Abstract: Described herein is a framework for automatically classifying a structure in digital image data are described herein. In one implementation, a first set of features is extracted from digital image data, and used to learn a discriminative model. The discriminative model may be associated with at least one conditional probability of a class label given an image data observation Based on the conditional probability, at least one likelihood measure of the structure co-occurring with another structure in the same sub-volume of the digital image data is determined. A second set of features may then be extracted from the likelihood measure.

Abstract: A method for computer aided detection of pulmonary emboli includes acquiring medical image data. A pulmonary embolism candidate comprising a cluster of voxels is identified. It is determined whether the candidate is a true pulmonary embolism or a false positive based on a spatial distribution of intensity values for the voxels of the cluster of voxels. The pulmonary embolism candidate is presented to a user when the candidate is determined to be a true pulmonary embolism.

Abstract: A system for automatically detecting pulmonary emboli from medical image data includes receiving image data, automatically detecting one or more pulmonary embolism candidates from the image data, segmenting an airway tract from the image data, segmenting an artery structure from the image data, calculating a distance between each of the candidates and a nearest portion of the segmented airway, determining whether each of the candidates is within or outside of the segmented artery structure, rejecting candidates based on the calculated distance between each of the candidates and the nearest portion of the segmented airway and the determination as to whether each of the candidates is within or outside of the segmented artery structure, and indicating the location of the non-rejected candidates within the image data.

Abstract: A method for modeling an image for multiple tasks includes providing an image with n image features, providing an indicator matrix which has m non-zero components corresponding to m selected features selected from the n image features, constructing a model of the image using the in selected features for each specific labeling task. There is a variable for a specific task to be performed on the image and a variable for a plurality of tasks to be performed on the image.

Abstract: A method for detecting pulmonary embolisms in computed tomographic (CT) angiography includes providing a digitized (CT) image acquired from inside a pulmonary vessel, the image comprising a plurality of intensities corresponding to a 3-dimensional grid of voxels, for each voxel in the image, extracting a first pulmonary embolism (PE) candidate and PE boundary from the voxel, and for each voxel in the PE boundary, selecting a voxel from the PE boundary, extracting a subsequent PE candidate and PE boundary from the voxel, merging the subsequent PE candidate with the first PE candidate, and merging the subsequent PE boundary with the first PE boundary.

Abstract: A method for training classifiers for Computer-Aided Detection in medical images includes providing an image feature training set {(xi, yi)}i=1l, wherein xi?Rd are input feature variables and yi?{?1,1} are class labels, and a cascade of K classifiers to be trained, minimizing, for each classifier k, a first cost function to initialize an ?k0 associated with each classifier k, fixing all classifiers except classifier k and minimizing a second cost function to solve for ?kc for a counter value c using the training dataset {(xik, yi)}i=1l, calculating a third cost function Jc(?lc, . . . , ?Kc) for each classifier k, and comparing Jc with a previous iteration Jc?1, wherein if Jc?Jc?1 is less than a predetermined tolerance, said classifier training is completed.

Abstract: Described herein is a framework for automatically classifying a structure in digital image data are described herein. In one implementation, a first set of features is extracted from digital image data, and used to learn a discriminative model. The discriminative model may be associated with at least one conditional probability of a class label given an image data observation Based on the conditional probability, at least one likelihood measure of the structure co-occurring with another structure in the same sub-volume of the digital image data is determined. A second set of features may then be extracted from the likelihood measure.

Abstract: Automated and semi-automated systems and methods for detection and classification of structures within 3D lung CT images using voxel-level segmentation and subvolume-level classification.

Abstract: A method for detecting an object within a structure includes performing tobogganing on image data to obtain one or more voxel clusters and to provide a rough indication of the structure. Each of the obtained voxel clusters is characterized as an object candidate and a set of features are determined for each object candidate. Correlations between pairs of the object candidates are measured. Each of the object candidates is classified as either a true object or a non-object based on the set of features and the measured correlations.

Abstract: A method for computer aided detection of anatomical abnormalities in medical images includes providing a plurality of abnormality candidates and features of said abnormality candidates, and classifying said abnormality candidates as true positives or false positives using a hierarchical cascade of linear classifiers of the form sign(wTx+b), wherein x is a feature vector, w is a weighting vector and b is a model parameter, wherein different weights are used to penalize false negatives and false positives, and wherein more complex features are used for each successive stage of said cascade of classifiers.

Abstract: Hierarchal modeling is used to distinguish one state or class from three or more classes. In a first stage, a normal or other class is distinguished from a diseased or other groups of classes. If the results of the first stage classification indicate diseased or data within the groups of different classes, a subsequent stage of classification is performed. In a subsequent stage of classification, the data is classified to distinguish one or more other classes from the remaining classes. Using two or more stages, medical information is classified by eliminating one or more possible classes in each stage to finally identify a particular class most appropriate or probable for the data.

Abstract: A method for computer aided detection of pulmonary emboli includes acquiring medical image data. A pulmonary embolism candidate comprising a cluster of voxels is identified. It is determined whether the candidate is a true pulmonary embolism or a false positive based on a spatial distribution of intensity values for the voxels of the cluster of voxels. The pulmonary embolism candidate is presented to a user when the candidate is determined to be a true pulmonary embolism.

Abstract: A system for automatically detecting pulmonary emboli from medical image data includes receiving image data, automatically detecting one or more pulmonary embolism candidates from the image data, segmenting an airway tract from the image data, segmenting an artery structure from the image data, calculating a distance between each of the candidates and a nearest portion of the segmented airway, determining whether each of the candidates is within or outside of the segmented artery structure, rejecting candidates based on the calculated distance between each of the candidates and the nearest portion of the segmented airway and the determination as to whether each of the candidates is within or outside of the segmented artery structure, and indicating the location of the non-rejected candidates within the image data.

Abstract: A computer-implemented method for determining a boundary for binary classification includes providing a data set, initializing a value for noise in the data set, and determining a hyperplane dividing the data set and a slack variable given a current value for noise. The method further includes updating the value for noise and the slack variable given the hyperplane, and determining the hyperplane to be the boundary for binary classification of the data set upon determining a termination criterion to be met, wherein elements of the data set are classified according to the boundary.

Abstract: A method for training classifiers for ICD-9 patient codes includes providing a set of documents regarding patient hospital visits, combining the documents for each patient visit to create a hospital visit profile, defining a feature as an ngram with a frequency of occurrence greater or equal to a predetermined value that does not appear in a standard list of ngrams, processing the profiles to remove redundancy at a paragraph level and perform tokenization and sentence splitting, performing feature selection, randomly dividing the documents into training, validation, and test sets, and training a set of binary classifiers using a weighted ridge regression, each binary classifier targeting a single ICD-9 code using the training set, wherein each classifier is adapted to determining a specific ICD-9 code by analyzing a patient's hospital records.

Abstract: A method for detecting pulmonary embolisms in computed tomographic (CT) angiography includes providing a digitized (CT) image acquired from inside a pulmonary vessel, the image comprising a plurality of intensities corresponding to a 3-dimensional grid of voxels, for each voxel in the image, extracting a first pulmonary embolism (PE) candidate and PE boundary from the voxel, and for each voxel in the PE boundary, selecting a voxel from the PE boundary, extracting a subsequent PE candidate and PE boundary from the voxel, merging the subsequent PE candidate with the first PE candidate, and merging the subsequent PE boundary with the first PE boundary.

Abstract: A method for training classifiers for Computer-Aided Detection in medical images includes providing an image feature training set {(xi, yi)}i=1l, wherein xi?Rd are input feature variables and yi?{?1,1} are class labels, and a cascade of K classifiers to be trained, minimizing, for each classifier k, a first cost function to initialize an ?k0 associated with each classifier k, fixing all classifiers except classifier k and minimizing a second cost function to solve for ?kc for a counter value c using the training dataset {(xik, yi)}i=1l, calculating a third cost function Jc(?lc, . . . , ?Kc) for each classifier k, and comparing Jc with a previous iteration Jc?1, wherein if Jc?Jc?1 is less than a predetermined tolerance, said classifier training is completed.

Abstract: A method and device having instructions for analyzing input data-space by learning classifiers include choosing a candidate subset from a predetermined training data-set that is used to analyze the input data-space. Candidates are temporarily added from the candidate subset to an expansion set to generate a new kernel space for the input data-space by predetermined repeated evaluations of leave-one-out errors for the candidates added to the expansion set. This is followed by removing the candidates temporarily added to the expansion set after the leave-one-out error evaluations are performed, and selecting the candidates to be permanently added to the expansion set based on the leave-one-out errors of the candidates temporarily added to the expansion set to determine the one or more classifiers.

Abstract: A method for detecting an object within a structure includes performing tobogganing on image data to obtain one or more voxel clusters and to provide a rough indication of the structure. Each of the obtained voxel clusters is characterized as an object candidate and a set of features are determined for each object candidate. Correlations between pairs of the object candidates are measured. Each of the object candidates is classified as either a true object or a non-object based on the set of features and the measured correlations.

Abstract: A method for modeling an image for multiple tasks includes providing an image with n image features, providing an indicator matrix which has m non-zero components corresponding to the m features selected from the n features, constructing a model of the image using the m selected features for each specific labeling task. There are a variable for a specific task to be performed on the image and a variable for a plurality of tasks to be performed on the image.