Enlitic, Inc.

Eric C Poblenz

Kevin Lyman

Li Yao

Anthony Upton

Ben Covington

Jordan Prosky

Chris Croswhite

Lionel Lints

Keith Lui

Ashwin Jadhav

A medical scan quality assurance system is operable to utilize artificial intelligence to train at least one computer vision model based on a training set of medical scans. A set of medical scans are received. Quality assurance data is generated for the set of medical scans utilizing artificial intelligence by performing at least one quality assurance function on the set of medical scans by utilizing the at least one computer vision model. A first medical scan is identified in the set of medical scans to include an artifact, detected by performing the at least one quality assurance function, that is determined to obscure at least a threshold percentage of a key anatomical part based on the quality assurance data. An artifact obstruction notification indicating the first medical scan is generated for transmission to a client device for display.

Artifact detection via a medical scan quality assurance system

A multi-model medical scan analysis system is operable to generate a plurality of training sets from a plurality of medical scans. Each of a set of sub-models is generated by performing a training step on a corresponding one of the plurality of training sets of the plurality of medical scans. A set of abnormality data is generated by applying a subset of a set of inference functions on a new medical scan. The subset of the set of inference functions utilize the subset of the set of sub-models, and each of the set of abnormality data is generated as output of performing one of the subset of the set of inference functions. The multi-model medical scan analysis system is further operable to generate final abnormality data that includes a global probability indicating a probability that any abnormality is present based on the set of abnormality data.

Generating final abnormality data for medical scans based on utilizing a set of sub-models

A medical scan header standardization system is operable to determine a plurality of counts for a plurality of entries of at least one of a standard set of fields for headers of a plurality of medical images. A standard set of header entries is determined for at least one of the standard set of fields based on including ones of the entries for the each of the standard set of fields with counts of the plurality of counts that compare favorably to a threshold. One of the standard set of header entries is selected to replace an entry of a field of a header of a medical image. A computer vision model is trained utilizing a training set of images that includes the medical image and the selected one of the standard set of header entries. Inference data for at least one new medical scan is generated based on utilizing the computer vision model.

Medical scan header standardization system and methods for use therewith

A location-based medical scan analysis system is operable to generate a generic model by performing a training step on image data of a plurality of medical scans. Location-based subsets of the plurality of medical scans are generated by including ones of the plurality of medical scans with originating locations that compare favorably to location grouping criteria for the each location-based subset. A plurality of location-based models are generated by performing a fine-tuning step on the generic model, utilizing a corresponding one of the plurality of location-based subsets. Inference data is generated for a new medical scan by utilizing one of the location-based models on the new medical scan, where an originating location associated with the new medical scan compares favorably to location grouping criteria for the location-based subset utilized to generate the location-based model. The inference data is transmitted to a client device for display via a display device.

Location-based medical scan analysis system

A model-assisted annotating system is operable to receive a first set of annotation data, corresponding to a broad type of annotation data output. A first training step is performed to train a computer vision model using the first set of annotation data. A second set of annotation data corresponding to the broad type of annotation data output is generated performing an inference function utilizing the computer vision model on medical scans. Additional annotation data further specifies the broad type of annotation data output is received. A second training step is performed to generate an updated computer vision model using set of additional annotation data. A third set of annotation data corresponding to the specified type of annotation data output is generated by performing an updated inference function utilizing the updated computer vision model on medical scans.

Model-assisted annotating system and methods for use therewith

A multi-model medical scan analysis system is operable to generate a generic model by performing a training step on image data of a plurality of medical scans and corresponding labeling data. A plurality of fine-tuned models are generated by performing a fine-tuning step on the generic model. Abnormality detection data is generated for a new medical scan by utilizing the generic model. A first one of the plurality of abnormality types that is detected in the new medical scan is determined based on a corresponding one of the plurality of probability values. Additional abnormality data is generated by performing a fine-tuned inference function on the image data of the new medical scan that utilizes one of the plurality of fine-tuned models that corresponds to the first one of the plurality of abnormality types. The additional abnormality data is transmitted for display.

Multi-model medical scan analysis system using fine-tuned models

A method includes generating first contrast significance data for a first computer vision model generated from a first training set of medical scans. First significant contrast parameters are identified based on the first contrast significance data. A first re-contrasted training set is generated based on performing a first intensity transformation function on the first training set of medical scans, where the first intensity transformation function utilizes the first significant contrast parameters. A first re-trained model is generated from the first re-contrasted training set, which is associated with corresponding output labels based on abnormality data for the first training set of medical scans. Re-contrasted image data of a new medical scan is generated based on performing the first intensity transformation function. Inference data indicating at least one abnormality detected in the new medical scan is generated based on utilizing the first re-trained model on the re-contrasted image data.

Re-training a model for abnormality detection in medical scans based on a re-contrasted training set

A label generating system operates to generate an artificial intelligence model by: training on a training data set that includes the plurality of medical scans with the corresponding global labels; generating testing global probability data by performing an inference function that utilizes the artificial intelligence model on the plurality of medical scans with the corresponding global labels, wherein the testing global probability data indicates a testing set of global probability values corresponding to the set of abnormality classes, and wherein each of the testing set of global probability values indicates a probability that a corresponding one of the set of abnormality classes is present in each of the plurality of medical scans with the corresponding global labels; comparing the testing set of global probability values to a corresponding confidence threshold for each of the plurality of medical scans selected based on the corresponding one of the global labels; generating an updated training data set by correcting ones of the plurality of medical scans having a corresponding one of the testing set of global probability values that compares unfavorably to the corresponding confidence threshold; and retraining the artificial intelligence model based on the updated training set.

AI-based label generating system and methods for use therewith

An accession number correction system is operable to determine that an accession number of a DICOM image does not link to any corresponding one of a plurality of medical reports. Medical report criteria is generated based on the DICOM image, and a set of medical reports are identified based on the medical report criteria. A computer vision model is trained from a training set of DICOM images, and inference data is generated for the DICOM image by performing at least one inference function utilizing the computer vision model. A selected one of the set of medical reports that corresponds to the DICOM image is determined based on comparing the inference data for the DICOM image to text included in at least one of the set of medical reports. Updated report header data for the selected medical report is generated, and storage of the updated report header data is facilitated.

Accession number correction system and methods for use therewith

An intensity transform augmentation system is operable to generate a plurality of sets of augmented images by performing a set of intensity transformation functions on each of a training set of medical scans. Each of the set of intensity transformation functions are based on density properties of corresponding anatomy feature present in the training set of medical scans. A computer vision model is generated by performing a training step on the plurality of sets of augmented images, where each augmented image of a set of augmented images is assigned same output label data based on a corresponding one of the training set of medical scans. Inference data is generated by performing an inference function on a new medical scan by utilizing the computer vision model on the new medical scan. The inference data is transmitted to a client device for display via a display device.

Intensity transform augmentation system and methods for use therewith

A medical scan triaging system is operable to train a computer vision model and to generate abnormality data indicating abnormality probabilities for medical scans via the computer vision model. A first subset of medical scans is determined by identifying medical scans with abnormality probabilities greater than a first probability value of a triage probability threshold. A second subset of medical scans is determined by identifying medical scans with abnormality probabilities less than the first probability value. An updated first subset of medical scans is determined by identifying medical scans with abnormality probabilities greater than a second probability value of an updated triage probability threshold. An updated second subset of the plurality of medical scans is determined by identifying medical scans with a abnormality probabilities less than the second probability value. The updated first subset of medical scans is transmitted to client devices.

Medical scan triaging system and methods for use therewith

A multi-label heat map generating system is operable to receive a plurality of medical scans and a corresponding plurality of medical labels that each correspond to one of a set of abnormality classes. A computer vision model is generated by training on the medical scans and the medical labels. Probability matrix data, which includes a set of image patch probability values that each indicate a probability that a corresponding one of the set of abnormality classes is present in each of a set of image patches, is generated by performing an inference function that utilizes the computer vision model on a new medical scan. Preliminary heat map visualization data can be generated for transmission to a client device based on the probability matrix data. Heat map visualization data can be generated via a post-processing of the preliminary heat map visualization data to mitigate heat map artifacts.

AI-based heat map generating system and methods for use therewith

An intensity transform augmentation system is operable to receive a training set of medical scans. Random intensity transformation function parameters are generated for each medical scan of the training set of medical scans. A plurality of augmented images are generated, where each of the plurality of augmented images is generated by performing a intensity transformation function on one of the training set of medical scans by utilizing the random intensity transform parameters generated for the one of the training set of medical scan. A computer vision model is generated by performing a training step on the plurality of augmented images. A new medical scan is received via the receiver. Inference data is generated by performing an inference function that utilizes the computer vision model on the new medical scan. The inference data is transmitted to a client device for display via a display device.

Computer vision model training via intensity transform augmentation

A multi-label heat map generating system is operable to receive a plurality of medical scans and a corresponding plurality of global labels that each correspond to one of a set of abnormality classes. A computer vision model is generated by training on the medical scans and the global labels. Probability matrix data, which includes a set of image patch probability values that each indicate a probability that a corresponding one of the set of abnormality classes is present in each of a set of image patches, is generated by performing an inference function that utilizes the computer vision model on a new medical scan. Heat map visualization data can be generated for transmission to a client device based on the probability matrix data that indicates, for each of the set of abnormality classes, a color value for each pixel of the new medical scan.

AI-based multi-label heat map generating system and methods for use therewith

A global multi-label generating system is operable to receive a plurality of medical scans and a corresponding plurality of global labels that each correspond to one of a set of abnormality classes. A computer vision model is generated by training on the medical scans and the global labels. Probability matrix data, which includes a set of image patch probability values that each indicate a probability that a corresponding one of the set of abnormality classes is present in each of a set of image patches, is generated by performing an inference function that utilizes the computer vision model on a new medical scan. Global probability data that includes a set of global probability values each indicating a probability that a corresponding one of the set of abnormality classes is present in the new medical scan is generated based on the probability matrix data for transmission to a client device.

Growing community of inventors

Palo Alto, CA, United States of America

Eric C Poblenz