Abstract: Techniques for processing one or more frames of an angiogram are disclosed. The processing may take place during or after an angiography exam. The one or more frames of the angiogram are acquired during the angiography exam. The one or more frames are processed to determine, based on at least one pre-defined criterion, whether the angiogram at least comprises one frame with a diagnostic value among the one or more frames. If the angiogram comprises at least one frame with the diagnostic value, based on the angiogram, a score quantifying the diagnostic value of the angiogram is determined using a trained machine-learning (ML) algorithm. Techniques for processing, e.g., ranking/sorting, multiple angiograms associated with an anatomical region of interest of a patient are also provided, by which a respective score for each of the multiple angiograms is determined using the techniques for processing one or more frames of an angiogram.

Abstract: A soft tissue emulation system, comprising: an input interface, configured to obtain imaging data of the soft tissue; a computing unit, configured to implement an artificial neural network, which is adapted to generate, using the obtained imaging data as input, and a biophysical model of the soft tissue, a digital twin of the soft tissue at different times, wherein the biophysical model describes the response of the soft tissue to at least one of thermal stimuli or electromechanical stimuli over time, and wherein the generation of the digital twin at one time is independent of the generation of the digital twin at another time; and an output interface, configured to output a representation of the soft tissue over time based on the digital twin generated by the artificial neural network.

Abstract: Systems and methods are provided for training an artificial intelligence model for detecting calcified portions of a vessel in an input medical image. One or more first medical images of a vessel in a first modality and one or more second medical image of the vessel in a second modality are received. Calcified portions of the vessel are detected in the one or more first medical images, The artificial intelligence model is trained for detecting calcified portions of the vessel in the input medical image in the second modality based on the one or more second medical images and the detected calcified portions of the vessel detected in the one or more first medical images.

Abstract: For shape determination of cardiac anatomy with a medical imager, irregularities in motion, poor image quality, and misalignment of imaging planes are counteracted by a process relying on alignment of contours in combination with selection and fitting of a motion model. Contours are extracted from 2D images and aligned for each frame, which is extracted from the sequence of 2D images. The alignment may use a translation for each frame and rotation across frames for improved performance. A motion model is fit to the aligned contours and tested. If insufficient (greater than threshold difference), other motion models are aligned and tested. Motion models may be created on demand for improved performance. If sufficient, the shape of the heart structure is determined from the fit model.

Abstract: Angiography angles are determined. Patient information and target vessel information are obtained, wherein the patient information defines individual medical information of a patient and wherein the target vessel information defines at least one target vessel to be imaged. At least one angiography angle is determined based on the patient information and the target vessel information. Angiograms obtained using the at least one angiography angle are analyzed to determine a vessel coverage of the target vessel and based on the vessel coverage determines additional angiography angles.

Abstract: Systems and methods for simulating radiation effect for cardiac ablation can include a processor generating a heart simulation model configured to simulate electrical activities of the heart of the patient. The processor can determine a simulated post-radiation state of the heart of the patient by adjusting the heart simulation model to account for an effect of a radiation treatment plan on simulated electrical activities of the heart of the patient. The processor can simulate, using the adjusted heart simulation model, the simulated post-radiation state of the heart with a stimulation to induce a heart rhythm disorder, determine whether the heart rhythm disorder is induced based on electrical activities generated when simulating the simulated post-radiation state of the heart with the stimulation, and output an indication of whether the heart rhythm disorder is induced. The processor can suggest modifications to the radiation plan if the heart rhythm disorder is induced.

Abstract: Systems and methods for predicting a location of a target heart region for cardiac ablation can include one or more processors generating a 3D model of a heart of a patient based on medical images of the patient, and estimating, using the 3D model of the heart and electrophysiology data of the patient, one or more mechanical properties that drive motion of the patient's heart. The one or more processors can generate, using the 3D model and the one or more mechanical properties, a simulated motion pattern of the patient's heart over at least a portion of a cardiac cycle, and identify a region of interest (ROI) of the heart of the patient to be radiated. The one or more processors can determine, using the simulated motion pattern of the heart of the patient, a location of the ROI at a predefined time instance within the cardiac cycle.

Abstract: Machine-based risk prediction or assistance is provided for peri-procedural complication, such as peri-procedural myocardial infarction (PMI). A machine-learned model is used to predict risk of PMI and/or recommend courses of action to avoid PMI in PCI. Various combinations of types or modes of information are used in the prediction, such as both imaging and non-imaging data. The prediction may be made prior to, during, and/or after PCI using the machine-learned model to more quickly reduce the chance of PMI. The workflows for prior, during, and/or post PCI incorporate the risk prediction and/or risk-based recommendations to reduce PMI for patients.

Abstract: Systems and methods for determining a semantic image understanding of medical imaging studies are provided. A plurality of medical imaging studies associated with a plurality of medical imaging modalities is provided. Metadata associated with each of the plurality of medical imaging studies is generated by performing a plurality of semantic image analysis tasks using one or more machine learning based networks. The metadata associated with each of the plurality of medical imaging studies is output.

Abstract: One or more example embodiments of the present invention relates to a method for the automated determination of examination results in an image sequence from multiple chronologically consecutive frames, the method comprising determining diagnostic candidates in the form of contiguous image regions in the individual frames for a predefined diagnostic finding; and for a number of the diagnostic candidates, determining which candidate image regions in other frames correspond to the particular diagnostic candidate, determining whether the candidate image regions of the particular diagnostic candidate in the other frames overlap with other diagnostic candidates, generating a graph containing the determined diagnostic candidates of the frames as nodes and the determined overlaps as edges, and generating communities from nodes connected via edges.

Abstract: A method and system for deep learning based cardiac electrophysiological model personalization is disclosed. Electrophysiological measurements of a patient, such as an ECG trace, are received. A computational cardiac electrophysiology model is personalized by calculating patient-specific values for a parameter of the computational cardiac electrophysiology model based at least on the electrophysiological measurements of the patient using a trained deep neural network (DNN). The parameter of the computational cardiac electrophysiology model corresponds to a spatially varying electrical cardiac tissue property.

Abstract: Systems and methods for determining input data is out-of-domain of an AI (artificial intelligence) based system are provided. Input data for inputting into an AI based system is received. An in-domain feature space of the AI based system and an out-of-domain feature space of the AI based system are modelled. The in-domain feature space corresponds to features of data that the AI based system is trained to classify. The out-of-domain feature space corresponds to features of data that the AI based system is not trained to classify. Probability distribution functions in the in-domain feature space and the out-of-domain feature space are generated for the input data and for the data that the AI based system is trained to classify. It is determined whether the input data is out-of-domain of the AI based system based on the probability distribution functions for the input data and for the data that the AI based system is trained to classify.

Abstract: A computer-implemented method for executing patient management workflows includes acquiring a pre-test dataset of clinically relevant information related to a patient and using a first intelligent agent to identify a diagnostic test for the patient based on the pre-test dataset. Following performance of the diagnostic test, a second intelligent agent is used to select a processing technique to be applied to data collected from the diagnostic test to obtain a diagnostic marker. Following application of the processing technique to the data collected from the diagnostic test, a third intelligent agent is used to generate an optimal patient management plan based on the pre-test dataset, the data collected from the diagnostic test, and the diagnostic marker.

Abstract: A method and system for non-invasive assessment and therapy planning for coronary artery disease from medical image data of a patient is disclosed. Geometric features representing at least a portion of a coronary artery tree of the patient are extracted from medical image data. Lesions are detected in coronary artery tree of the patient and a hemodynamic quantity of interest is computed at a plurality of points along the coronary artery tree including multiple points within the lesions based on the extracted geometric features using a machine learning model, resulting in an estimated pullback curve for the hemodynamic quantity of interest.

Abstract: Systems and methods for generating an ablation map identifying target ablation locations on a heart of a patient are provided. One or more input medical images of a heart of a patient and a voltage map of the heart of the patient are received. An ablation map identifying target ablation locations on the heart is generated using one or more trained machine learning based models based on the one or more input medical images and the voltage map. The ablation map is output.

Abstract: Systems and methods for determining one or more measures of interest for optimizing throughput of a catheterization laboratory are provided. A priori medical procedure data relating to a medical procedure to be performed on a patient in a catheterization laboratory is received. One or more measures of interest are predicted based on the received a priori medical procedure data using a trained machine learning model. The one or more measures of interest include an overall time for performing the medical procedure on the patient in the catheterization laboratory. The one or more predicted measures of interest are output.

Abstract: Heart strain determination includes receiving a series of 2D-slice images as input. A pose estimation module estimates a slicing-pose of the inputted series of 2D-slice images in the heart. A 3D deformation estimation module estimates a 3D deformation field from the series of 2D-slice images and the estimated slicing-pose. A strain measurement module computes a heart strain measure from the 3D deformation field and a predefined definition for strain computation.

Abstract: For segmentation in medical imaging, a shape generative adversarial network (shape GAN) is used in training. By including shape information in a lower dimensional space than the pixels or voxels of the image space, the network may be trained with a shape loss or optimization. The adversarial loss and the shape loss are used to train the network, so the resulting generator may segment complex shapes in 2D or 3D. Other optimization may be used, such as using a loss in image space.

Abstract: A medical knowledge base in a digital, clinical system is upgraded. A storage with a knowledge base, being a SNOMED knowledge base, is provided in a web ontology format. Procedural data, representing clinical procedures for evaluation of a patient's health state, is received. The received procedural data is mapped in a set of SNOMED expressions. The SNOMED expressions are converted into statements in the web ontology format. The SNOMED knowledge base is upgraded with the received procedural data by adding the statements in the SNOMED knowledge base for providing a processable file with an upgraded version of the SNOMED knowledge base.

Abstract: For non-invasive EP mapping, a sparse number of electrodes (e.g., 10 in a typical 12-lead ECG exam setting) are used to generate an EP map without requiring preoperative 3D image data (e.g. MR or CT). An imager (e.g., a depth camera) captures the surface of the patient and may be used to localize electrodes in any positioning on the patient. Two-dimensional (2D) x-rays, which are commonly available, and the surface of the patient are used to segment the heart of the patient. The EP map is then generated from the surface, heart segmentation, and measurements from the electrodes.