Ideen (IDs): Unterschied zwischen den Versionen
Zur Navigation springen
Zur Suche springen
Claas (Diskussion | Beiträge) (4-eye mode for ML based diagnosis) |
Claas (Diskussion | Beiträge) (NFT and DL segmentation) |
||
Zeile 1: | Zeile 1: | ||
* NFTs for medical images (track ownership, track distribution) | |||
* For deep learning segmentation use images reconstructed twice with different filters (edge preserving, smoothing) in training and application | |||
* 4-eye mode for ML based diagnosis | * 4-eye mode for ML based diagnosis | ||
* DLCT: Make every 2nd pixel in upper layer thinner | * DLCT: Make every 2nd pixel in upper layer thinner |
Aktuelle Version vom 22. Februar 2022, 09:06 Uhr
- NFTs for medical images (track ownership, track distribution)
- For deep learning segmentation use images reconstructed twice with different filters (edge preserving, smoothing) in training and application
- 4-eye mode for ML based diagnosis
- DLCT: Make every 2nd pixel in upper layer thinner
- Use 5G in medical imaging
- Kassenbon wird automatisch auf Handy übertragen
- Tantalum-Filter on tube for switching contrast of tantalum contrast agent
- Region-of-Interest MLIR
- 2D stereo anti scatter grid (smaller lamellae for z-direction)
- Stereo CT where detector is located asymmetrical along z
- Twin-beam CT: Use x-deflection for different spectra at central detector rows
- Use stereo for motion estimation
- Stereo Tube at beginning/end of helix
Submitted
- Use technetium for measurement of iodine contrast in ventricle
- Cardiac CT: Frequency split
- HD: Dual-Layer detector with different pixel sizes in each layer
- Use survey scan for detection of patient parts outside of SFOV
- Stereo-Tube with inplane shift of focal-spots
- x-ray tube with oscillating anode speed
- DART for EFOV
- Stereo CT with central collimator for full-scan acquisitions (intersection point at rotation axis)
- Use helical scout scan as start image for axial MLIR / detect motion (based on bone segmentation)
- Different slice thicknesses in SFOV and EFOV
- Use thresholded start image for MLIR