top of page
Automatic localization and classification of spatially correlated objects in medical images
March 2017 - Present
Research assistant in Kiel University of Applied Science
Automatically analyse the medical images for diagnosis and therapy purposes. A central aspect is to automatically localize and classify organs or landmarks in the medical images:
-
Localize based on gray level by using Regression Tree Ensembles
-
Model topological relationship by using Conditional Random Field
My tasks:
+ Write the Python script for the localization of the joints of fingers after Hough-transform applied using Mean Shift with Gaussian Kernel, resulting over 90% correction localization over 412 testing images with the threshold being post-upscaled image height / 256.
+ Currently, write the Python script to localize the joints of knees, ankles and femurs by running Regression Tree Ensembles and estimating true points in down-scaled images with Polynomial Regression.
bottom of page