Image processing device security is provided herein. A method can include assembling, by a first system comprising a processor using a first virtual machine enabled via the first system, raw input data captured by an image capture device from an input image, resulting in assembled input data; generating, by the first system using a second virtual machine that is enabled via the first system and distinct from the first virtual machine, an output image from the assembled input data; reading, by the first system in response to the generating, the output image; and preventing, by the first system, a second system, distinct from the first system, from accessing the output image in response to the reading resulting in execution of unauthorized instructions at the first system.

A blood smear full-view intelligent analysis method, and a blood cell segmentation model and recognition model construction method. The analysis method comprises: collecting a plurality of original blood smear single-view images, establishing an original blood smear single-view image group, and establishing a blood smear full-view image on the basis of the plurality of original blood smear single-view images; constructing an image restoration model on the basis of a first training set and a first verification set; constructing an image segmentation model on the basis of a second training set and a second verification set, obtaining a third training set and a third verification set on the basis of a plurality of segmented individual blood cell images, and constructing an image recognition model; and finally obtaining a blood cell classification result. According to the method, full-view blood cells are analyzed on the basis of an artificial intelligence algorithm, thereby greatly reducing interference of human factors, improving objectivity of an inspection result, and improving blood cell analysis and classification accuracy; recognition and analysis can be realized for picture input meeting requirements, the algorithm robustness and accuracy are higher than those of conventional image recognition algorithms, and the overall time is greatly shortened.

A classifying device for classifying cells in real-time, comprising: as alignment unit configured to align a cell to be classified along the cell's major axis; and a classifying unit configured to classify the aligned cell using a multilayer perceptron, MLP; wherein the MLP classifies the aligned cell based on one or more images of the aligned cell. By executing the classifying device, an improved and efficient cell classification in real-time based on cell images can be provided, while labelling of the cells to be classified can be avoided.

This disclosure describes single-use test cartridges, cell analyzer apparatus, and methods for automatically performing microscopic cell analysis tasks, such as counting blood cells in biological samples. A small unmeasured quantity of a biological sample such as whole blood is placed in the disposable test cartridge which is then inserted into the cell analyzer. The analyzer isolates a precise volume of the biological sample, mixes it with self-contained reagents and transfers the entire volume to an imaging chamber. The geometry of the imaging chamber is chosen to maintain the uniformity of the mixture, and to prevent cells from crowding or clumping, when it is transferred into the imaging chamber. Images of essentially all of the cellular components within the imaging chamber are analyzed to obtain counts per unit volume. The devices, apparatus and methods described may be used to analyze a small quantity of whole blood to obtain counts per unit volume of red blood cells, white blood cells, including sub-groups of white cells, platelets and measurements related to these bodies.

Efficient methods for identifying biomarkers are described. The method may include identifying a tumor area. The method may further include identifying a plurality of regions. The method may also include defining, for each region, a bounding area for the region that encompasses the region. The method may include determining, for each region of a first subset of the plurality of regions, that the region is to be ascribed to the tumor, where the bounding area is fully within the tumor area. The method may further include determining, for each region of a second subset of the plurality of regions, whether to ascribe the region to the tumor based on an intersection of the region and the tumor area. The method may also include accessing a metric characterizing a biological observation and generating a result based on the metrics. The result may be used as a biomarker.

Systems and methods are provided for automatically imaging and analyzing cell samples in an incubator. An actuated microscope operates to generate images of samples within wells of a sample container across days, weeks, or months. A plurality of images is generated for each scan of a particular well, and the images within such a scan are used to image and analysis metabolically active cells in the well. Tins analysis includes generating a “range image” by subtracting the minimum intensity value, across the scan, for each pixel from the maximum intensity value. This range image thus emphasizes cells or portions of cells that exhibit changes in activity over a scan period (e.g., neurons, myocytes, cardiomyocytes) while de-emphasizing regions that exhibit consistently high intensities when images (e.g., regions exhibiting a great deal of autofluorescence unrelated to cell activity).

A method and apparatus of a device that classifies an image is described. In an exemplary embodiment, the method includes tiling at least one region of interest of the input image into a set of tiles. For each tile, the method includes extracting a feature vector of the tile by applying a convolutional neural network, wherein a feature is a local descriptor of the tile; and computing a score of the tile from the extracted feature vector, said tile score being representative of a contribution of the tile into a classification of the input image. The method also includes sorting a set of the tile scores and selecting a subset of the tile scores based on their value and/or their rank in the sorted set. The method also includes applying a classifier to the selected tile scores in order to classify the input image.

An optical method of characterizing an object comprises providing an object to be characterized, the object having at least one nanoscale feature; illuminating the object with coherent plane wave optical radiation having a wavelength larger than the nanoscale feature; capturing a diffraction intensity pattern of the radiation which is scattered by the object; supplying the diffraction intensity pattern to a neural network trained with a training set of diffraction intensity patterns corresponding to other objects with a same nanoscale feature as the object to be characterized, the neural network configured to recover information about the object from the diffraction intensity pattern; and making a characterization of the object based on the recovered information.

Presented herein are systems and methods for determining scores from biomedical images. A computing system may identify a plurality of tiles in a first biomedical image derived from a sample of a subject. Each tile may correspond to features of the sample. The computing system may apply the plurality of tiles to a machine learning (ML) model. The ML model may include: an encoder to generate a plurality of feature vectors based on the plurality of tiles; a clusterer to select a subset from the plurality of feature vectors; and an aggregator to determine a first score indicative of a time to an event for the subject resulting from the features of the sample. The model may be trained in accordance with a loss derived from second scores determined for second biomedical images. The computing system may store an association between the score and the first biomedical image.

Disclosed herein are systems and methods for direct classification of biological datasets. The datasets may include raw mass spectrometry data. Some aspects include training a classifier for direct classification of raw data, and some aspects include applying the classifier.