Deep-learning-based medical image interpretation system for animals
A deep-learning-based medical image interpretation system and method uses one or more neural network models to analyze medical images of animals. Result sets from the one or more neural network models are used to create one or more medical diagnoses. One or more treatment options are generated in response to the one or more medical diagnoses. One or more animal products may be suggested related to care of the one or more medical diagnoses.
1 . A method, comprising:
creating a training set for one or more neural network models using a set of medical images for a representational set of animals, the medical images include case specific medical images and normal, associated, medical images, each medical image having one or more descriptive labels relating to desired analysis predictions;
the creating the training set comprises preprocessing the set of medical images using one or more transformations or augmentations on each medical image in the set of medical images to isolate desired regions or remove unwanted regions of the medical image;
training one or more neural network models using the training set, the one or more neural networks determine if an input image is normal or indicates a case condition;
generating a set of results for a set of input medical images using the one or more neural network models;
processing the set of results using one or more rules engines to apply an error correction to the set of results to improve an overall quality of the set of results, the one or more rules engines determine whether a case condition prediction in the set of results is within an acceptable error rate based on the set of input medical images, the one or more rules engines are trained on outputs from a validation set to determine optimized error corrected results;
processing the processed set of results to communicate, for a specific delivery method, case predictions that include one or more medical diagnoses.
2 . The method as recited in claim 1 , wherein the case predictions include one or more animal product recommendations based on the processed set of results.
3 . The method as recited in claim 1 , wherein the processing the processed set of results further comprises:
generating a confidence metric for at least one of the one or more medical diagnoses;
displaying the confidence metric in proximity to the at least one of the one or more medical diagnoses having a generated confidence metric.
4 . The method as recited in claim 1 , wherein the processing the processed set of results further comprises:
generating a confidence metric for at least one of the one or more medical diagnoses;
graphically displaying the confidence metric in proximity to the at least one of the one or more medical diagnoses having a generated confidence metric.
5 . The method as recited in claim 1 , wherein the processing the processed set of results further comprises:
generating a confidence metric for the one or more medical diagnoses;
graphically displaying the confidence metric of the one or more medical diagnoses having a generated confidence metric.
6 . The method as recited in claim 1 , wherein the processing the processed set of results further comprises:
identifying a medical diagnosis of the one or more medical diagnoses as critical to care;
wherein the report places the medical diagnosis at or near the top of a section of the report based on the identification that the medical diagnosis is critical to care.
7 . The method as recited in claim 1 , wherein the case predictions include one or more treatment options based on the processed set of results.
8 . One or more non-transitory computer-readable storage media, storing one or more sequences of instructions, which when executed by one or more processors cause performance of:
creating a training set for one or more neural network models using a set of medical images for a representational set of animals, the medical images include case specific medical images and normal, associated, medical images, each medical image having one or more descriptive labels relating to desired analysis predictions;
the creating the training set comprises preprocessing the set of medical images using one or more transformations or augmentations on each medical image in the set of medical images to isolate desired regions or remove unwanted regions of the medical image;
training one or more neural network models using the training set, the one or more neural networks determine if an input image is normal or indicates a case condition;
generating a set of results for a set of input medical images using the one or more neural network models;
processing the set of results using one or more rules engines to apply an error correction to the set of results to improve an overall quality of the set of results, the one or more rules engines determine whether a case condition prediction in the set of results is within an acceptable error rate based on the set of input medical images, the one or more rules engines are trained on outputs from a validation set to determine optimized error corrected results;
processing the processed set of results to communicate, for a specific delivery method, case predictions that include one or more medical diagnoses.
9 . The one or more non-transitory computer-readable storage media as recited in claim 8 , wherein the case predictions include one or more animal product recommendations based on the processed set of results.
10 . The one or more non-transitory computer-readable storage media as recited in claim 8 , processing the processed set of results further comprises:
generating a confidence metric for at least one of the one or more medical diagnoses;
displaying the confidence metric in proximity to the at least one of the one or more medical diagnoses having a generated confidence metric.
11 . The one or more non-transitory computer-readable storage media as recited in claim 8 , wherein the processing the processed set of results further comprises:
generating a confidence metric for at least one of the one or more medical diagnoses;
graphically displaying the confidence metric in proximity to the at least one of the one or more medical diagnoses having a generated confidence metric.
12 . The one or more non-transitory computer-readable storage media as recited in claim 8 , wherein the processing the processed set of results further comprises:
generating a confidence metric for the one or more medical diagnoses;
graphically displaying the confidence metric of the one or more medical diagnoses having a generated confidence metric.
13 . The one or more non-transitory computer-readable storage media as recited in claim 8 , wherein the processing the processed set of results further comprises:
identifying a medical diagnosis of the one or more medical diagnoses as critical to care;
wherein the report places the medical diagnosis at or near the top of a section of the report based on the identification that the medical diagnosis is critical to care.
14 . The one or more non-transitory computer-readable storage media as recited in claim 8 , wherein the case prediction include one or more treatment options based on the processed set of results.
15 . An apparatus, comprising:
one or more hardware processors; and
a memory storing instructions, which when executed by the one or more processors, cause the one or more processors to:
create a training set for one or more neural network models using a set of medical images for a representational set of animals, the medical images include case specific medical images and normal, associated, medical images, each medical image having one or more descriptive labels relating to desired analysis predictions;
the create the training set comprises preprocessing the set of medical images using one or more transformations or augmentations on each medical image in the set of medical images to isolate desired regions or remove unwanted regions of the medical image;
train one or more neural network models using the training set, the one or more neural networks determine if an input image is normal or indicates a case condition;
generate a set of results for a set of input medical images using the one or more neural network models;
process the set of results using one or more rules engines to apply an error correction to the set of results to improve an overall quality of the set of results, the one or more rules engines determine whether a case condition prediction in the set of results is within an acceptable error rate based on the set of input medical images, the one or more rules engines are trained on outputs from a validation set to determine optimized error corrected results;
process the processed set of results to communicate, for a specific delivery method, case predictions that include one or more medical diagnoses.
16 . The apparatus as recited in claim 15 , wherein the case predictions include one or more animal product recommendations based on the processed set of results.
17 . The apparatus as recited in claim 15 , wherein the process the processed set of results further comprises:
generate a confidence metric for the one or more medical diagnoses;
graphically display the confidence metric of the one or more medical diagnoses having a generated confidence metric.