Tissue damage assessment method and system
Inventors
Makarenko, Vladimir • Uplinger, II, James Robert
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11154211-B2
Publication Date
2021-10-26
Expiration Date
Abstract
A method and system to assess data from a dielectric probe to determine the health of tissue by denoising the data and clustering the data points. The assessment is used to assist medical professionals in the care patients.
Core Innovation
The invention relates to de-noising dielectric probe data from a tissue based on measurements of tissue permittivity in response to received dielectric probe data containing measurements of the tissue permittivity. A tissue type is identified by a dielectric response, and the received dielectric probe data are clustered by varying one or both of a minimum number of data points within a cluster and a maximum distance between data points within the cluster to determine a boundary of the cluster. Data points that are not on the boundary or within an interior of the cluster are discarded as noise.
The de-noised data points are mapped, and the mapped de-noised data points are used to assess tissue damage. In some embodiments, remaining data points that are on the boundary and within the cluster are mapped onto n-dimensional space, and the output of the mapping is used to assess tissue damage. The mapping of resulting clusters onto the n-dimensional space is performed after discarding non-clustered data points that are classified as noise.
In embodiments with training, validation, and testing data sets, the clustering algorithm is trained with a training data set and validated with a validation data set, and then applied to a received testing data set. A parameter of data point distance and a parameter of data point nearness are selected for the trained and validated clustering algorithm, and a value of minimum points for a cluster is selected and then varied successively until an algorithm clustering variation is generally stable.
A system architecture is also described with a distance layer, a minimum point layer, and a nearness layer, where the minimum point layer and nearness layer successively vary a minimum point parameter and a nearness parameter to produce clustering and to define boundaries of resulting clusters.
Claims Coverage
Independent claim coverage spans 4 independent claims, describing de-noising dielectric probe data, clustering permittivity measurements to determine cluster boundaries, discarding non-clustered/noise points, mapping resulting clusters, and assessing tissue damage.
De-noising dielectric probe data using clustering boundaries and noise discarding
In response to received dielectric probe data containing measurements of tissue permittivity, identifying a tissue type by a dielectric response; clustering by varying one or both of a minimum number of data points within a cluster and a maximum distance between data points within said cluster to determine a boundary of said cluster; discarding data points that are not on said boundary or within an interior of said cluster as noise; de-noised data points, when mapped, assessing tissue damage.
Training, validation, and testing with successive parameter variation until clustering is stable
Processing a received testing data set with a clustering algorithm trained with a training data set and validated with a validation data set; selecting a parameter of data point distance and a parameter of data point nearness; selecting a value of minimum points for a cluster; varying successively the data point nearness parameter and the value of minimum points until an algorithm clustering variation is generally stable; discarding non-clustered data points classified as noise; mapping resulting clusters onto n-dimensional space; mapped clusters assessing tissue damage.
Mapping boundary and interior points onto n-dimensional space to assess tissue damage
Clustering received dielectric probe data with a trained and validated dimensionless clustering algorithm to create one or more clusters; identifying cluster boundary data points; discarding data points not within said cluster or on a boundary of said cluster; mapping remaining data points that are on said boundary and within said cluster onto n-dimensional space; outputting said mapping to assess tissue damage.
Layered system for distance measurement, successive minimum-point and nearness variation, and cluster boundary definition
A distance layer that measures data points received from a dielectric probe and retains resulting data point distance; a minimum point layer and a nearness layer with memory portions; the minimum point layer and nearness layer successively varying a minimum point parameter and a nearness parameter to produce clustering; the nearness layer defining boundaries of resulting clusters; data points not within said resulting clusters being removed to result in de-noised data points; said de-noised data points assessing tissue damage.
Across the independent claims, the core inventive approach is de-noising dielectric probe data by clustering permittivity measurements to determine cluster boundaries, discarding non-clustered/noise points, mapping resulting clusters, and using the mapped clusters or mapped output to assess tissue damage. Training, validation, testing, and successive variation of nearness and minimum-points until clustering stability are included in one independent claim, and a layered system architecture with distance measurement plus successive minimum-point and nearness varying is included in another independent claim.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?