PIL technology
We are developing medical image processing techniques for various purposes such as high resolution, noise reduction, and contrast improvement. We are developing not only conventional filtering methods but also methods using machine learning.
Image synthesis
We are developing a processing method to change image style to different medical imaging modalities. For example, an MRI image can be converted to a CT image, a DRR image to an X-ray image, and an X-ray image to a DRR image.
We are developing nonlinear registration (deformable image registration) using 3D images. The speedup is achieved by using GPU for the speedup.
We are developing 2D-3D rigid image registration, which is used in radiotherapy to align X-ray images with treatment planning CT images. This technology is used in radiotherapy for patient positioning to align X-ray images with CT images for treatment planning, and is accelerated by using GPUs.
Recently, 4DCT is increasingly used in the radiotherapy field, but it is sometimes difficult to obtain good quality 4DCT images due to irregular breathing patterns. To solve this problem, we have developed a technology to generate a 4DCT image from a 3DCT image.
4DCT artifact reduction
In 4DCT imaging, it is sometimes difficult to acquire 4DCT images with precise geometric shapes due to irregular breathing patterns (4DCT artifact). We have developed a method to improve this artifact by using deep learning.
Markerless tumor tracking
In radiotherapy of the torso region, respiratory synchronous irradiation is used, in which the inside of the body is monitored while X-ray fluoroscopic images are taken during treatment, and the treatment beam is irradiated when the tumor reaches the position determined in the treatment plan. There is a method to insert a metal marker near the tumor, but we have developed a technology to track the tumor position in the X-ray fluoroscopic image in real time without the need to insert a metal marker.
We are developing CT/CBCT reconstruction algorithms and artifact reduction methods for raw databases based on our experience in the joint development of area detector CT with medical equipment manufacturers and cone beam CT (CBCT) with treatment equipment manufacturers. We are also working on the implementation of high-speed computation using GPUs.
Bone suppression filter on X-ray images
The removal of bone structures in X-ray images has been performed by the dual energy subraction method, which uses X-rays of different energy (tube voltage), and by image processing. However, it is difficult to achieve real-time processing in these methods. Therefore, we developed a method to achieve bone removal by placing a filter on the X-ray detector and performing X-ray imaging.
DICOM files stored in PACS can be searched and retrieved based on the search key.
Create DRR (digitally reconstructed radiographs) images from CT images at high speed.