Fujifilm Corporation

Deepak Keshwani

Yoshiro Kitamura

A first learning unit performs learning of a first neural network that extracts a feature vector in each pixel of a target image including a plurality of objects and that outputs a feature map in which feature vectors of pixels belonging to individual objects included in the target image are clustered and distributed as a plurality of the feature vector groups in a feature space which is a space of the feature vector. A second learning unit performs learning of a second neural network that outputs a class classification result of a plurality of objects belonging to the same category included in the target image in response to input of the feature vector of the target image.

Learning apparatus, learning method, and learning program, class classification apparatus, class classification method, and class classification program, and learned model

A learning unit derives, from a target image including at least one tubular structure, in a case where an image for learning and ground-truth data of a graph structure included in the image for learning are input to an extraction model which extracts a feature vector of a plurality of nodes constituting a graph structure of the tubular structure, a loss between nodes on the graph structure included in the image for learning on the basis of an error between a feature vector distance between nodes belonging to the same graph structure and a topological distance which is a distance on a route of the graph structure between the nodes, and performs learning of the extraction model on the basis of the loss.

Learning apparatus, learning method, and learning program, graph structure extraction apparatus, graph structure extraction method, and graph structure extraction program, and learned extraction model

Provided is a machine learning device and method that enables machine learning of labeling, in which a plurality of labels are attached to volume data at one effort with excellent accuracy, using training data having label attachment mixed therein. A probability calculation unit () calculates a value (soft label) indicating a likelihood of labeling of a class Ci for each voxel of a second slice image by means of a learned teacher model (). A detection unit () detects 'bronchus' and “blood vessel” for the voxels of the second slice image using a known method, such as a region expansion method and performs labeling of “bronchus” and “blood vessel”. A correction probability setting unit () replaces the soft label with a hard label of “bronchus” or “blood vessel” detected by the detection unit (). A distillation unit () performs distillation of a student model () from the teacher model () using the soft label after correction by means of the correction probability setting unit (). With this, the learned student model () is obtained.

Machine learning device and method

Provided are an image processing apparatus, an image processing method, and a program that can reduce the time and effort required to correct the segmentation of a medical image. An image processing apparatus includes: an image acquisition unit () that acquires a medical image (); a segmentation unit () that performs segmentation on the medical image acquired by the image acquisition unit and classifies the medical image into prescribed classes for each local region; a global feature acquisition unit () that acquires a global feature indicating an overall feature of the medical image; and a correction unit () that corrects a class of a correction target region that is a local region whose class is to be corrected in the medical image according to the global feature with reference to a relationship between the global feature and the class.

Image processing apparatus, image processing method, and program for segmentation correction of medical image

Provided are an image processing apparatus, an image processing method, and a program that can suppress an error in the segmentation of a medical image. An image processing apparatus includes: a segmentation unit () that applies deep learning to perform segmentation which classifies a medical image () into a specific class on the basis of a local feature of the medical image; and a global feature classification unit () that applies deep learning to classify the medical image into a global feature which is an overall feature of the medical image. The segmentation unit shares a weight of a first low-order layer which is a low-order layer with a second low-order layer which is a low-order layer in the global feature classification unit.

Image processing apparatus, image processing method, and program

Provided are a machine learning device and a method capable of performing machine learning of labeling for accurately attaching a plurality of labels to volume data at once by using learning data with mixed inconsistent labeling. A neural network () receives an input of multi-slice images of learning data Di (i=1, 2, . . . n) of which a class to be labeled is n types, and creates a prediction mask of n anatomical structures i by a convolutional neural network (CNN) or the like (S). A machine learning unit () calculates a prediction accuracy acc(i) of the class corresponding to the learning data Di for each learning data Di (S). The machine learning unit () calculates a weighted average M of an error di between the prediction accuracy acc(i) and a ground truth mask Gi. The machine learning unit () calculates a learning loss by a loss function Loss (S). The machine learning unit () changes each coupling load of the neural network () from an output layer side to an input layer side according to a value of the learning loss calculated by the loss function Loss (S).

There are provided machine learning device and method which can prepare divided data suitable for machine learning from volume data for learning. A machine learning unit () calculates detection accuracy of each organ O(j,i) in a predicted mask Pj using a loss function Loss. However, the detection accuracy of the organ O(k,i) with a volume ratio A(k,i)&#x3c;Th is not calculated. That is, in the predicted mask Pk, the detection accuracy of the organ O(k,i) with a volume ratio that is small to some extent is ignored. The machine learning unit () changes each connection load of a neural network () from an output layer side to an input layer side according to the loss function Loss.

Growing community of inventors

Tokyo, Japan

Deepak Keshwani