Deep-learning classifier identifies and stages cancer

Researchers from the University of California have devised an automatic classification method for distinguishing between indolent and clinically significant carcinoma using multiparametric (mp) MRI (WO/2017/165801). The hierarchical classification framework utilizes a convolutional neural network that automatically extracts deep features. This avoids deficiencies in current schemes, such as the need to provide handcrafted features predefined by a domain expert or precise delineation of lesion boundaries. The framework is trained using previously acquired mp-MRI data with known cancer classification characteristics. It can then be applied to mp-MRI images of new patients to provide identification and computerized cancer classification results of a suspicious lesion.

Tileable detectors target positron emission mammography

A positron emission mammography (PEM) system with a stationary detector large enough to cover the entire breast with slight compression could collect more signal while using a lower radiotracer dose. With this goal, a team at Thunder Bay Regional Health Research Institute has described tileable block detectors for use in nuclear medicine applications such as PEM (WO/2017/143442). Each block detector includes a scintillator array, a photodetector and a light guide that's optically coupled to the scintillator array and photodetector. The block detectors are four-side tileable, enabling construction of seamless detector arrays. When arrayed as such, the block detectors provide a full-size, seamless detector that achieves full coverage of an object (for example, a gently immobilized breast), improves data collection, and enables high-resolution imaging with a significantly lower radiation dose than with existing PEM systems.

Ultrasound imager needs no physical contact

A system for noncontact ultrasound imaging is described in patent application WO/2017/160858. Developed by MIT researchers, the method can be used to safely generate images of eyes or injured tissue, for example. A photoacoustic excitation source directs light signals with wavelengths of 1400–1600 nm into the patient, generating acoustic disturbances that induce propagating photoacoustic waves. The acoustic disturbances can be translated in defined directions to cause coherent summation of the propagating photoacoustic waves and, thereby, generate an acoustic wave to probe structures within the patient. Vibrations created by the backscatter of the resultant wave are detected at the surface of the patient, and ultrasound images of structures within the patient can be generated without physical contact.

CT scheme minimizes radiation, reduces cycle time per scan

Siemens has invented a CT method that minimizes radiation while simultaneously reducing the cycle time per scan (WO/2017/157491). The operator initially sets an expected approximate scanning area. A first sample scanning line is a scanned in a mid-scanning row within this area, and the tissue texture captured by this line is analysed to see whether it captures the desired region to be examined. The system then searches for the first organ boundary, with the aid of further sample scanning lines, proceeding from the mid-scanning row up to a first limit of the expected scanning area. A further organ boundary is ascertained between the mid-scanning row and a second limit of the expected scanning area. Scanning lines for the actual examination are only captured between the two organ boundaries, eliminating the need for a preliminary scan or human intervention when adjusting the scanning area. At the same time, a more accurate scan is produced, avoiding errors and reducing radiation and scan time.

Rapid filtration methods improve dual-energy CT

Scientists at Rensselaer Polytechnic Institute have published systems and methods for performing dual-energy X-ray CT that can improve spectral separation and decrease motion artefacts without increasing radiation dose (WO/2017/176976). The methods can be used with either a kVp-switching source or a single-kVp source. When used with a kVp-switching source, an absorption grating and filter grating can be placed between the X-ray source and the sample or patient being imaged. Relative motion of the filter and absorption gratings can by synchronized to the kVp switching frequency of the X-ray source. Different filter regions can be exposed to X-rays at various times to produce low- and high-energy X-rays accordingly. When used with a single-kVp source, a combination of absorption and filter gratings can be used and can be driven in an oscillation movement that is optimized for a single-kVp X-ray source. With a single-kVp source, the absorption grating can also be omitted and the filter grating can remain stationary.

Analysing blood vessel geometry sheds light on disease

Bio-Tree Systems has developed methods for extracting blood vessel geometry from optical coherent tomography (OCT) images, for use in analysing biological structures for diagnostic and therapeutic applications relating to diseases that can be detected by vascular changes in the retina (WO/2017/173090). Some embodiments include a method of identifying a region-of-interest in a retina from a number of retinal blood vessels in at least one OCT image of at least part of the retina. Other embodiments include a method of distinguishing between several retinal layers from vessel morphology information of retinal blood vessels in at least one OCT image.