# US 12,004,860 B2 · Cardiac function assessment using machine-learning algorithms > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,004,860 B2 - Document ID: US12004860B2 - Status: Granted - Kind: B2 - Application number: 17/305,391 - Filed: 2021-07-07 - Published: 2022-02-03 - Granted: 2024-06-11 - Earliest claimed: 2020-07-31 - Assignees: Siemens Healthineers AG - Applicants: Siemens Healthineers AG - Inventors: Paul Klein; Ingo Schmuecking; Costin Florian Ciusdel; Lucian Mihai Itu; Tiziano Passerini; Puneet Sharma - Web page: https://patentgrep.com/patent/US12004860B2 ## CPC - A61B 5/308: for electrocardiography [ECG] - A61B 5/7264: {Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems} - A61B 5/02028: {Determining haemodynamic parameters not otherwise provided for, e.g. cardiac contractility or left ventricular ejection fraction} - A61B 5/026: Measuring blood flow {(A61B3/1233, A61B3/1241 take precedence)} - A61B 5/7267: {involving training the classification device} - G16H 30/40: for processing medical images, e.g. editing - G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems - G16H 50/30: for calculating health indices; for individual health risk assessment - G16H 50/50: for simulation or modelling of medical disorders - A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D) - G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA ## Abstract A method includes processing at least one input dataset (using a multi-level processing algorithm, one or more of the at least one input dataset comprising imaging data of an echocardiography of a cardiovascular system of a patient. The multi-level processing algorithm comprises a multi-task level and a consolidation-task level. An input of the consolidation-task level is coupled to an output of the multi-task level. The multi-task level is configured to determine multiple diagnostic metrics of the cardiovascular system based on the at least one input dataset. The consolidation-task level is configured to determine a fitness of the cardiovascular system of the patient. ## Claims ### Claim 1 A method, comprising: processing at least one input dataset using a multi-level machine learning processing algorithm, one or more of the at least one input dataset comprising imaging data of a cardiovascular system of a patient, wherein the multi-level processing algorithm comprises a multi-task level and a consolidation-task level, wherein an input of the consolidation-task level is coupled to an output of the multi-task level, wherein the multi-task level is configured to determine multiple diagnostic metrics of the cardiovascular system based on the at least one input dataset, wherein the consolidation-task level is configured to determine a fitness of the cardiovascular system of the patient, wherein the multi-level processing algorithm further comprises a pre-processing level, the pre-processing level being configured to determine a quality index for each one of the at least one input dataset, the pre-processing level further configured to filter at least one of the at least one input dataset based on the quality index and/or determine a confidence level of the fitness based on the quality indexes. ### Claim 2 (depends on 1) The method of claim 1, wherein the multiple diagnostic metrics comprise a classification of perspectives of views of the imaging data. ### Claim 3 (depends on 1) The method of claim 1, wherein the multiple diagnostic metrics comprise a classification of cardiac phases the imaging data. ### Claim 4 (depends on 1) The method of claim 1, wherein the multiple diagnostic metrics comprise a classification of anatomic features imaged by the imaging data. ### Claim 5 (depends on 1) The method of claim 1, wherein the multiple diagnostic metrics comprise a classification of properties of a heart chamber of the heart of the cardiovascular system, wherein the properties are from the group comprising: linear measurements; area measurements; volumetric measurements; mass; cardiac deformation; and/or segmentation. ### Claim 6 (depends on 1) The method of claim 1, wherein the multiple diagnostic metrics comprise a classification of a wall motion of the left ventricle of the heart of the cardiovascular system. ### Claim 7 (depends on 1) The method of claim 1, wherein the multiple diagnostic metrics comprise a classification of hemodynamics of a whole-body hemodynamic circulation associated with the cardiovascular system. ### Claim 8 (depends on 1) The method of claim 1, wherein the consolidation-task level is configured to determine the fitness further based on a classification of hemodynamics of a whole-body hemodynamic circulation associated with the cardiovascular system. ### Claim 9 (depends on 1) The method of claim 1, wherein consolidation-task level is configured to determine the fitness of the cardiovascular system of the patient by classifying a wall motion of a left ventricle. ### Claim 10 (depends on 1) The method of claim 1, wherein the at least one input dataset comprises multiple input datasets acquired at multiple stages of a disease of the cardiovascular system, wherein the consolidation-task level is configured to determine a trend of the fitness across the multiple stages of the disease. ### Claim 11 (depends on 1) The method of claim 1, wherein the at least one input dataset comprises a medical report associated with the disease of the cardiovascular system, wherein the multi-level processing algorithm is configured to extract at least one of the multiple diagnostic metrics from the medical report. ### Claim 12 (depends on 1) The method of claim 1, further comprising: selecting between a first configuration of the multi-level processing algorithm and a second configuration of the multi-level processing algorithm depending on at least one of a current stage or a type of a disease associated with the cardiovascular system. ### Claim 13 (depends on 1) The method of claim 1, further comprising: training one or more neural networks of at least one of the multi-task level or the consolidation-task level based on the at least one input dataset and one or more labeling functions, the one or more labeling functions determining a ground truth for the at least one input dataset based on one or more of the following inputs to the one or more labeling functions from the group of: a fractional flow reserve measurement; a clinical decision; a patient outcome; a stenosis grade determined based on a coronary angiography; and a wall motion score index. ### Claim 14 A method, comprising: processing, using at least one machine-learning processing algorithm, multiple input datasets indicative of at least one of measurements on a cardiovascular system of a patient or diagnostic metrics of the cardiovascular system, the multiple input datasets being associated with different stages of a disease of the cardiovascular system, wherein the at least one processing algorithm is configured to determine a trend of a fitness of the cardiovascular system across the multiple stages of the disease of the cardiovascular system, wherein the at least one processing algorithm is configured to select a subset of data from the multiple input datasets based on the trend of the fitness and to provide the subset of data to a human-machine-interface for expert review. ### Claim 15 (depends on 14) The method of claim 14, wherein one or more of the multiple input datasets comprise a medical report associated with the disease of the cardiovascular system, wherein the method further comprises: pre-processing the one or more of the multiple input datasets comprising the medical report to extract associated one or more diagnostic metrics of the disease of cardiovascular system. ### Claim 16 (depends on 14) The method of claim 14, further comprising: determining a quality index for one or more of the multiple input datasets, and based on the quality index, filtering one or more of the multiple input datasets prior to processing the multiple input datasets using the at least one processing algorithm. ### Claim 17 (depends on 14) The method of claim 14, further comprising: determining a quality index for one or more of the multiple input datasets, wherein the at least one processing algorithm is configured to determine a confidence level of the trend of the fitness based on the quality index. ### Claim 18 (depends on 14) The method of claim 14, further comprising: determining a quality index for one or more of the multiple input datasets, wherein the at least one processing algorithm is configured to provide, based on the quality index and to at least one of the human-machine-interface or a control engine of a measurement apparatus, a request for an additional measurement to complement the multiple input datasets. ### Claim 19 (depends on 14) The method of claim 14, wherein the at least one processing algorithm is an auto-encoder neural network, wherein the trend of the fitness is determined based on a comparison between encoded representations of pairs of the input dataset obtained from an encoder of the auto-encoder neural network. ### Claim 20 (depends on 14) The method of claim 14, further comprising: selecting, from each one of the multiple input datasets, a subset of data, wherein the subsets of data of the multiple input datasets are compared with each other by the at least one processing algorithm. ### Claim 21 A method of training one or more neural networks, the method comprising: configuring a multi-level processing algorithm comprising a multi-task level and a consolidation-task level to determine a fitness of the cardiovascular system of the patient based on at least one input dataset comprising medical measurements of the cardiovascular system, wherein an input of the consolidation-task level is coupled to an output of the multi-task level, wherein the multi-task level is configured to determine multiple diagnostic metrics of the cardiovascular system based on the at least one input dataset, wherein the consolidation-task level is configured to determine the fitness of the cardiovascular system of the patient; and training the one or more neural networks based on the at least one input dataset and one or more labeling functions, the one or more labeling functions determining a ground truth for the at least one input dataset based on one or more of the following inputs to the one or more labeling functions from the group of: a fractional flow reserve measurement; a clinical decision; a patient outcome; a stenosis grade determined based on a coronary angiography; and a wall motion score index one or more characteristics of a subsequent clinical procedure of treatment of the cardiovascular system. ### Claim 22 (depends on 21) The method of claim 21, wherein multiple neural networks of the multi-task level are trained using multiple labeling functions, the multiple labeling functions having inter-dependencies.