Abstract: A candidate generator generates a set of candidate three-dimensional image patches from an input volume. A candidate classifier classifies the set of candidate three-dimensional image patches as containing or not containing disease. Classifying the set of candidate three-dimensional image patches comprises generating an attention mask for each given candidate three-dimensional image patch within the set of candidate three-dimensional image patches to form a set of attention masks, applying the set of attention masks to the set of candidate three-dimensional image patches to form a set of masked image patches, and classifying the set of masked image patches as containing or not containing the disease. The candidate classifier applies soft attention and hard attention to the three-dimensional image patches such that distinctive image regions are highlighted proportionally to their contribution to classification while completely removing image regions that may cause confusion.

Abstract: Methods and systems for training computer-aided condition detection systems. One method includes receiving a plurality of images for a plurality of patients, some of the images including an annotation associated with a condition; iteratively applying a first deep learning network to each of the images to produce an attention map, a feature map, and an image-level probability of the condition for each of the images; iteratively applying a second deep learning network to each feature map produced by the first network to produce a plurality of outputs; training the first network based on the attention map produced for each image; and training the second network based on the output produced for each of the patients. The second network includes a plurality of convolution layers and a plurality of convolutional long short-term memory (LSTM) layers. Each of the outputs includes a patient-level probability of the condition for one of the patients.

Abstract: Methods and systems for training computer-aided condition detection systems. One method includes receiving a plurality of images for a plurality of patients, some of the images including an annotation associated with a condition; iteratively applying a first deep learning network to each of the images to produce a segmentation map, a feature map, and an image-level probability of the condition for each of the images; iteratively applying a second deep learning network to each feature map produced by the first network to produce a plurality of outputs; training the first network based on the segmentation map produced for each image; and training the second network based on the output produced for each of the patients. The second network includes a plurality of convolution layers and a plurality of convolutional long short-term memory (LSTM) layers. Each of the outputs includes a patient-level probability of the condition for one of the patients.

Abstract: Methods and systems for training computer-aided condition detection systems. One method includes receiving a plurality of images for a plurality of patients, some of the images including an annotation associated with a condition; iteratively applying a first deep learning network to each of the images to produce a segmentation map, a feature map, and an image-level probability of the condition for each of the images; iteratively applying a second deep learning network to each feature map produced by the first network to produce a plurality of outputs; training the first network based on the segmentation map produced for each image; and training the second network based on the output produced for each of the patients. The second network includes a plurality of convolution layers and a plurality of convolutional long short-term memory (LSTM) layers. Each of the outputs includes a patient-level probability of the condition for one of the patients.

Abstract: Methods and systems for training computer-aided condition detection systems. One method includes receiving a plurality of images for a plurality of patients, some of the images including an annotation associated with a condition; iteratively applying a first deep learning network to each of the images to produce an attention map, a feature map, and an image-level probability of the condition for each of the images; iteratively applying a second deep learning network to each feature map produced by the first network to produce a plurality of outputs; training the first network based on the attention map produced for each image; and training the second network based on the output produced for each of the patients. The second network includes a plurality of convolution layers and a plurality of convolutional long short-term memory (LSTM) layers. Each of the outputs includes a patient-level probability of the condition for one of the patients.

Abstract: A candidate generator generates a set of candidate three-dimensional image patches from an input volume. A candidate classifier classifies the set of candidate three-dimensional image patches as containing or not containing disease. Classifying the set of candidate three-dimensional image patches comprises generating an attention mask for each given candidate three-dimensional image patch within the set of candidate three-dimensional image patches to form a set of attention masks, applying the set of attention masks to the set of candidate three-dimensional image patches to form a set of masked image patches, and classifying the set of masked image patches as containing or not containing the disease. The candidate classifier applies soft attention and hard attention to the three-dimensional image patches such that distinctive image regions are highlighted proportionally to their contribution to classification while completely removing image regions that may cause confusion.

Abstract: Systems and methods for developing a disease detection model. One method includes training the model using an image study and an associated disease label mined from a radiology report. The image study including a sequence of a plurality of two-dimensional slices of a three-dimensional image volume, and the model including a convolutional neural network layer and a convolutional long short-term memory layer. Training the model includes individually extracting a set of features from each of the plurality of two-dimensional slices using the convolutional neural network layer, sequentially processing the features extracted by the convolutional neural network layer for each of the plurality of two-dimensional slices using the convolutional long short-term memory layer, processing output from the convolutional long short-term memory layer for each of the plurality of two-dimensional slices to generate a probability of the disease, and updating the model based on comparing the probability to the label.

Abstract: Systems and methods for developing a disease detection model. One method includes training the model using an image study and an associated disease label mined from a radiology report. The image study including a sequence of a plurality of two-dimensional slices of a three-dimensional image volume, and the model including a convolutional neural network layer and a convolutional long short-term memory layer.

Abstract: Systems and methods for developing a disease detection model. One method includes training the model using an image study and an associated disease label mined from a radiology report. The image study including a sequence of a plurality of two-dimensional slices of a three-dimensional image volume, and the model including a convolutional neural network layer and a convolutional long short-term memory layer. Training the model includes individually extracting a set of features from each of the plurality of two-dimensional slices using the convolutional neural network layer, sequentially processing the features extracted by the convolutional neural network layer for each of the plurality of two-dimensional slices using the convolutional long short-term memory layer, processing output from the convolutional long short-term memory layer for each of the plurality of two-dimensional slices to generate a probability of the disease, and updating the model based on comparing the probability to the label.

Abstract: Systems and methods for developing a disease detection model. One method includes training the model using an image study and an associated disease label mined from a radiology report. The image study including a sequence of a plurality of two-dimensional slices of a three-dimensional image volume, and the model including a convolutional neural network layer and a convolutional long short-term memory layer.

Abstract: A method for recognizing heart diseases in a cardiac echo video of a heart with an unknown disease using a spatio-temporal disease model derived from a training echo video, comprising the steps of: generating a plurality of training models for heart diseases, wherein the cardiac echo videos are each derived from a known viewpoint and the disease of the heart is known; analyzing the video of the heart with the unknown disease by fitting a model of shape and motion for each frame and combining the results across the frames; and, reporting the disease using a classification method for choosing among the diseases of interest.

Abstract: According to one embodiment of the present invention, a method for echocardiogram view classification is provided. According to one embodiment of the present invention, a method comprises: obtaining a plurality of video images of a subject; aligning the plurality images; using the aligned images to generate a motion magnitude image; filtering the motion magnitude image using an edge map on image intensity; detecting features on the motion magnitude image, retaining only those features which lie in the neighborhood of intensity edges; encoding the remaining features by generating, x, y image coordinates, a motion magnitude histogram in a window around the feature point, and a histogram of intensity values near the feature point; and using the encoded features to classify the video images of the subject into a predetermined classification.

Abstract: Systems and methods providing automated extraction of information contained in video data and uses thereof are described. In particular, systems and associated methods are described that provide techniques for extracting data embedded in video, for example measurement-value pairs of medical videos, for use in a variety of applications, for example video indexing, searching and decision support applications.

Abstract: Continuous wave Doppler images in a data base comprising cardiac echo studies are processed to separate Doppler frames. The frames are pre-processed to extract envelope curves and their corner shape features. Shape patterns in Doppler images from echo studies of patients with known cardiac (valvular) diseases are employed to infer the similarity in valvular disease labels for purposes of automated clinical decision support. Specifically, similarity in appearance of Doppler images from the same disease class is modeled as a constrained non-rigid translation transform of velocity envelopes embedded in these images. Shape similarity between two Doppler images is then judged by recovering the alignment transform using a variant of dynamic shape warping. Since different diseases appear as characteristic shape patterns in Doppler images, measuring the similarity in the shape pattern conveyed within the velocity region of two Doppler images can infer the similarity in their diagnosis labels.

Abstract: A method for recognizing heart diseases in a cardiac echo video of a heart with an unknown disease using a spatio-temporal disease model derived from a training echo video, comprising the steps of: generating a plurality of training models for heart diseases, wherein the cardiac echo videos are each derived from a known viewpoint and the disease of the heart is known; analyzing the video of the heart with the unknown disease by fitting a model of shape and motion for each frame and combining the results across the frames; and, reporting the disease using a classification method for choosing among the diseases of interest.

Abstract: Systems and methods providing automated extraction of information contained in video data and uses thereof are described. In particular, systems and associated methods are described that provide techniques for extracting data embedded in video, for example measurement-value pairs of medical videos, for use in a variety of applications, for example video indexing, searching and decision support applications.

Abstract: Systems and methods for removing or suppressing information in images and video frames is described herein. In particular, systems and methods provide for removing information from images capable of identifying individuals related to the image. For example, embodiments provide for the removal of protected health information (PHI) from source images, including medical images, video frames, and documents converted to images or video frames. In addition, embodiments operate on actual images and video frames as opposed to data extracted from such sources. In particular, embodiments provide for the creation of a PHI filter for an individual of interest comprised of identifying information. Images are filtered using the PHI filter and information potentially identifying the individual of interest is located and removed from the image.

Abstract: A method for recognizing heart diseases in a cardiac echo video of a heart with an unknown disease using a spatio-temporal disease model derived from a training echo video, comprising the steps of: generating a plurality of training models for heart diseases, wherein the cardiac echo videos are each derived from a known viewpoint and the disease of the heart is known; analyzing the video of the heart with the unknown disease by fitting a model of shape and motion for each frame and combining the results across the frames; and, reporting the disease using a classification method for choosing among the diseases of interest.

Abstract: Systems and methods providing automated extraction of information contained in video data and uses thereof are described. In particular, systems and associated methods are described that provide techniques for extracting data embedded in video, for example measurement-value pairs of medical videos, for use in a variety of applications, for example video indexing, searching and decision support applications.

Abstract: According to one embodiment of the present invention, a method for echocardiogram view classification is provided. According to one embodiment of the present invention, a method comprises: obtaining a plurality of video images of a subject; aligning the plurality images; using the aligned images to generate a motion magnitude image; filtering the motion magnitude image using an edge map on image intensity; detecting features on the motion magnitude image, retaining only those features which lie in the neighborhood of intensity edges; encoding the remaining features by generating, x, y image coordinates, a motion magnitude histogram in a window around the feature point, and a histogram of intensity values near the feature point; and using the encoded features to classify the video images of the subject into a predetermined classification.