Kyoto University

Honda Motor Co., Ltd.

Shin Ishii

Shinichi Maeda

Takumi Tanaka

Yuji Yasui

Kosuke Nakanishi

Wataru Sasaki

Daigo Yoshikawa

Provided is a model parameter learning device and the like capable of learning model parameters such that the influence of a noise in input data can be suppressed. A model parameter learning device () alternately carries out first learning processing for learning model parameters W, b, Wand bsuch that an error between data Xand data Xis minimized, and second learning processing for learning model parameters W, b, W, b, Wand bsuch that a loss function Lis minimized.

Model parameter learning device, control device, and model parameter learning method

This invention provides an X-ray CT image processing method that allows more flexible expression by expressing X-ray absorption coefficients probabilistically, makes it possible to acquire reconstructed images that are comparable to those obtained by conventional methods but involve lower X-ray doses, and can reduce beam-hardening artifacts. A probability distribution for the observation of projected X-rays is set and statistical inference is performed. Said probability distribution is expressed in terms of the process of observing a multiple-X-ray sum resulting from multiple projected X-rays being incident upon a detector. Bayesian inference in which the expected value of the posterior distribution is used for statistical inference is performed on the basis of a prior distribution for X-ray absorption coefficients, said prior distribution having parameters for the material and the observation process in terms of which the multiple-X-ray sum is expressed.

X-ray CT image processing method, X-ray CT image processing program, and X-ray CT image device

An x-ray CT image processing method which carries out a statistical estimate in prior knowledge relating to the movement and the x-ray absorption coefficient of a measurement subject comprises: a step of hypothesizing that the measurement subject changes smoothly over time, and defining a first probability model (prior knowledge of the measurement subject over all times) relating to movement and a second probability model (projected image observation model over all times) relating to observation; and a step of carrying out a statistical estimate dependent on both the first probability model and the second probability model. The first probability model relating to the movement of the measurement subject and the second model relating to observation are defined as probability models with which statistical estimates are carried out at the outset, and motion and CT images are simultaneously estimated by carrying out induction based on the probability models.

Motion-tracking X-ray CT image processing method and motion-tracking X-ray CT image processing device

Growing community of inventors

Kyoto, Japan

Shin Ishii