Systems, methods, and computer-readable media for using descriptors to identify when a subject is likely to have a dysmorphic feature
Inventors
Gelbman, Dekel • Gurovich, Yaron
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-10722181-B2
Publication Date
2020-07-28
Expiration Date
Abstract
Systems, methods, and computer-readable media are disclosed for identifying when a subject is likely to be affected by a medical condition. For example, at least one processor may be configured to receive information reflective of an external soft tissue image of the subject. The processor may also be configured to perform an evaluation of the external soft tissue image information and to generate evaluation result information based, at least in part, on the evaluation. The processor may also be configured to predict a likelihood that the subject is affected by the medical condition based, at least in part, on the evaluation result information.
Core Innovation
The invention provides a computer-implemented method for determining, from a series of pixels in an image of a subject, whether the subject is likely to be affected by one or more medical conditions. The method uses a plurality of images of positive and negative individuals, including images of a first set of first positive individuals having a first medical condition and images of at least one negative individual who lacks the first medical condition. It calculates descriptors derived from features in each cell within a grid imposed on the positive and negative images, and it calculates descriptors in the same cell grid for an image of the subject.
For each of the positive and negative images, the method produces descriptor-based vectors by aggregating the descriptors of the plurality of cells. It then produces one or more vectors for the image of the subject by aggregating the subject-derived cell descriptors. The method implements a comparison between the vectors of the positive and negative individuals and the one or more vectors of the subject by computing a distance metric between vectors.
Based on the comparison, the method determines a subset of the plurality of vectors more similar to the one or more vectors than a remainder of the plurality of vectors. It determines a first probability score indicating a likelihood that the subject has the first medical condition by analyzing whether vectors in the determined subset exceed a threshold of similarity with the one or more vectors derived from the image of the subject.
Claims Coverage
The document contains three independent claims: a computer-implemented method, an electronic system, and a non-transitory computer-readable medium. Across these independent claims, the core inventive features are centered on grid-based descriptor calculation, aggregation into vectors, vector comparison using a distance metric, selection of a more similar subset, and determining a probability score based on exceeding a similarity threshold.
Grid-based descriptor calculation from image cells
Using a plurality of images of positive and negative individuals, calculating descriptors derived from features in each of a plurality of cells within a grid imposed on each of the plurality of images, including images of a first set of first positive individuals having a first medical condition and images of at least one negative individual who lacks the first medical condition, and calculating descriptors derived from features in each of a plurality of cells within a grid imposed on the image of the subject.
Aggregating cell descriptors into vectors
Producing a plurality of vectors by aggregating the descriptors of the plurality of cells in the plurality of images of the first positive individuals and by aggregating the descriptors of the plurality of cells in the plurality of images of the at least one negative individual, and producing one or more vectors by aggregating the descriptors of the plurality of cells in the image of the subject.
Vector comparison using a distance metric and similarity threshold
Implementing a comparison of the plurality of vectors of the images of the first positive individuals and the at least one negative individual with the one or more vectors of the image of the subject, by computing a distance metric between each of the plurality of vectors and each of the one or more vectors, and determining a first probability score indicating a likelihood that the subject has the first medical condition by analyzing whether vectors in a determined subset exceed a threshold of similarity with the one or more vectors derived from the image of the subject.
Selecting a subset of more similar vectors
Determining, based on the comparison, a subset of the plurality of vectors more similar to the one or more vectors than a remainder of the plurality of vectors.
The independent claims consistently require grid-based descriptor generation over image pixels, aggregation of cell descriptors into vectors, comparison via a distance metric, selection of a subset of more similar vectors, and determining a probability score when similarity exceeds a threshold.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?