Device for the qualitative evaluation of human organs
Inventors
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12016696-B2
Publication Date
2024-06-25
Expiration Date
Abstract
This method for qualitatively evaluating human livers comprises: a step (301) of computing normalized histograms of colour channels from a portion of a photograph of a liver;a step (304) of loading coefficients obtained at the end of a training phase;a step (305) of extracting from the histograms values corresponding to variables retained at the end of the training phase;a step (306) of computing a linear combination of the extracted values weighted with the loaded coefficients; anda step (308, 309) of displaying information representative of the result of the computation of the linear combination.
Core Innovation
The invention relates to a device and method for qualitatively evaluating human livers by capturing an image of at least a portion of a liver when the liver is in a donor's body, already removed, or placed in a hypothermic, normothermic and/or subnormothermic graft perfusion machine. An image processor selects at least one portion of the captured image, and an estimation means extracts image characteristics from the selected portion and compares them with corresponding image characteristics saved in a database that associates training image characteristics with liver variables indicating health conditions.
A classifier model trained to classify liver variables in accordance with image characteristics is used to assign a quality evaluation to the liver represented by the captured image based on classification of the image characteristics. The estimation means is further configured to calculate a predicted classification error in accordance with variables in the classification and to compute a value representative of steatosis of the liver.
The document further specifies image processing and display aspects that support the qualitative evaluation, including normalized color-channel histogram features and training-derived coefficients. It also describes calculation and display of an uncertainty interval and steatosis classes, and image-quality control through sharpness estimation based on filtering.
Claims Coverage
The document includes two independent claims, a device claim and a method claim. The independent claims center on image capture and portion selection, comparison against a database of training-derived image characteristics mapped to liver variables for classification-based quality evaluation, and computation of steatosis via a histogram-based linear combination with multiplicator coefficients, including calculation of predicted classification error.
Capture and portion selection for liver health evaluation
A camera for capturing an image of at least a portion of a liver when the liver is in a donor's body, already removed, or placed in a hypothermic, normothermic and/or subnormothermic graft perfusion machine, and an image processor configured to select at least one portion of the captured image.
Training-derived database mapping image characteristics to liver variables
A database including image characteristics extracted from a training image of a liver and associated with liver variables indicating health conditions of the liver shown on the training image.
Classifier model for quality evaluation of liver variables
A classifier model trained to classify liver variables in accordance with image characteristics of liver images saved in the database.
Quality evaluation based on classification compared to database image characteristics
Means for estimating configured to extract image characteristics of the selected portion and compare the image characteristics with corresponding image characteristics in the database using the classifier model, and to assign a quality evaluation to the liver represented by the captured image in accordance with classification of said image characteristics.
Predicted classification error estimation
The means for estimating further configured to calculate a predicted classification error in accordance with variables in said classification.
Histogram-based steatosis computation using multiplicator coefficients
The means for estimating configured to compute a value representative of steatosis of the liver, the value being a linear combination of pixel numbers of color values, wherein said color values are components of a histogram of the image, and wherein the pixel numbers are assigned multiplicator coefficients.
Capture and portion selection for classifier-based liver evaluation
Capturing an image of at least a portion of a liver when the liver is in a donor's body, already removed, or placed in a hypothermic, normothermic and/or subnormothermic graft perfusion machine, selecting at least a portion of the captured image, and extracting image characteristics of the selected portion for comparing with corresponding image characteristics in a database.
Training and classification using database image characteristics
Training a classifier model to classify liver variables in accordance with image characteristics of liver images saved in the database, comparing extracted image characteristics with corresponding image characteristics saved in the database including image characteristics and liver variables associated with health conditions from training images, and assigning a quality evaluation to the liver represented by the captured image in accordance with classification of the image characteristics using the classifier model.
Predicted classification error calculation in the method
Calculating a predicted classification error in accordance with variables in the classification.
Histogram-based steatosis computation using multiplicator coefficients
Computing a value representative of steatosis of the liver, the value being a linear combination of pixel numbers of color values, wherein the color values are components of a histogram of the image, and wherein the pixel numbers are assigned multiplicator coefficients.
Across both independent claims, the inventive core is a classifier-based qualitative evaluation of human liver health using image characteristics extracted from a selected portion of a captured liver image and compared to a database of training-derived image characteristics mapped to liver variables, with additional computation of predicted classification error and a steatosis value computed as a linear combination of histogram-based color components using multiplicator coefficients.
Stated Advantages
Assigning a quality evaluation to a liver represented by a captured image in accordance with classification of image characteristics.
Calculating a predicted classification error in accordance with variables in the classification.
Computing a value representative of steatosis of the liver using a histogram-based linear combination of color components with multiplicator coefficients.
Documented Applications
Qualitatively evaluating human livers by using captured images of a liver portion while the liver is in a donor's body, already removed, or placed in a hypothermic, normothermic and/or subnormothermic graft perfusion machine.
Qualitatively evaluating hepatic steatosis and steatosis classes from smartphone/digital camera RGB images, including computing a steatosis value and associated predictive outputs.
Automatic liver segmentation using a fully convolutional neural network, with performance reporting using Dice similarity coefficient, Recall, and Precision.
Interested in licensing this patent?