Image interpretation model development

A computerized method, performed by a computing system having one or more hardware computer processors and one or more non-transitory computer readable storage device storing software instructions executable by the computing system to perform the computerized method comprising:

• determining a set of training medical imaging exams each including a training report and a training medical image;
• determining one or more finding items and associated finding item criteria;
• for each of the training medical imaging exams:
• use the finding item criteria to reorganize text of the training report into a list of finding items, each associated with text extracted from the training report text;
• use natural language processing to analyze the resultant text associated with each finding item to determine a finding item indication associated with each finding item;
• and
• store, in a training dataset, the training medical imaging exam, the associated finding items, the matching text, and the finding item indications resulting from the analysis of the matching text;
• and
• training a neural network using the training dataset, wherein the neural network learns to output, for other medical imaging exams, one or more finding items and associated finding item classifications based on analysis of respective medical imaging exams.

A computerized method, performed by a computing system having one or more hardware computer processors and one or more non-transitory computer readable storage device storing software instructions executable by the computing system to perform the computerized method comprising:

• determining a set of training medical imaging exams each including a training report and a training medical image;
• determining one or more finding items and associated finding item criteria in a finding item template;
• for each of the training medical exams:
• for each of the finding items in the finding item template:
• applying the associated finding item criteria to text of the training report to identify any matching text;
• analyzing the matching text associated with the finding item to generate a finding item indication indicating whether the matching text indicates a finding or no finding;
• and
• storing, in a training dataset, an association between the training medical exam, the finding items, the matching text for each finding item, and the finding item indication for each finding item;
• training a neural network using the training dataset, wherein the neural network learns to predict finding item indications of medical images not included in the set of training medical exams based on analysis of respective medical images;
• and
• outputting a trained image classification model configured to analyze features of medical images to predict finding item indications of the medical images based on learned relationships between the training medical images and the associated finding item indications.