1, disease screening:
The steps of lung screening by artificial intelligence are: using image segmentation algorithm to process lung scanning sequence to generate lung area map, and then generating lung image according to lung area map. The lung region image generated by lung segmentation and the nodule labeling information are used to generate the nodule region image, and the lung nodule segmenter based on convolutional neural network is trained.
Then the image is divided into lung nodules, and the suspected lung nodule area is obtained. After the suspected pulmonary nodules are found, the 3D convolutional neural network is used to classify the pulmonary nodules, and the position and confidence of the real pulmonary nodules are obtained.
2, lesion contour:
The delineation of target area and the design of treatment plan occupy a lot of time of oncologists, and the oncologists in primary medical institutions are inexperienced. Radiotherapy is the most mainstream of the three major treatments for tumors (the other two are surgery and chemotherapy). Compared with diagnosis, treatment cuts into the core of medical treatment.
There are about 200 CT images of each tumor patient, and doctors need to mark the position of organs and tumors on each image when sketching. According to the traditional method, the doctor needs 3-5 hours. After finding the location of the tumor, the doctor also needs to design a specific irradiation plan or surgical plan according to the size and shape of the tumor.
It also contains different radiation doses at different locations. The average waiting time for cancer patients is 2-3 weeks. If the doctor spends his time on one patient and the other continues to wait, he may miss the best treatment period. The delineation of the target area and the design of the treatment plan have certain technical content, which requires the experience of doctors, but it contains a lot of repetitive work.
These labor-intensive jobs are the expertise of artificial intelligence, and using AI to do these things will save oncologists a lot of time.
3. Three-dimensional imaging of internal organs:
Three-dimensional imaging of organs is an artificial intelligence to locate and segment the target organs based on medical image data such as MRI and CT, and display the internal situation of patients on the computer. Input the patient's MRI, CT and other image data, and display the patient's internal situation on the computer. Where the probe in the doctor's hand points, the system will update the display in real time, so that the doctor can see the anatomical position of the patient at a glance.
Make the operation faster, more accurate and safer. The real 3D model of the organ can be automatically reconstructed, so that doctors can directly see the anatomical details of the patient's real human structure in the virtual space of augmented reality through special facilities, and can analyze the three-dimensional geometric shapes of organs and lesions in real time through gestures and voice operations, and accurately measure the position, volume, diameter and distance of the target structure.