Self-aware image segmentation methods and systems

Abstract

The following relates generally to image segmentation. In one aspect, an image is received and preprocessed. The image may then be classified as segmentable if it is ready for segmentation; if not, it may be classified as not segmentable. Multiple, parallel segmentation processes may be performed on the image. The result of each segmentation process may be marked as a potential success (PS) or a potential failure (PF). The results of the individual segmentation processes may be evaluated in stages. An overall failure may be declared if a percentage of the segmentation processes marked as PF reaches a predetermined threshold.

1 . An image segmentation method, comprising:
 classifying, with a computer-implemented binary classifier, an input image as segmentable using a computer-implemented segmentation process or not segmentable using the computer-implemented segmentation process;   segmenting the input image using the computer-implemented segmentation process if the input image is classified as segmentable; and   performing a remedial process if the input image is classified as not segmentable.   
     
     
         2 . The method of  claim 1  further comprising:
 performing computer-implemented pre-processing of the input image prior to the classifying, the classifying being performed on the pre-processed input image; 
 wherein the remedial process includes performing further computer-implemented pre-processing of the input image. 
 
     
     
         3 . The method of  claim 1  further comprising:
 acquiring the input image using a medical imaging system; 
 wherein the remedial process comprises acquiring a new input image using the medical imaging system with a different imaging configuration. 
 
     
     
         4 . The method of  claim 1  further comprising:
 during a training phase performed prior to the classifying, training the binary classifier using a computer-implemented training process operating on a set of training images each labeled as segmentable or not segmentable. 
 
     
     
         5 . The method of  claim 4  wherein the training phase further comprises:
 segmenting each training image using the computer-implemented segmentation process and labeling the training image as segmentable or not segmentable based on an output of the segmenting. 
 
     
     
         6 . The method of  claim 1 , wherein the computer-implemented segmentation process comprises multiple, parallel segmentation processes. 
     
     
         7 . The method according to  claim 6 , wherein each segmentation process of the multiple, parallel segmentation processes is different from every other segmentation process of the multiple, parallel segmentation processes. 
     
     
         8 . The method according to  claim 6 , wherein each segmentation process of the multiple, parallel segmentation processes has a different segmentation process initialization generated by a random perturbation of a baseline segmentation process initialization. 
     
     
         9 . The method according to  claims 6 , wherein the computer-implemented segmentation process further comprises (1) grouping segmentation results of the multiple, parallel segmentation processes to identify a group of mutually similar segmentation results and (2) generating a final segmentation result for the input image based on the group of mutually similar segmentation results.