X-ray Image Analysis

It is estimated that 8 million people in the world will suffer from tuberculosis every year. If these patients are not properly treated and treated, there will be 2 million deaths from tuberculosis or its recurring diseases. X-Ray image analysis is one of the important methods for the detection and prevention of tuberculosis.

Our X-ray image analysis system for image recognition uses deep learning technology to simulate human neural networks to learn and analyze these data, and identify some recurring or similar characteristics in diseases, summarize rules, and combine the experience of medical customers. Existing biological information can predict the possible mutations and other symptoms of diseases in the future. This brings the safety of life to the majority of patients and improves the efficiency and accuracy of doctors' diagnoses.

Model Components

Classification Model
Our X-ray image analysis system consists of a classification model based on deep learning (Figure 1), which mainly performs two analyses. 1. Perform positioning analysis for the location of the disease. 2. To classify the pathological symptoms of the disease.

Figure 1. Classification Model

According to the images, the pathological types of tuberculosis are mainly divided into the following 4 categories: A) cavities, B) consolidation, C) effusion, D) miliary. After sample labeling and learning and training, our classification model can accurately find the pathological location of the sample and mark those pathological diagnosis classifications (Figure 2). The average processing time of artificial intelligence diagnosis for each photo is 53.4 milliseconds.

Figure 2. A Cavity with surrounding consolidations in the right upper

Figure 2. B Consolidations in the left upper lobe

Figure 2. C Small left-sided pleural effusion

Figure 2. D Extensive bilateral miliary tuberculosis

Advantages

Accurate
Our X-ray image analysis are trained from nearly 200 patients with co-infection of human immunodeficiency virus (HIV) and tuberculosis. A large amount of training data and robust algorithms make our model's accuracy rate as high as 95%, which is significantly higher than manual recognition accuracy.

Efficient
Supported by high-performance servers and robust algorithms, the X-ray image analysis can diagnose a large number of X-ray image in a short period of time and keep working 24/7, which can greatly reduce the workload of doctors and improve their diagnosis effectiveness.

Applicable
Our X-ray image analysis model can accurately and quickly complete image screening, and make decision assistance based on AI screening results. According to the test results, our X-ray image analysis model can help junior radiologists to improve the quality of diagnosis, so that their diagnostic level can be improved to be equivalent to that of senior radiologists in a short time.