Systems and methods for analysis of anatomical images

A method of training a statistical classifier for identification of at least one of a plurality of visual findings in at least one of two anatomical images of a target individual, comprising:

• providing, for each of a plurality of sample individuals, two anatomical 2D images of a body portion of the respective sample individual each captured at a respective unique orientation of at least the body portion of the target individual, and a text based radiology report including a plurality of sentences;
• creating, for each of the plurality of sample individuals, a respective tag according to an analysis that maps at least one of the plurality of sentences to a respective certain indication of a plurality of indications of distinct visual findings present in at least one of the two anatomical 2D images of the respective individual;
• and
• training a statistical classifier for identification of at least one of a plurality of visual findings in at least one of two anatomical 2D images of the body portion of a target individual according to a training dataset comprising the respective two anatomical images and respective tag for each of the plurality of sample individual,
• wherein the statistical classifier includes a first processing path comprising a first CNN component that outputs a first feature vector when fed a first anatomical 2D image captured at a first unique orientation, a second processing path comprising a second CNN component that outputs a second feature vector when fed a second anatomical 2D image captured at a second unique orientation, the first feature vector outputted by the first processing path and the second feature vector outputted by the second processing path are concatenated and fed into a combined processing path comprising a fully connected layer component of the statistical classifier that outputs the indication of at least one of the plurality of visual findings in at least one of two anatomical 2D images.

A system for identification of an indication of at least one of a plurality of visual findings in at least one of two anatomical images of a target individual, comprising:

• a non-transitory memory having stored thereon a code for execution by at least one hardware processor, the code comprising:
• code for providing two anatomical 2D images of a body portion of a target individual, each of the two anatomical 2D images captured at a unique orientation of at least the body portion of the target individual;
• code for inputting a first anatomical 2D image of the two anatomical 2D images captured at a first unique orientation into a first processing path comprising a first convolutional neural network (CNN) component of a statistical classifier to output a first feature vector;
• code for inputting a second anatomical image 2D of the two anatomical 2D images captured at a second unique orientation into a second processing path comprising a second CNN component of the statistical classifier to output a second feature vector;
• code for concatenating the first feature vector outputted by the first processing path and the second feature vector outputted by the second processing path;
• code for inputting the concatenation of the first feature vector and the second feature vector into a combined processing path comprising a fully connected layer component of the statistical classifier;
• and
• code for computing an indication of at least one of a plurality of distinct visual findings present in at least one of the two anatomical 2D images by the fully connected layer,
• wherein the statistical classifier is trained on a training dataset including, for each of a plurality of sample individuals, two anatomical 2D images of a body portion of the respective sample individual each captured at a respective unique orientation of at least the body portion of the target individual, and a tag associated with the two anatomical images, wherein the tag is created based on an analysis that maps respective individual sentences of a text based radiology report to one of a plurality of indications of visual findings.