The State University of New York

Konstantin Dmitriev

Arie E Kaufman

Arie Kaufman

The exemplary system, method, and computer-accessible medium for generating a multiclass image segmentation model(s) can include receiving multiple single-class image datasets, receiving a target mask for each of the single-class image datasets, receiving a condition of an object associated with each of the single-class image datasets, and generating the multiclass image segmentation model(s) based on the single-class image datasets, the target masks, and the identification of the target objects. The single-class image datasets can include computer tomography images of abdominal organs. The single-class image datasets can be non-overlapping single-class image datasets. The single-class image datasets can include medical imaging datasets or cityscape datasets. The condition can include (i) an identification of a target object associated with each image in each single-class image dataset, (ii) a classification of each image associated with each single-class image dataset or (iii) an identifiable detail regarding each image in each single-class image datasets.

System, method, and computer-accessible medium for generating multi-class models from single-class datasets

A system, method, and computer-accessible medium for using medical imaging data to screen for a cystic lesion(s) can include, for example, receiving first imaging information for an organ(s) of a one patient(s), generating second imaging information by performing a segmentation operation on the first imaging information to identify a plurality of tissue types, including a tissue type(s) indicative of the cystic lesion(s), identifying the cystic lesion(s) in the second imaging information, and applying a first classifier and a second classifier to the cystic lesion(s) to classify the cystic lesion(s) into one or more of a plurality of cystic lesion types. The first classifier can be a Random Forest classifier and the second classifier can be a convolutional neural network classifier. The convolutional neural network can include at least 6 convolutional layers, where the at least 6 convolutional layers can include a max-pooling layer(s), a dropout layer(s), and fully-connected layer(s).

Growing community of inventors

Port Jefferson, NY, United States of America

Konstantin Dmitriev