arteries to detect any pulmonary embolisms. The radiologist's identification of any pulmonary embolisms is compared with the analyzer's identification of any pulmonary embolisms to determine if there are any embolisms identified by the analyzer that were not identified by the radiologist.

Abstract: An object in a video sequence is tracked by object masks generated for frames in the sequence. Macroblocks are motion compensated to predict the new object mask. Large differences between the next frame and the current frame detect suspect regions that may be obscured in the next frame. The motion vectors in the object are clustered using a K-means algorithm. The cluster centroid motion vectors are compared to an average motion vector of each suspect region. When the motion differences are small, the suspect region is considered part of the object and removed from the object mask as an occlusion. Large differences between the prior frame and the current frame detect suspected newly-uncovered regions. The average motion vector of each suspect region is compared to cluster centroid motion vectors. When the motion differences are small, the suspect region is added to the object mask as a disocclusion.

Abstract: Red, Green, Blue (RGB) pixels in a Beyer pattern are converted to YUV pixels by a converter. The converter does not interpolate RGB pixels to fill in missing RGB color values but instead performs interpolation during conversion to YUV. An edge-enhancement filter is applied to the preliminary Y values to generate final Y values with sharpened edges. The final Y values are combined with R or B pixels from the Beyer pattern to generate U and V chrominance values. Since the preliminary luminance Y values are edge-enhanced, and then the edge-enhanced Y values are used to generate the U, V, values, enhancement improves U and V values as well. Rather than use full-frame intermediate buffers, a 7-line RGB buffer, a 5-line preliminary Y value buffer, and a 3-line final Y buffer can be used.

Abstract: An object in a video sequence of frames is tracked by object masks generated for frames in the sequence. Macroblocks are motion compensated. Blocks matching entirely within a prior-frame object mask are used to generate an average object motion. When the average motion is below a motion threshold, frames are skipped at larger intervals, but more frequent frames are processed when high motion occurs. When the macroblock best matches a prior-frame block that has the object's boundary passing through the block, the macroblock is uncertain and is sub-divided into smaller sub-blocks that are again motion compensated. Sub-blocks matching blocks within the object mask in the base frame are added to the new object mask for the current frame while sub-blocks matching a block containing the object boundary are uncertain and can again be sub-divided to further refine the object boundary. Frame skipping and adaptive-size blocks on the object boundary reduce computational load.

Abstract: Bayer-pattern pixels captured by an image sensor have only one of the three primary colors (RGB) per pixel location. Rather than interpolate the Bayer-pattern to generate the missing RGB color components for each pixel location, a direct conversion is performed to YUV pixels. A luminance calculator receives a 3×3 block of Bayer-pattern pixels and generates a luminance (Y) pixel for the center pixel location. Different coefficients are multiplied by each of the 9 Bayer-pattern pixels before summing to produce the center Y pixel, depending on the pattern location. A chrominance calculator first receives a 3×3 block of Y pixels generated by the luminance calculator. The 9 Y pixels are averaged to produce an average luminance. Two red or blue pixels in the 3×3 block are averaged and the average luminance subtracted. Then a constant is multiplied to generate the U and V pixels. Intermediate interpolated RGB avoided.