Jia Kang

Constructed a workflow for the 3D detection and analysis of neuron numbers in different regions of mouse brain with microscopic images, and integrating the pipeline into a Python package to automate the process.

Preprocessed and registered mouse brain to the Allen Brain Atlas using active contour algorithm, feature based alignment and Elastix.

Developed a “blob structure detection and cell segmentation” path for cell counting in brain slices with 3 steps, which can get rid of the unbalanced distribution problem.

Employed multi-scale Hessian-based vessel enhancement method and graph-cut algorithm to enhance and binarize the 3D CT images and segmented the vessel structure out.

Implemented multi-material decomposition to derive the component maps from dual energy CTs.

Designed an adaptive thresholding method along the centerline of the vessel tree in order to overcome the uneven distribution of contrast agent.

Invented a deep-learning based method to help HoloLens track and predict the position and posture of the handle in VR- AR surgery system, which increased the precision to 97.5% and the speed to about 46 fps.

Employed dual dictionary learning and sparse representation method to solve the 6D pose estimation problem, which can be used to guide the movement of the probes inside patients’ airway.

Conducted liquid-liquid extraction and solid-phase micro-extraction of 11 different kinds of alcohols.

Built SVM classifiers to distinguish liquor’s types and brands with mass spectrometric data from GC×GC-TOFMS analysis.