Shimadzu Corporation

Shota Oshikawa

Wataru Takahashi

Ayako Akazawa

Zhengyan Gao

Kazue Mizuno

Yohei Sugawara

Yuichiro Hirano

In a creation method of a trained model, a reconstructed image () obtained by reconstructing three-dimensional X-ray image data () is generated. A projection image () is generated from a three-dimensional model of an image element () by a simulation. The projection image is superimposed on the reconstructed image to generate a superimposed image (). A trained model () is created by performing machine learning using the superimposed image, and the reconstructed image or the projection image.

Creation method of trained model, image generation method, and image processing device

This image generation device is provided with: a learning image generation unit () for generating a training input image and a training output image based on three-dimensional data; a noise addition unit () for adding the same noise to the training input image and the training output image; a learning unit () for learning a learning model for extracting or removing a specific portion by performing machine learning based on the training input image to which the noise has been added and the training output image to which the noise has been added; and an image generation unit () for generating an image from which the specific portion has been extracted or removed by using a learned learning model.

Image generation device, image generation method, and learned model generation method

An image generating device for generating an image which is an X-ray image of an area including a bone portion of a subject with the bone portion removed has a control unitincluding: a DRR imagerconfigured to generate a first DRR image of an area including a bone portion and a second DRR image showing the bone portion, by performing, for a set of CT image data of an area including the bone portion of a subject, a virtual fluoroscopic projection simulating a geometric fluoroscopy condition of an X-ray irradiator and an X-ray detector for the subject; a training sectionconfigured to generate a machine learning model for recognizing the bone portion, by performing machine learning using the first DRR image and the second DRR image serving as a training image; an image converterconfigured to perform conversion of the X-ray image of the area including the bone portion of the subject, using the machine learning model trained in the training section, to generate an image showing the bone portion; and a bone portion subtractorconfigured to subtract the image showing the bone portion from the X-ray image of the area including the bone portion of the subject.

Image creation device

A training label image correction method includes performing a segmentation process on an input image () of training data () by a trained model () using the training data to create a determination label image (), comparing labels of corresponding portions in the determination label image () and a training label image () with each other, and correcting label areas () included in the training label image based on label comparison results.

Training label image correction method, trained model creation method, and image analysis device

A control partincludes: a DRR image creation partthat creates a DRR image including a specific site; a specific site projection partthat creates a projection region image representing the region of the specific site; a discriminator learning partthat learns a discriminator for recognizing the region of the specific site by performing machine learning with use of the DRR image and the projection region image as a training label image; a specific site region detection partthat detects the region of the specific site by performing discrimination using the discriminator learned by the discriminator learning parton an X-ray fluoroscopic image including the specific site; and a radiation signal generation partthat transmits a treatment beam radiation signal to an irradiation devicewhen the region of the specific site detected by the specific site region detection partis included in the radiation region of a treatment beam.

Tracking device for radiation treatment, position detection device, and method for tracking moving body

A full-size training image is reduced by an image reduction unit () and input to an FCN (Fully Convolutional Neural Network) computation unit (), and the FCN computation unit () performs calculation under a set filter coefficient and outputs a reduced label image. The reduced label image is enlarged to a full size by an image enlargement unit (), the error calculation unit () calculates an error between the enlarged label image and a full-size ground truth based on a loss function, and the parameter update unit () updates a filter coefficient depending on the error. By repeating learning under the control of the learning control unit, it is possible to generate a learning model for performing optimal segmentation including an error generated at the time of image enlargement. Further, by including the image enlargement processing in the learning model, a full-size label image can be output, and the accuracy evaluation of the model can also be performed with high accuracy.

Growing community of inventors

Kyoto, Japan

Shota Oshikawa