Examples of Artificial Intelligence and Machine Learning in Medicine

Jun 18, 2020

The progression of technology has affected every industry and healthcare is no exception. Doctors will always heavily rely on their years of studying, training, and work experience to diagnose and treat illness, but revolutionary technology has made their job easier and often improves the results in many facets of medicine. While many people are worried that engineered solutions will begin replacing human workers at a large scale, most industries have found that humans and technology working together is the ideal solution. Computers are very good at performing repetitive or computational heavy tasks and they never tire. Humans are better at interacting with other humans and performing complex tasks in which a set of rules is not easily defined. There are many examples of technology helping human doctors, but this blog will highlight the rise of machine learning and artificial intelligence in the field of radiology.

 

Machine learning and artificial intelligence are popular buzzwords in tech news, but what do they actually mean? By most definitions, machine learning is actually part of the larger field of artificial intelligence. Artificial intelligence (AI) is broadly considered technology that interprets inputs to make decisions that will help achieve a goal. The input could be data collected by sensors in the surrounding environment or information entered by a user. The goal of an AI system is set by the creators and could vary from playing chess to providing customer support in a website chat. Machine learning pertains to AI that is able to “learn” as time goes on. Machine learning algorithms have been around for decades, but have become more applicable with the rise of affordable computing power, relevant applications, and developments in theory. There are many variations of machine learning, but a system typically operates by first using a defined set of data to train and then uses the information gathered in training to begin making decisions on real data. For example, a machine learning algorithm designed to recognize stop signs would be trained using a set of pictures with and without stop signs in them. After each decision, the algorithm learns from its successes and mistakes similar to a student studying flashcards! Eventually the algorithm would be used to interpret real time images of the road to detect stop signs.

 

The rise of artificial intelligence and machine learning is affecting many industries and radiology is no exception. Radiology is a field of medicine that has been quick to begin developing artificial intelligence and machine learning as tools for making the review of imaging more efficient. Every patient in the hospital receives imaging, resulting in a significant volume of work for radiologists. In addition, radiologists serve as consultants to many other physicians and healthcare professionals. As a result, radiologists can benefit greatly from the help of technology. Radiologists review the imaging exams of patients to identify pathology, structural abnormalities, and trauma. Image inspection is a great application for artificial intelligence and machine learning because it has consistent input and output. The software can be designed to take relatively consistent images of a certain type and format as input and provide a set of interpreted output details about the image. Image inspection algorithms are used in countless fields such as social media, manufacturing, automation, and media production. In radiology software, images can be inspected pixel by pixel at an extremely fast rate to search for cardiovascular issues, musculoskeletal injuries, cancer, and many more conditions. The machine learning software can be fed a large set of images with data indicating if a certain diagnosis was found from each image. A doctor is typically looking for certain shapes or patterns to make a diagnosis and as the algorithm trains, it will also look for these commonalities.

 

The next logical question is: will artificial intelligence software replace radiologists? The answer is not anytime soon. The artificial intelligence and machine learning fields are moving faster than ever, but they are still moving slowly. Applying the technology to medicine comes with even more obstacles. Large medical image datasets required to train machine learning software are difficult to obtain because of ownership rights and privacy concerns. Current solutions are typically developed with a narrow scope, such as diagnosing one or two related conditions. Increasing the range of possible conditions to find would increase the complexity of the product. However, the field will continue to develop and more products will be approved. As mentioned in the beginning, technology is often best at complementing a human to complete a job. Software is better utilized to help complete repeatable tasks and to supplement the doctor’s work on important tasks. A machine learning algorithm’s diagnosis from a patient image can help the doctor be more confident in their diagnosis and it could also help bring potential problems to attention faster or even notice an issue that is difficult to detect with the human eye. Radiologists are intelligent, highly trained professionals, and it would take a lot of technological advances to replace them.

 

Despite the obstacles, many doctors and engineers are working to bring AI to the field of radiology. Aidoc is a company that provides software used to find patient issues in CT scans, so the doctor can prioritize cases that show trouble signs. Aidoc is based in Tel-Aviv, Israel, but is providing its system to hospitals all over the world. An American company named Viz.ai utilizes artificial intelligence to detect warning signs for strokes in brain images. Arterys touts a cloud native AI platform to provide medical insights. The cloud aspect means that Arterys utilizes reliable, scalable, remote servers that can be accessed anywhere! Arterys’s platform offers apps for cardiovascular, lung, and chest MSK analysis. These are just a few of the companies driving innovation in this field, and many more will likely arise in the coming years.

 

As technology progresses and software continues performing more tasks in the world, many questions form that our society must consider. The regulation of artificial intelligence is a highly debated topic and has increased in popularity with the rise of driverless car technology.  Artificial intelligence “doctors” would open a host of administrative, legal, and philosophical questions. The organization and procedure of medical visits could change dramatically if software is leading your examination or procedure. If a software application can perform a diagnosis, the patient may not even have to leave their home. Lastly, the event of a misdiagnosis or malpractice becomes difficult to solve legally when an algorithm is to blame. The details would likely change for each situation, but perhaps the owner and proprietor of the software would be liable for mistakes. Medical artificial intelligence and machine learning still have a long way to go before they are upending medical practices, but the implications create some potentially difficult scenarios. We must continue to prepare and adapt because technology is not slowing down.

Close

Enjoy 25% off

on your next purchase

When you join our mailing list!