Rib instance segmentation, counting and positioning method and system

Abstract

The invention provides a rib instance segmentation, counting and positioning method and system, and the method comprises the following steps: carrying out the semantic segmentation of ribs in a chest CT, and obtaining a rib semantic segmentation binary mask; collecting a manual label corresponding to the rib semantic segmentation binary mask to obtain a rib instance segmentation mask; processing the rib semantic segmentation binary mask and the rib instance segmentation mask, and making a sequence segmentation training sample set; establishing a deep convolutional neural network model; training the deep convolutional neural network model by using the training sample set to obtain a deep convolutional neural network capable of predicting the topmost ribs; inputting the rib semantic segmentation binary mask into the trained deep convolutional neural network, obtaining a prediction mask of the topmost layer of ribs from top to bottom in sequence, and further obtaining a predicted rib instance segmentation mask. According to the method, explicit modeling is carried out on the rib counting problem, the ribs are segmented one by one from the lung top to the lung bottom, and rib counting and positioning are achieved.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular, to a rib instance segmentation, counting and positioning method and system, especially a rib instance segmentation, counting and positioning method based on deep learning. Background technique [0002] Diagnosis, description, and reporting of fractures (diseases) appearing in CT are one of the important contents for radiologists to read films. When a fracture (disease) is found, it is necessary to describe the lesion according to its anatomical location for follow-up analysis or reference for other departments. With the popularity of thin-section CT, doctors can find subtle fractures (diseases), but due to the increase in the number of layers, it has become a difficult problem to confirm the location of the lesion, especially the description of the ribs. People usually have 12 pairs of ribs, and each rib has an independent number, which can be divided into the first rib,...

AI Technical Summary

Problems solved by technology

However, this method does not consider the morphological information of the ribs. When the first pair of ribs is not scanned by CT, it will give a wrong count; in the case of severe fractures or segmentation failures, each connected domain no longer corresponds to a rib. Therefore it is difficult to devise reasonable rules for assigning rib numbers to each region

The second is a method based on voxel segmentation. Such methods usually regard rib counting as a segmentation problem, and use a 2D or 3D segmentation model based on deep learning to predict each rib as an independent category. This method can be marked by A large amount of rib counting data, learning from the data to avoid manual design rules, but limited by memory and calculation, the model can only use part of the CT data as input, resulting in inaccurate segmentation due to the lack of sufficient context; at the same time, the segmentation network has A large number of parameters require a large number of CTs and their corresponding counting labels. Considering that the types and locations of fractures are various, and individual differences are large, it is difficult to collect training data with sufficient diversity in practice to ensure the stability of the model. Finally, due to the fact that the segmentation network operates on the original data, it has extremely high computational complexity, and there is a huge resource overhead in actual deployment.

However, this patent document can only obtain the rib semantic segmentation result, and cannot obtain the instance segmentation mask for each rib separation, and still has the defect of inaccurate segmentation

Images