Mature algorithm: lung nodule detection technology enters the era of artificial intelligence

The fatal number of lung cancer is the highest among all cancers, male or female. According to statistics, in 2016, patients who died of lung cancer in the United States accounted for 27% of total cancer deaths. Early screening is an important means of reducing mortality. However, because early lung cancer patients generally lack obvious clinical symptoms and no specific biomarkers, the main method of screening is to detect whether there are suspicious lesions in the lungs by radiography.

The most commonly used chest radiography techniques include x-ray chest X-ray or CT chest examination. Compared to the flat two-dimensional picture of the x-ray chest, CT can provide three-dimensional information of the chest, so its screening accuracy is much higher than x-ray.

Low-dose CT scan is very suitable for annual physical examination because of its fast scanning speed (average operation time is less than 1 minute per test subject), low cost and small radiation dose. The National Lung Screening Trial, a multi-year national lung screening study in the United States, uses low-dose CT to perform chest physical examinations in high-risk populations (aged, smoking history, or family history) and lung cancer deaths within 7 years. The rate is 20% lower than that of people who have a normal chest examination.

Because of the excellent screening results of low-dose CT scans, this screening has gradually become a routine physical examination program in the United States, and it is believed that other countries will gradually promote the application of this method.

A major obstacle to the promotion of this screening work is the enormous workload of CT imaging diagnostics. Early lung cancer is characterized by pulmonary nodules, which are small in size, low in contrast, and high in shape heterogeneity. Therefore, screening is done by imaging experts. However, there are at least 100 CT images of the chest of each subject, and even 600 scans of fine-level scans. Therefore, with the rapid growth of the number of medical examinations, the manual processing method is more and more difficult to perform this task.

Method: three steps

Over the past decade or so, a variety of computer-aided diagnosis (CAD) systems for CT screening of pulmonary nodules have been developed, including representative systems such as ISICAD, SubsolidCAD, LargeCAD, and ETROCAD. These CAD systems usually consist of three steps: 1. Data image preprocessing; 2. Establish a suspected nodule set; 3. Reduce false positives of the suspected set.

The task of step 1 is to standardize the input image, fix the image resolution and layer spacing; divide the lung tissue, cut off other tissue areas; reduce data noise. The task of step 2 is to use a variety of algorithms to pick out as many of the nodule regions as possible in the image. In order to enhance the sensitivity of the algorithm to nodules, this step generally does not impose strict requirements on the false positive rate. The goal of step 3 is to reduce the false positive false positive rate of the system as much as possible in the suspected concentration generated in the previous step. The main methods of each step are described as follows:

Image preprocessing

The pre-processing steps adopted by each CAD system in this step are basically the same, and it includes these tasks:

a. Import the CT data file (usually DICOM file) that meets the requirements into the system, rejecting the data in the wrong format or the broken file in the scanning layer;

b. Convert the data to the standard Heinz unit of the CT signal;

c. Adjust the layer spacing to ensure that the true physical spacing (in millimeters) of each pixel is consistent;

d. Divide the lung area and exclude the non-lung area values ​​to prevent noise signals from being generated in these areas.

After these processes, the system will get a clean set of input data.

Pitaya Powder

Dragon fruit powder is rich in protein vitamins and trace elements such as selenium, calcium, phosphorus, potassium, iron, magnesium and natural pigments such as anthocyanins.

Pitaya Powder,Red Pigment,Pitaya Fruit Powder,Dragon Fruit Powder

Shaanxi Zhongyi Kangjian Biotechnology Co.,Ltd , https://www.zhongyibiology.com

Posted on