top of page

Deep Learning based CT Reconstruction

  • Denoising low dose CT images

  • Reconstruction of high-quality CT images from undersampled sinogram using deep learning algorithm

  • Research in progress: Development of deep learning network structure for beam hardening artifact reduction

Deep Learning based CT Reconstruction

Brain Vessel Extraction using Deep Learning

  • Vessel visualization using MR Images without MR Angiography or Venography.

  • Research in progress: Development of deep learning network structure to visualize brain vessel structure for surgical planning and scan time reduction.

Hyungseob @ mai-lab

Deep Learning for Brain Metastasis Detection

​Deep Learning for Brain Metastasis Detection
​Deep Learning for Brain Metastasis Detection

  • Black-blood (BB) imaging is used to complement contrast-enhanced 3D gradient-echo (CE 3D-GRE) imaging for detecting brain metastases, requiring additional scan time.

  • We proposed deep-learned 3D BB imaging with an auto-labelling technique and 3D convolutional neural networks for brain metastases detection without additional BB scan.

Intelligent MR Image Quality Scoring

Intelligent MR Image Quality Scoring
Intelligent MR Image Quality Scoring

  • A learning-based system that evaluates various image degradation factors in medical images

  • SROCC value is 0.9 or higher compared to the value of the radiologist

A.I. Segmentation

Vertebral Body Segmentation
Vertebral Body Segmentation

  • Medical images region detection and segmentation research in progress

  • Deep Learning algorithm is used for spine vertebral body segmentation

  • High accuracy (about 97.3%) compared to conventional methods

Deep Learning based MR Reconstruction

​Deep Learning based MR Reconstruction

  • Removing aliasing artifacts caused by fast MRI using deep learning algorithm and acquiring clear images

  • Higher evaluation values compared to conventional methods (PSNR: more than 4dB ↑)

​Deep Learning based MR Reconstruction

Intelligent High-Dynamic-Range Imaging

Deep-learning based High-Dynamic-Range Imaging
Deep-learning based High-Dynamic-Range Imaging
Deep-learning based High-Dynamic-Range Imaging

  • Analyze the contrast of the image to create a natural and realistic image

  • Extend the dynamic range of the image using deep learning algorithm

  • Extract HDR specific features: color, contrast, and sharpness which are the three factors that greatly affect human perception

  • Evaluate enhanced images considering subjective perception

bottom of page