Wearable monitor keeps an eye on cardiac function
An urgent need exists for continuous, non-invasive monitoring of heart function in patients at risk of congestive heart failure. An important diagnostic indicator of congestive heart failure is reduction in the left ventricle ejection fraction (the amount of blood pumped out by the left ventricle in each heartbeat). As such, a team from the University of California has invented an integrated cardiorespiratory (ICR) system for continuous ejection fraction measurement using a wearable device containing a number of acoustic sensors (WO/2019/045996). The ICR system performs signal processing to characterize acoustic signals generated by cardiac haemodynamic flow, cardiac valve and tissue motion, and uses advanced machine learning methods to provide accurate computation of ejection fraction.
Deep learning network identifies, removes image artefacts
Philips has published details of an ultrasound system that uses deep learning networks to enhance ultrasound images by identifying artefacts for removal (WO/2019/034436). By analysing orthogonal information — in this implementation, the structural information of a B mode image and motion information of the same field-of-view as that image — the system eliminates haze artefacts in B mode images of the carotid artery. In another embodiment, the neural network reduces haze artefacts by reducing time-gain control at the depth of artefacts.
Hybrid NMR–OCT device improves analysis of excised tissue
Clear-Cut Medical has devised a hybrid system that combines nuclear magnetic resonance (NMR) and optical coherence tomography (OCT) to enable real-time imaging and analysis of surgically excised tissue (WO/2019/030620). Imaging using NMR creates a pixel map of the tissue surface, with each pixel colour-coded according to its probability of containing malignant tissue. Based on these probabilities, the system uses OCT to extract microscopic images from locations in the suspicious pixels. It then analyses these images to make a diagnosis based on the cellular microstructure of the tissue.
Software generates 3D brain map from MRI data
NEUROPHET has developed a method and program for generating a 3D brain map based on MR imaging (WO/2019/050226). The approach involves acquiring a brain MRI of a subject, segmenting this image into multiple regions, and using the segmented image to generate a 3D brain image. Based on the properties of each region in the 3D image, the method generates an individual 3D brain map of the subject. This map can then be used in simulations of electrical stimulation of the subject’s brain, for example, to help electrode positioning for therapeutic applications. The segmentation process includes a step in which the subject’s brain MRI is input into a deep-learning model that has been trained using a number of processed brain MRI images.
Data processing enhances shear wave elastography
Researchers at the Mayo Foundation for Medical Education and Research have described processing methods for data acquired using probe oscillation shear wave ultrasound elastography (WO/2019/032803). This elastography method uses continuous vibration of the ultrasound probe to generate shear waves in the tissue and pulse-echo ultrasound detection to track the resulting shear waves. The described approach can effectively separate the shear wave signals from signals corresponding to residual motion artefacts from the transducer vibration. The filing also details systems and methods for real-time visualization of shear waves propagating in the subject.
CT fluoroscopy promises ultralow dose
CT fluoroscopy is an invaluable tool for use in CT-guided interventions. LiteRay Medical of Madison, WI, has described a system for ultralow-dose CT fluoroscopy (WO/2019/055288). The method involves acquiring pairs of projections of an interventional device using CT fluoroscopy, by rotating the gantry of a CT scanner. Each pair of projections is obtained at a predetermined angular separation, which is greater than the angular separation used for a full-dose CT scan of the target object. The full-dose scan acquires at least twice the number of projections per gantry rotation than used for projection pairs of the interventional device. The position of the interventional device is identified in real time for each pair of projections, using back-projection of images of the device from the respective projections pair. The system can also superimpose a 2D or 3D image of the interventional device on a CT image of an anatomical region at the identified device position.
