Reconstruction From Projection TRIZ Cases

Pre-learned dictionaries speed low-dose electron microscopy reconstruction, enabling live previews for focusing and alignment with less beam damage.

Spatially distributed x-ray focal spots and spectral filters cut scatter and metal artifacts in volumetric CT while enabling dual-energy imaging.

By modeling e-beam expansion and attenuation, SEM tomography reconstructs 3D maps of non-thin specimens without destructive lamella prep.

Matching media and a boundary-model lid help head EMT scanners image through skull shielding and complex field distributions more precisely.

A boundary-model lid, matching media, and superheterodyne antenna control improve head EMT accuracy despite bone shielding and dielectric contrast.

Continuous translation and angled composite imaging enable 3D tomography of elongated samples with uniform electron dosage and less stitching.

Probe-beam scanning with induced current or voltage reconstruction adds depth to charge collection analysis, revealing local defects in samples.

A three-block refractive color splitter redirects green light to the target pixel, improving color separation, light intake, and crosstalk control.

Matching media and an anatomy-mimicking lid improve head EMT calibration, helping signals penetrate skull shielding for more accurate brain imaging.

Combining OCT height profiles with reflected-light images cuts scan count and improves workpiece surface positioning and feature localization.

Passive temporal shearing and compressed sensing reconstruct 2D transient events with high spatial resolution from few ultrafast measurements.

A dual-branch CSformer fuses CNN local features with transformer context to improve compressive sensing reconstruction with lower complexity.

Cloud compression and remote processing turn slow multi-dimensional ultrasound reconstruction into real-time imaging with less local compute burden.

A deep CNN learns the inverse map from compressive measurements to signals, cutting recovery time while preserving reconstruction quality.

Compressed sensing cuts MPI calibration measurements for 3D system function mapping while preserving reconstruction accuracy and signal quality.

Compressed sensing reconstructs the MPI system matrix from fewer voxel measurements, cutting calibration time while preserving accuracy.

An m-estimator and sparsity transform iteratively rebuild incomplete, noisy signal data with high reconstruction quality and manageable compute time.

Pixel-parameter calibration corrects CT projection data without exact spectral curves, improving material density image accuracy.

Displacement vector fields align overlapping CT partial images from different heart cycles to reduce motion artifacts and stack discontinuities.

A trained correction model removes beam hardening, detector shading, and position artifacts from spectral radiography projection data.

Joint self-supervised reconstruction and coil sensitivity calibration improves parallel MRI image quality without ground-truth data or pre-calibration.

Multiple energy windows separate scatter and tailing photons, enabling iterative NM image reconstruction with cleaner CZT detector images.

Decoupled filtering and resampling shrink CT detector data for faster transfer and processing while preserving spatial resolution and noise.

Sensing coils and a residual U-Net separate changing electromagnetic interference from MRI signals to reconstruct clearer images without shielding.

Synthetic forward projections from a tomosynthetic volume create seamless large-area x-ray views and align previews to current device orientation.

Dual-energy X-ray imaging uses a spectrally distinct contrast agent to separate vessels from bone and calcification at lower radiation dose.

CT voxel material classification replaces linear scaling to build more accurate PET attenuation maps for MR hardware and metal implants.

Projection-data contrast tracking sets the prescan-to-main-scan timing in photon-counting CT, reducing contrast dose, lag, and radiation exposure.

Using L-BFGS in unrolled MRI reconstruction improves data consistency, cutting iterations and memory use while preserving image quality.

Missing coincidence counts are tracked during PET buffering and transmission, then used to correct SUV values and improve image accuracy.

Weighted self-supervised MRI reconstruction cuts scan time and artifacts while avoiding groundtruth training data and heavy memory use.

Different regularizers for truncated and non-truncated CT volume regions suppress truncation artifacts and improve reconstruction stability.

Out-of-plane artifact correction in DBT slice images preserves lesion visibility while reducing image count, review burden, and storage needs.

Passive TMR sensing locates small ferromagnetic foreign bodies in the body with 3D probe guidance, avoiding radiation and generated EM fields.

Self-supervised cardiac MRF reconstruction uses acquired k-space data and a self-consistency loss to avoid large training sets and dictionary matching.

Multiple robotic-arm X-ray views and object-specific simulation correct scatter artifacts, improving 3D reconstruction while reducing radiation exposure.

When patient motion interrupts PET or SPECT scans, adaptive acquisition timing extends data collection to preserve image quality.

Beamforming on raw multi-channel k-space data creates virtual MRI coils that favor ROI signals and suppress outside-region interference.

Single-angle pre-scan chord lengths guide X-ray exposure settings, reducing dose while improving scan efficiency and reconstruction accuracy.

Free-breathing 3D radial UTE imaging uses self-navigation and motion-compensated reconstruction to reduce lung MRI artifacts and enable ventilation mapping.

A machine learning model converts PET image sequences into high-quality parametric images, reducing noise while avoiding slow direct reconstruction.

A space-time-coil neural network reconstructs motion-resolved MRI faster by avoiding iterative k-space consistency while reducing aliasing.

A two-stage ML training approach reconstructs low-field MRI images by removing correlated noise and aliasing without clean reference images.

Pre-learned spatial subspaces turn single k-space readouts into real-time 3D MR images, improving temporal resolution for therapy tracking.

Undersampled PROPELLER k-space blades are rebuilt with deep learning to shorten MRI scans while preserving image quality and motion robustness.

Separate B0 and gradient field coefficients correct MRI geometric distortion more completely, especially across larger fields of view.

Classifying PET scattering and non-scattering events with separate correction parameters improves reconstruction accuracy and imaging quality.

A neural calibration model maps point-source simulations to extended-source antenna responses, improving tomographic reconstruction under multiple scattering.

High-absorption regions are detected and excluded from motion-vector calculation to reduce CT image distortion during motion correction.

Confocal height feedback and masked femtosecond ablation speed semiconductor 3D imaging while preserving flat layers and sample integrity.

A neural network reconstructs ultra-fast-pitch helical CT data to suppress motion-related artifacts and restore anatomical detail on single-source scanners.

Conventional PET time resolution limits image detail; a trained neural network upgrades low-resolution histo-images before reconstruction.