TWM570689U - Automatic detection, identification and report generation system for breast calcification - Google Patents

Abstract

The invention provides an automatic detection, identification and report generation system for breast calcification, comprising: a server having a breast X-ray image file database for storing breast calcification morphology and distribution; a neural network learning module; A breast calcification identification and labeling module; and a data output module.

Description

Automatic detection, identification and report generation system for breast calcification

This new type relates to the field of detection of breast calcification. In particular, the present invention relates to an automatic detection, identification and report generation system for breast calcification.

Breast calcification refers to the calcified white spots seen by mammography. Calcification can occur in cell death, inflamed tissue, scar tissue or cancerous tissue. Calcification is an image presentation, not a specific disease. Breast calcification is mostly a benign disease, but it may also be breast cancer. A quarter of breast cancer patients have been found to have calcification, and their microcalcifications have no specific appearance. They are generally thought to be caused by necrosis of breast cancer cells or their secretions. The more similar the shape of the calcification point is, the closer to the circle, or the higher the benign chance of punctate; the more irregular, linear or branching of the shape, Or a large difference between calcification and calcification (pleomorphic) or microcalcification in the shape of a milk duct, which is more likely to be malignant. In mammography, the part that X-rays can penetrate will appear grayish black, and the part that X-rays cannot penetrate will appear white. We can use mammography to look for tumors and malignant calcification in the breast. If the breast tissue is dense, the mammogram is white, and it is relatively difficult to find the white calcification point. With age, the mammary gland is degraded. In older women, dark adipose tissue will slowly replace the white mammary gland, and calcification will be easier to identify. At present, the identification and malignant judgment of calcification points in mammography images depends on the experience of physicians. Therefore, there is still a need to develop automatic detection of breast calcification, to reduce errors due to differences in physician experience, or to miss due to reading too fast.

The invention provides an automatic detection, identification and report generation system for breast calcification, comprising: a server having a breast X-ray image file database for storing breast calcification morphology and distribution; a neural network learning module, The neural network training model is based on the morphological and distribution characteristics of the breast; a breast calcification identification and labeling module, which automatically displays the position, shape and distribution of the calcification by using the trained model, and inputs the result into the module; And a data output module, the breast calcification position, size, shape and distribution and number output form a test result report. In some embodiments, the server stores breast calcification patterns and distribution information for different types of mammography. These different types of breast calcification patterns and distribution information are collected from multiple patients to establish big data. This information is collected from patient data in each hospital. In general, the more consistent the shape of the calcification particles, the closer to a circle, or the punctate, the higher the benign chance. The more irregular, linear or branching of calcification particles, or the greater the difference between calcification and calcification (pleomorphic) or the shape of a milk duct (casting, casting) The small calcification, the higher the chance of being malignant. The image of benign calcification showed that the shape was mostly round, popcorn-like, and the teacup-shaped distribution was more uniform and symmetric than the calcification point of 1 mm. Malignant calcification images appear to have an irregular size, and are almost all smaller than 0.5 mm (so called microcalcifications), and clustered at 5 mm more than 5 points, such as branching, linear (linear), etc., in the distribution to promote clustering or milk duct formation linear distribution, sometimes combined with the occurrence of density changes in the mammary glands near the calcification point. The American Society of Radiology provides criteria for determining the form and distribution of breast calcification as benign or malignant (http://www.radiologyassistant.nl/en/p4793bfde0ed53/breast-calcifications-differential-diagnosis.html). In some embodiments, a neural network learning module is used to train a neural network based model based on breast calcification morphology and distribution characteristics. The training model first performs machine learning to identify whether it is a breast calcification point, and then through deep neural network learning to learn to find the characteristic value of the breast calcification point image to identify the calcification point. A neural network is a mathematical model or computational model that mimics the biological neural network (the central nervous system of animals, especially the structure and function of the brain, for estimating or approximating functions. By deep neural network learning model In the group, the breast calcification feature is trained to achieve the identification of breast calcification characteristics. In one embodiment, the neural network is a Convolutional Neural Networks (CNN) learning module. CNN is a feedforward neural network. The network, its artificial neurons can respond to a part of the coverage of the surrounding units, for the image processing has excellent performance, using CNN to identify the calcification morphology and distribution images. In some implementations, the breast calcification identification and labeling module The trained model is used to automatically indicate the location, shape and distribution of calcification, and the results are input into the module. In some implementations, the data output module forms the position, size, shape, distribution and number of breast calcification. Test results report. The American Society of Radiology developed a mammography report in order to make the breast examination report consistent. The data interpretation system (BI-RADS) can be used to standardize the interpretation of mammography. The interpretation criteria of the aforementioned BI-RADS system are as follows: BI-RADs(0): Additional image evaluation is required (eg: Breast ultrasound), or need to compare previous mammography results. BI-RADs (1): negative. No other suggestion, no abnormalities or suspicious calcification of the breast. BI-RADs (2): benign findings. X-ray results are normal, but with benign findings such as calcified fibroids, dispersible calcified tissue, signs of fat (eg, oil cysts, lipoma, breast cysts), intramammary lymph nodes, vascular calcification. BI-RADs (3): It may be a benign discovery. It is recommended to conduct a follow-up examination in the short term to ensure that there are no further changes. BI-RADs(4): Suspected that it may be abnormal tissue, it is recommended to perform biopsy. Atypical malignant tissue is found, and BI Compared with the results of RADs(3), such organizations have a higher probability of developing malignancy. BI-RADs(5): Highly suspected of malignant tissue, appropriate action should be taken. BI-RADs(6): Health check and diagnosis of malignancy Weaving. The output of module data report is formed based on the detection result of BI-RADs.

Referring to the first figure, the novel system comprises four components, namely a server 1, a neural network learning module 2, a breast calcification recognition and labeling module 3, and a data output module 4. The server 1 has a database for storing breast calcification patterns and distribution information of different types of mammography. Based on the calcification morphology and distribution characteristics of the breast, the neural network learning module 2 training model is used. Next, the breast calcification identification and labeling module 3 automatically displays the calcification position, shape and distribution using the trained model, and inputs the result into the module. Based on the American Radiological Medicine Association to provide a judging criteria for judging the morphology and distribution of breast calcification, the data output module 4 reports the position, size, shape, distribution and number of breast calcifications to form a test result report.

1‧‧‧Server

2‧‧‧ Neural Network Learning Module

3‧‧‧ Breast calcification identification and labeling module

4‧‧‧ Data Output Module

The first figure is a schematic diagram of the connection of various components of the automatic detection, identification and report generation system for breast calcification.

Claims (5)

An automatic detection, identification and report generation system for breast calcification, comprising: a server having a breast X-ray image file database for storing breast calcification morphology and distribution; a neural network learning module based on breast The calcification morphology and distribution characteristics are based on a neural network training model; a breast calcification identification and labeling module, which uses the trained model to automatically display the calcification position, shape and distribution, and input the result into the module; The output module reports the position, size, shape, distribution and number of breast calcifications to form a test result report. The system of claim 1, wherein the mammography profile of the breast calcification morphology and distribution is collected from a plurality of patients to establish big data. The system of claim 1, wherein the neural network learning module is a convolutional neural network. The system of claim 1, wherein the criteria for determining the morphology and distribution of breast calcification to be benign or malignant are based on the standards of the American Society of Radiological Medicine. The system of claim 7, wherein the interpretation of the results of mammography is standardized using a Mammography Reporting and Data Interpretation System (BI-RADS).