Introduction and AI in diagnosing pneumothorax

Artificial intelligence and particularly machine learning are very important in medical research and applications, especially in diagnosis. This week, Dr. Kang-Yun Lee will introduce the role of AI in lung disease diagnosis.

In this video, we are going to focus on the diagnosis of respiratory medicines for artificial intelligence, especially image studies. Topics of other applications, pulmonary function tests, and diagnosis for other lung diseases such as abstracted and restrictive lung disease, are also included.

In recent years, more research focused on the study of lung cancer images, and lung cancer pathology, including radiomics and radio genomics. We will briefly talk about each of them this week.

Dr. Lee will next talk on pneumothorax studies. Pneumothorax is a disease involving the pleura of the lung when some air leak from the lung itself or from outside the lung, the air get into the pleural space, making the lung collapse, and then, the increased air in the pleural space will increase the pressure of their pleura. In an emergency situation, the tension caused by the air in the lung is too big enough and compresses a patient’s heart causing a shock in the patient’s circulation. This will cause the death of the patient.

Appling AI to detect these conditions can help to prevent the emergency. An automatic method of prioritizing chest X-rays images with pneumothorax may reduce the time to patient treatment and that will be very important for clinical conditions.

Next, Dr. Lee will explain a research project from Dr. Tayder from the University of California. The research team develops software, Automated detection of moderate and large pneumothorax on frontal Chest X-rays using deep convolutional neural network. The software can detect images with severe pneumothorax and be able to report it in time and notify the doctor to treat the patient. According to the data, they developed 16 models, different models have different sensitivity and specificity so when physicians need very sensitive tools or very specific tools, they can use different models to apply to different conditions.

In conclusion, these AI models provide a high specificity screen solution to detect moderate or large pneumothorax and images collected when the human review might be delayed. If the data showed a moderate or large amount of new drugs and the machine could inform the doctor immediately, and the doctor can call the patient back and treat the patient for this pneumothorax. This will prevent very life-threatening conditions. AI in this X-ray reading for pneumothorax is very important for clinical doctors.

Sharing and learning

• How does artificial intelligence use frontal chest X-rays to identify the severity of a patient’s pneumothorax?
• If physicians use frontal chest X-rays to identify the severity of a patient’s pneumothorax, what is the maximum time it might take?