18 results about "Voxel intensity" patented technology

Systems or techniques for facilitating learnable visual prompt engineering are provided. In various embodiments, a system can access a medical image and a pre-trained machine learning model that is configured to perform a diagnostic or prognostic inferencing task. In various aspects, the system can apply a pre-processing transformation to one or more pixels or voxels of the medical image, thereby yielding a transformed version of the medical image, wherein the pre-processing transformation can convert an input pixel or voxel intensity value to an output pixel or voxel intensity value via one or more parameters that are iteratively learned. In various instances, the system can perform the diagnostic or prognostic inferencing task, by executing the pre-trained machine learning model on the transformed version of the medical image.

The invention relates to a human brain extraction method and system for enhancing anatomical structure and surface information perception, electronic equipment and a storage medium, and belongs to the technical field of image processing. The anatomical structure sensing module aims to enhance understanding of global brain tissues, and the surface information sensing module emphasizes a local mode of a brain surface so as to solve challenges brought by low-level inter-class similarity of voxel intensity between brain tissues and non-brain tissues; furthermore, a novel local-global feature enhancement module is employed to enhance the association between local surface patterns and global brain tissue, providing a robust local-global representation that facilitates precise brain extraction. According to the method, the anatomical structure and surface information perception of the human brain is enhanced, so that accurate brain extraction is realized.

A computer-implemented method includes obtaining, from a 3D imaging measurement, a 3D image of at least a portion of an electrochemical cell unit; using a trained supervised machine learning engine to associate, based on intensities of voxels of the 3D image, each of said voxels of the 3D image to labels representative of electrochemical cell unit components; calculating, from said segmented 3D image, values of one or more physical property of the electrochemical cell unit; obtaining one or more physical measurements of said one or more physical property; calculating one or more difference between said one or more physical measurements and said calculated values; and if said one or more difference exceeds a threshold, modifying one or more of said 3D image and one or more parameter of the preceding steps.

A computer-implemented method for removing background noise from a medical image comprising voxels, each voxel having a voxel intensity, the method comprising: generating a mask based on the medical image, the mask including a foreground portion designating foreground voxels, a background portion designating background voxels, and a perimeter separating the foreground portion from the background portion; specifying a perimeter portion of the mask, the perimeter portion surrounding the perimeter, a subset of the foreground portion, and a subset of the background portion; specifying a threshold for the perimeter portion to separate voxels based on voxel intensity; filtering the medical image with the mask and the threshold to obtain a filtered image; and displaying the filtered image on a display.

A method for perfusion imaging in a region of interest of an individual, the method comprising a computer-implemented processing method, the processing method comprising: ■ Generating intensity-time curves from pixel / voxel intensities of N blocks of J elementary ultrasound images of the region of interest generated from raw ultrasound data, ■ Identifying a reference intensity-time curve of a predetermined reference zone of the region of interest, ■ Generating at least one perfusion indicator from the reference intensity-time curve.

The invention discloses a super-resolution reconstruction method based on a low-resolution nuclear magnetic resonance image, and belongs to the technical field of medical image processing, and the method comprises the following steps: firstly, inputting a high-resolution nuclear magnetic resonance image into a data preprocessing module for simulating a low-resolution image; secondly, inputting the preprocessed nuclear magnetic resonance image into an interpolation up-sampling module, and adjusting the size and the resolution of the preprocessed nuclear magnetic resonance image to be consistent with those of a high-definition image; and inputting the up-sampled low-resolution image into a three-dimensional convolutional neural network model based on a residual module, and realizing accurate prediction of each voxel intensity of the high-resolution nuclear magnetic resonance image in combination with a feature encoder, a feature decoder and a regression head. The method has the advantages that distribution characteristics of clinical low-definition MRI are better adapted, MRI data with non-fixed axial resolution are effectively processed, and the model can reconstruct high-frequency details of the skull and the face while restoring details of an anatomical structure of a brain region.

X-ray sources emit X-rays into an examination region along different projection views. An X-ray detector array detects the X-rays emitted by the X-ray sources after passing through the examination region. X-ray imaging data are acquired by cycling through the X-ray sources with each step of the cycle including: switching an active X ray source on to emit X-rays and the other X ray sources off to not emit X rays, and acquiring X ray imaging data along the projection view corresponding to the active X-ray source that is switched on. Tomosynthesis image reconstruction is performed on the X ray imaging data to generate at least one volumetric image. A time sequence of spatially aligned images is produced from the time sequence of volumetric images. A perfusion image and / or a ventilation image is generated based on voxel intensity variation over time of the time sequence of spatially aligned images.

The invention discloses a volume electron microscope isotropic reconstruction method based on implicit neural representation, which relates to the technical field of biological image processing, and comprises the following steps of: cascading continuous coordinate position codes with multi-scale features and degradation descriptors, inputting the cascaded codes into an implicit neural representation decoder, and outputting voxel intensity in a continuous space, the reconstruction loss aligned with the training pair and the average pooling re-input item are combined for optimization; the method comprises the following steps: respectively receiving two paths of low-resolution inputs in which an original body is regarded as a first transverse direction and a second transverse direction by a decoder, and uniformly reconstructing along the axial direction to obtain two parts of candidate isotropic body data; a voxel co-location relation is established between the two candidate isotropic volume data, voxel-level arithmetic average is adopted as a fusion mode, corresponding voxels are summed and scaled at equal ratio, and final isotropic volume data is obtained to serve as output of the method; and the method is suitable for multi-source samples and imaging conditions, is suitable for assembly line integrated processing, and is convenient for archiving management.

The embodiment of the invention discloses a three-dimensional ultrasonic imaging method and device, computing equipment and a computer storage medium. The method comprises the following steps: acquiring two-dimensional ultrasonic echo point data and constructing a three-dimensional voxel grid; performing position classification based on the two-dimensional ultrasonic echo point data to obtain at least one position category, and calculating a principal component vector of each position category; determining a principal component vector of each voxel according to the position category to which each voxel belongs; obtaining a candidate echo point of each voxel; for any voxel, identifying effective echo points from the candidate echo points of the voxel through XY plane projection constraint verification and principal component vector constraint verification according to the principal component vector of the voxel; and for any voxel, calculating a voxel intensity value of the voxel according to the intensity value of the effective echo point of the voxel. By adopting the scheme, the edge imaging precision can be improved, the overall imaging quality can be improved, the calculation amount can be reduced, and the three-dimensional imaging efficiency can be improved.

The invention discloses a mild cognitive impairment classification and identification method based on a double-attention mechanism hierarchical CNN model, and the method comprises the following steps: segmenting the inner olfactory cortex, the middle temporal lobe, the hippocampus, the amygdaloid nucleus and the brain island region of a left brain and a right brain according to a DKTatlas cortex segmentation map and an Aseg brain map, and selecting 10 ROI regions; calculating morphological characteristics of each ROI region by using the sMRI image, wherein the morphological characteristics comprise grey matter volume, cortex thickness mean value, cortex thickness standard deviation, surface area and voxel intensity mean value; creating a 3D CNN of a hierarchical ROI of an embedded channel attention and space attention double-attention mechanism and a double-attention mechanism hierarchical CNN model of a multi-layer perceptron classifier based on the ROI; the ROI-based multi-layer perceptron classifier is integrated with the 3D CNN of the double-attention mechanism hierarchical ROI, local depth semantic features are extracted from different ROI regions to generate global depth semantic features, and the global depth semantic features and calculated morphological features of different ROI regions are connected in series and fused for classification and recognition of mild cognitive impairment.

The application discloses a super-resolution reconstruction method based on low-resolution nuclear magnetic resonance images and belongs to the technical field of medical image processing, and comprises the following steps: firstly, high-resolution nuclear magnetic resonance images are input into a data preprocessing module for simulating low-resolution images; secondly, the preprocessed nuclear magnetic resonance images are input into an interpolation upsampling module to adjust the size and resolution of the images to be consistent with those of high-definition images; and then the upsampled low-resolution images are input into a three-dimensional convolutional neural network model based on a residual module, and a feature encoder, a feature decoder and a regression head are combined to realize accurate prediction of the intensity of each voxel of the high-resolution nuclear magnetic resonance images. The application has the beneficial effect that it better adapts to the distribution characteristics of clinical low-definition MRI, effectively processes MRI data with non-fixed axial resolution, and the model can recover the details of the anatomical structure of the brain region and also reconstruct the high-frequency details of the skull and face.

This application discloses a method and apparatus for training a parameter acquisition model based on medical images, comprising: firstly, segmenting the template medical image to obtain various template regions based on the number of voxels and voxel intensity values, and determining the category and physical property parameters of the template regions based on the average voxel intensity and standard deviation of voxel intensity; secondly, acquiring various sample medical images and matching each sample medical image with each template region to determine the sample regions and physical property parameters of each category in the sample medical image; finally, using the sample regions of each category as samples and their physical property parameters as labels, training the parameter acquisition model. This solves the technical problem of low efficiency in training data acquisition and low training efficiency of the parameter acquisition model due to the large amount of human intervention required for training, thereby improving the efficiency of training data acquisition and the training efficiency of the parameter acquisition model.

The invention relates to the technical field of image data processing, and discloses an image segmentation method and system based on multi-modal feature fusion, a terminal and a storage medium, and the method comprises the steps: cutting the voxel intensity of an initial image set to obtain a target image set, and then carrying out the resampling and data enhancement to fuse all modals, storing the file as a preset file; inputting the training image set into a self-defined segmentation network, performing size conversion and feature fusion for multiple times through an encoder and a decoder, and outputting a final image set matched with the segmentation label; and training a segmentation network by using a plurality of loss functions in the final image set output process, predicting the to-be-segmented image by using the trained segmentation network, and outputting an image segmentation result. According to the invention, through a new segmentation network model, the capability of capturing local feature information and the capability of extracting multi-modal and multi-channel 3D medical image features of the model are enhanced.

This invention relates to the field of medical image artificial intelligence analysis technology, specifically to an automated preoperative precise risk stratification system and method for gastrointestinal stromal tumors. It includes: a data preprocessing module for acquiring the original sequences of multi-phase computed tomography (CT) scans of the patient's abdomen, and performing three-dimensional volume data construction, spatial dimension standardization, and voxel intensity normalization to generate a standardized three-dimensional tensor; a deep learning model module for receiving the three-dimensional tensor and, based on a 3D Swin Transformer architecture, outputting classification probabilities corresponding to four risk levels: very low risk, low risk, intermediate risk, and high risk; and an interpretable visualization module for generating heatmaps identifying key decision-making regions of the model using gradient-weighted class activation mapping technology, and overlaying these heatmaps back onto the original CT images for visualization.

This invention discloses a magnetic resonance imaging (MRI) image reconstruction method based on three-dimensional Gaussian volume rendering. The method includes: first, acquiring point clouds of slice data; second, constructing a three-dimensional Gaussian primitive mesh and fitting the point cloud to obtain basic voxel intensities; then, introducing a neural residual field to learn and predict intensity residuals to correct the basic intensities; finally, fusing the basic intensities and residuals to generate high-fidelity final voxel intensities, and reconstructing the slice data into a three-dimensional volume based on this, mapping it to the target image space. By using this invention, high-resolution MRI volume reconstruction tasks can be completed quickly, meeting the time-sensitive requirements of clinical workflows. This invention can be widely applied in the field of image processing.

The present disclosure concerns a free-breathing system for reconstructing a super-resolution volume of a 3D portion of a breathing body, including: a medical imaging device generating at least two snapshots for each plane within a set of K parallel planes transverse to the 3D portion while the body is freely breathing; a contour extraction module for extracting at least part of a contour of 2D cross-sections from the snapshots; an iterative 3D shift estimation module for iteratively estimating a 3D shift in a 3D image space of the extracted contours; and a super-resolution reconstruction module for: repositioning in the image 3D space all snapshots according to the computed 3D shift, sampling the repositioned snapshots with a super-resolution factor, and computing voxel intensities in the sampled image 3D space by averaging voxel values from the snapshots, so as to reconstruct the super-resolution volume of the 3D portion.

The invention relates to a B spline image registration optimization method based on a boundary consistency distance field, and belongs to the technical field of medical image processing methods. The B-spline image registration optimization method based on the boundary consistency distance field is adopted, an input image is preprocessed, and a boundary distance map is constructed; gradient response near the boundary is enhanced through exponential decay transformation; iteratively calculating a deformation field through multi-stage B spline registration; and deformation field information is extracted from the final optimization result, and a deformation field transformation object is generated. According to the medical image registration method, the response away from the boundary region can be effectively suppressed by using the exponential decay transformation, and meanwhile, a relatively large gradient value is reserved in the region close to the boundary, so that the registration precision of the network in the boundary region is improved, the complex pixel intensity and the structure contour information are formed, and the registration arrangement precision and stability are improved.

This invention discloses an isotropic reconstruction method for volumetric electron microscopy based on implicit neural representations, relating to the field of biological image processing technology. The method includes: cascading continuous coordinate position encoding with multi-scale features and degenerate descriptors, inputting this into an implicit neural representation decoder, outputting voxel intensity in continuous space, and jointly optimizing with a reconstruction loss aligned with the training pair and an average pooling back-projection term; receiving two inputs from the decoder, treating the original volume as low-resolution along a first lateral direction and a second lateral direction respectively, and uniformly reconstructing along the axial direction to obtain two candidate isotropic volume data sets; establishing voxel isotope relationships between the two candidate isotropic volume data sets, using voxel-level arithmetic mean as the fusion method, summing corresponding voxels and scaling proportionally to obtain the final isotropic volume data as the method output; facilitating batch quality inspection and delivery, adapting to multi-source samples and imaging conditions, suitable for pipeline integrated processing, and convenient for archiving management.