Method of determining contrast phase of a computerized tomography image
Inventors
Sitek, Arkadiusz • Graf, Benedikt • Xie, Yiting • Katouzian, Amin • Takeuchi, Yusuke • Dufort, Paul
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11302044-B2
Publication Date
2022-04-12
Expiration Date
Abstract
A computer-implemented method for classifying and presenting a contrast phase (CP) of a contrast enhanced computerized tomography (CECT) scan is provided. The method includes training an artificial intelligence (AI) algorithm utilizing a set of CPs labeled CECT data to associate a set of characteristics of the data with a probability associated with the CP. The method includes receiving a new set of unlabeled CECT data, and applying the AI algorithm to the new unlabeled CECT data to associate a first probability of a first CP and a second probability of a second CP. The method also includes providing a graphical representation including the first probability of the first CP and the second probability of the second CP.
Core Innovation
The invention provides a computer-implemented method for classifying and presenting a contrast phase (CP) of a contrast enhanced computerized tomography (CECT) scan. The method trains an artificial intelligence (AI) algorithm using a set of CPs labeled CECT data to associate a set of characteristics of the data with a probability associated with the CP. The method then receives a new set of unlabeled CECT data and applies the trained AI algorithm to associate a first probability of a first CP and a second probability of a second CP.
The invention addresses CP classification for new CECT data by using labeled CPs to create an AI mapping between data characteristics and probabilities for CPs. It outputs probabilities for CPs corresponding to the new unlabeled CECT data. The described presentation of those CP probabilities is integrated into the classification output by providing a graphical representation including the first probability of the first CP and the second probability of the second CP.
In certain embodiments, the graphical representation includes structural encodings that associate polygon features with CP probabilities. A polygon is used such that the number of polygon sides maps to the number of CPs, and each polygon vertex corresponds to a distinct CP. Locations of the vertices of the polygon correspond to 100% probabilities for respective CPs, and markers are provided where distances to polygon vertices encode probabilities for the CP of the new unlabeled CECT data.
Claims Coverage
The independent claims cover a method, a computer system, and a computer program product that train an AI algorithm on CP-labeled CECT data, apply it to new unlabeled CECT data to produce CP probabilities, and present those probabilities graphically. The inventive features include polygon-based probability visualization and, in dependent claims, threshold-triggered CP-specific anomaly determination with an AI model.
AI training for CP probability association
Training an artificial intelligence (AI) algorithm utilizing a set of CPs labeled CECT data to associate a set of characteristics of the data with a probability associated with the CP.
Applying AI to unlabeled CECT to produce CP probabilities
Applying the AI algorithm to the new unlabeled CECT data to associate a first probability of a first CP and a second probability of a second CP.
Graphical representation of CP probabilities
Providing a graphical representation including the first probability of the first CP and the second probability of the second CP.
Polygon visualization mapping CPs to vertices
Providing the graphical representation as a polygon such that the number of sides maps to the number of CPs and each polygon vertex corresponds to a distinct CP.
Vertex locations correspond to 100% probabilities
Configuring the graphical representation so that the locations of the vertices of the polygon represent 100% probabilities for the respective CP values.
Marker distances encode CP probabilities
Providing a graphical representation including at least one marker on an edge of the polygon or within the polygon, where distances from the marker to the respective vertices correspond to probabilities that the CP of new unlabeled CECT data corresponds to the CP associated with those vertices.
Thresholded selection of CP-specific AI model for anomaly detection
Determining that a first probability exceeds a threshold probability; assigning an AI model associated with a first contrast phase; and processing new unlabeled CECT data with the AI model to determine at least one anomaly.
Across the independent claims, the core coverage is the training of an AI algorithm on CP-labeled CECT data to associate data characteristics with per-CP probabilities for new unlabeled CECT data, and the graphical presentation of those probabilities. Dependent claims further specify polygon-based visualization mappings and threshold-based selection of a contrast-phase-associated AI model for anomaly determination.
Stated Advantages
Documented Applications
No documented applications found
Interested in licensing this patent?