Today, Machine Learning (ML) and Artificial Intelligence (AI) have begun to affect doctors, patients, hospitals, and all areas related to health. For example, Artificial Intelligence (AI) and Machine Learning (ML) methods are currently used in computed tomography, diabetic retinopathy analysis, and heart attack risk determination by EKG. Since there is a lot of data in these fields, algorithms can be as successful as expert doctors diagnose.
Artificial Intelligence (AI) studies continue in areas such as, where the patient–doctor relationship is at the forefront, and patient privacy must be ensured.
Doctors spend more time on data entry and desk work than patients. Considering that the world’s population is getting older and the need for doctors is constantly increasing, every second gained will help many people regain their health.
In the case of using Artificial Intelligence (AI) methods in data entry and analysis of results, the physician can be informed. In this way, while helping the physician make the diagnosis, it can be beneficial for the physicians to spend more productive time with the patients.
At the same time Artificial Intelligence (AI) Methods, besides saving time for physicians, can also compile and analyze skipped and wasted data.
About one million children die from pneumonia and other lower respiratory tract infections each year. This is more than deaths from malaria and HIV. Some of these deaths could have been prevented if all children in the world had access to advanced health checkups. Unfortunately, 95% of people living in developing countries do not have access to x–rays, the ideal technology for diagnosing pneumonia.
John Hopkins University researchers may have found a solution to this problem: An intelligent stethoscope that uses Artificial Intelligence (AI) to self–diagnose pneumonia.
Anyone Can Use It Anywhere
A standard stethoscope is an invaluable and low–cost tool for detecting lung conditions. However, this device does not work well if the user does not know what to listen to or if the ambient sound suppresses the patient’s breathing. Taking these limitations into account, the John Hopkins team decided to rethink the stethoscope. First, the team designed a device to ignore outside sounds while recording lung sounds. They then developed an Artificial Intelligence (AI)–powered app that could “listen” to this lung sound for signs of pneumonia.
The John Hopkins team is currently testing prototypes of this stethoscope in Peru, Bangladesh, Malawi, and the USA. Hopes are high for the final version of the device. Researchers say this device will be much cheaper than the current $500 electronic stethoscope on the market.
“We think that this stethoscope, which we have rearranged, with investigations that can be done at the scene by any local paramedic with an inexpensive device, will impact the global health crisis of pneumonia in children. We hope that hundreds of thousands of lives can be saved at the macro level. But at the micro–level, if even a single parent sees their child improve through early detection, all our efforts will be worth it.” he said.
In the line of development that started with games, Artificial Intelligence (AI) has advanced enough to sign revolutionary results today. But, of course, we can predict that shortly; We will consider what they can do today at a beginner level because we can now add the diagnosis of pneumonia among the jobs where Artificial Intelligence (AI) is more successful than humans.
In addition, a group of scientists from the Stanford University Machine Learning (ML) Group managed to train a computer to diagnose pneumonia using Deep Learning techniques.
Although the report’s subject is the diagnosis of pneumonia, the CheXNet algorithm also produced successful results in 14 diseases identified in the ChestX–ray14 dataset. Furthermore, the results revealed that CheXNet outperformed the other two Machine Learning (ML) experiments on 14 pathologies, including atelectasis, cardiomegaly, induration, edema, effusion, emphysema, fibrosis, hernia, infiltration, mass, nodule, pleural swelling, pneumonia, and pneumothorax.
To accurately diagnose pneumonia, the CheXNet neural network needed to learn to accurately detect features in images and link them to natural conditions such as lung air spaces, fluid, large masses, small nodules, and an enlarged heart. The chest X–ray 14 clusters, released earlier this year, plays a vital role in acquiring this skill, as it is more significant than past X–ray collections.
The CheXNet algorithm holds excellent potential for diagnosing pneumonia and other diseases. Researchers say that more than 2 billion X–rays are taken each year, and more than 1 million adults are hospitalized for pneumonia.
Given that approximately 50,000 people die from pneumonia each year, the automation of accurate diagnosis via computer could help many people get treatment before it’s too late.
