51 results about "Chromosome maps" patented technology

A system and method for computer-assisted karyotyping includes a processor which receives a digitized image of metaphase chromosomes for processing in an image processing module and a classifier module. The image processing module may include a segmenting function for extracting individual chromosome images, a bend correcting function for straightening images of chromosomes that are bent or curved and a feature selection function for distinguishing between chromosome bands. The classifier module, which may be one or more trained kernel-based learning machines, receives the processed image and generates a classification of the image as normal or abnormal.

The invention provides an automatic chromosome karyotype analysis and anomaly detection method. According to the method, an attention mechanism and a convolutional neural network are combined, the target is positioned in two stages, and the area of the target is roughly positioned in the first stage; in the second stage, an attention mechanism is added to the target area in the first stage, deepersemantic features are extracted to predict a mask of the target, meanwhile, the position of the target is roughly positioned by means of category prediction and a detection box regression task, and segmentation is conducted; finally, the trained model is used for segmenting and detecting the chromosome image, chromosome segmentation and anomaly detection can be accurately achieved, and thereforechromosome karyotype analysis automation is achieved. The method provided by the invention can overcome the defects that a traditional chromosome karyotype analysis method mainly depends on genetics to carry out artificial interaction analysis, time is consumed, overlapped chromosomes are not easy to separate, abnormal chromosomes cannot be automatically detected and the like.

The invention discloses a chromosome image instance segmentation method and device based on an improved Mask RCNN. The method comprises the steps of obtaining each first ROI formed by feature information of each scale in a chromosome image and each corresponding first axis alignment bounding box; correcting the first rotary bounding box in each first ROI to obtain each second rotary bounding box to form a rotary bounding box set; performing NMS operation on the second rotary bounding box according to the angle weighted IOU of the second rotary bounding box and the rotary bounding box set to obtain a final axis alignment bounding box of the first ROI; and according to the final ROI axis alignment bounding box and the corresponding feature information, obtaining a second ROI, inputting the second ROI into a Mask branch, and obtaining a chromosome image segmentation mask. Compared with the prior art, the obtained first ROI is subjected to the NMS operation combined with the angle weightedIOU, so that the problems of excessive inhibition of cross overlapping chromosomes, mutual interference of overlapping monomers during mask regression and the like are solved, and the segmentation effect of the chromosome image is improved.

The invention discloses a cross chromosome image instance segmentation method based on chromosome trisection feature point positioning, which comprises the following steps of: dividing an image containing chromosomes into S * S grid units, and allocating chromosome instances to a plurality of grid units according to the positions of chromosome instance trisection feature points, wherein each gridunit is responsible for predicting the instance category of one instance. Each output channel is responsible for predicting an instance mask map of a chromosome object to which the grid cell belongs.Compared with an instance positioning segmentation network which only adopts an instance center point to complete instance segmentation, the method provided by the invention adopts chromosome trisection feature points to complete chromosome instance segmentation, has higher segmentation accuracy for crossed and overlapped chromosomes, solves the problems that the segmentation effect of a chromosome image is still not ideal and the chromosomes with small size are easily undetected. Therefore, the robustness of the chromosome instance segmentation algorithm is greatly improved, and the segmentation speed of chromosome instances is increased.

The invention discloses an automatic identification method and system for a chromosome metaphase splitting phase microscopic image based on a convolutional neural network. The automatic identification method comprises the following steps: acquiring the chromosome metaphasesplitting phase microscopic image; extracting a chromosome image from the chromosome metaphase splitting phase microscopic image; dividing the extracted chromosome image into a first single independent chromosome image and an adhesion chromosome group image; separating the adhesion chromosome group image into a plurality of second single independent chromosome images; and inputting all the first single independent chromosome images and all the second single independent chromosome images into at least one convolutional neural network model to identify a double centromere chromosome image. The double-centromere chromosome image is identified through the convolutional neural network model, so that automation in the identification process of the double-centromere chromosome is realized.

The invention discloses a small-molecule fluorescent probe and an application thereof in chromosome analysis and cytogenetics, and the small-molecule fluorescent probe can be used as a universal toolfor chromosome image segmentation and is used for identifying and distinguishing overlapped, aggregated or contacted chromosomes. The small-molecule fluorescent probe is used as a reference marker forpositioning centromere positions, is used for classifying homologous chromosomes, distinguishing different forms of chromosomes and estimating chromosome arm lengths, is used in cooperation with karyotype analysis for chromosome deletion and polyploid, is used in cooperation with fluorescence in situ hybridization (FISH), and is used for positioning designated genes and judging chromosome abnormalities. The invention provides a rapid and reliable method for auxiliary chromosome analysis, and has great potential in conventional clinical practice.

The invention discloses an overlapping chromosome segmentation method based on a deformable U-shaped convolutional neural network, so as to further improve the accuracy of the existing segmentation technology. An experiment is performed around a human chromosome image of the same region in the metaphase of mitotic metaphase. The method comprises the following steps: 1) combining DAPI and Cy3 images into a grayscale image, and extracting 46 individual chromosomes one by one; 2) respectively carrying out image enhancement processing on the 46 chromosome images and superposing the 46 chromosome images in pairs to construct an overlapped chromosome image library; 3) dividing the overlapped chromosome image library into a training sample set, a verification sample set and a test sample set; and4) constructing a deformable U-shaped network, completing network training through the training sample set, verifying the sample set to evaluate and test the network, and finally performing overlapping chromosome segmentation on the test sample set. According to the method, various characteristics including chromosome morphological changes such as angles and curl states are extracted by utilizinga deformable U-shaped network structure, so that the accuracy of overlapping chromosome segmentation is improved.

The invention discloses an automatic overlapping chromosome segmentation method based on adversarial learning multi-scale features. Challenges of human chromosome analysis are overlapping of automaticchromosome segmentation, which hinders medical diagnosis and biomedical research. Therefore, the invention provides an adversarial multi-scale feature learning framework, adopts a nested U-shaped network as a generator, and aims to explore 'optimal' representation of chromosome images by using multi-scale features. Conditional generative adversarial network (cGAN) adversarial learning are used topromote output distribution to be closer to a gold standard image; a least square GAN target is adopted to improve the training stability of the framework; and the continuity optimization is carriedout by using the Lovasz-Softmax loss, so that better performance is obtained. Experimental results show that the method is superior to other traditional algorithms in subjective visual effect and objective evaluation standard.

The invention provides a chromosome important feature visualization method based on a convolutional neural network. The method comprises the following steps: training a preset convolutional neural network model by using a training data set; inputting the training data set into the trained convolutional neural network model to obtain a classification result, a weight result and a feature result of each chromosome picture; multiplying the feature result and the weight result of each chromosome picture to obtain first importance information of each chromosome picture; averagely mapping the first importance information of each chromosome picture to a chromosome stripe corresponding to the chromosome picture to obtain longitudinal importance information; performing statistical analysis on each chromosome picture according to the longitudinal importance information and the classification result of each chromosome picture, obtaining and displaying the important stripe features of the convolutional neural network model for identifying various chromosomes, and realizing visual display of the chromosome important stripe features of the convolutional neural network for discriminating chromosome categories.

The invention discloses an automatic segmentation method of overlapping and adherent chromosomes based on a full convolution network. The method comprises the following steps: comparing a pure color image and an original image of a manual mark with a full convolution network; comparing a pure color line image and an original image of the manual mark with the original image, generating a first training database and a second training database; reading the original chromosome map; First detecting the chromosome picture by using the first training database, generating a first detecting map and storing the first detecting map; Using a second training database to perform a second detection on a chromosome picture, generating a second detection map and storing the second detection map; performingpixel processing on the first detection map and the second detection map; outline detection is carried out on the processed picture; the chromosome images of all the contour regions are cut out by using the position coordinate data of the detected contour regions, and the overlapping and adhered chromosomes are segmented quickly and effectively by the full convolution network, so that the automatic segmentation of the overlapping and adhered chromosomes can be achieved efficiently, the coverage range is wide, and the workload and difficulty of manual segmentation are reduced.