What if I told you an algorithm could figure out the cause of your lower back pain faster than your doctor? These days, doctors can more easily collect and keep tons of info on our mental health, genetic predisposition, physical ailments, some of the most intimate details about our lives. And the amount of data humans produce, and the amount of data available to healthcare researchers, is growing exponentially. But all this data is insurmountable for humans to process in a timely manner. And when it comes to health, time can be limited. Because of this, artificial intelligence is proving to be increasingly accurate and reliable in detecting a range of illnesses. But is it really a good idea to hand all of our personal health information over to an algorithm? Will robots replace doctors? How can we safely and ethically integrate this new and potentially life-saving technology into our healthcare system? Let's take a look at what's on the horizon.
Healthcare data is poised to grow more than data in any other sector. 30% of the world's data is produced by the healthcare industry. Let's put this into perspective. In 2018, humans produced around 33 zettabytes of data. To get just one zettabyte, you'd have to fill 34.4 billion smartphones to capacity. If you put 34.4 billion Samsung S5s end-to-end, they would circle the Earth almost 122 times. By 2025, our output is expected to increase from 33 zettabytes to 175. So a third of that is healthcare data.
When it comes to health information, medical professionals and researchers believe there's one area of healthcare where this large volume of information can make a big impact. Medical diagnostics. And when it comes to receiving a diagnosis, timeliness is often one of the most important factors impacting rate of survival. One way data can help diagnose patients is with imaging studies. Research hospitals and medical schools around the world are developing sophisticated uses of imaging data to provide warning signs of disease at its earliest stages. With the wealth of medical images available today from ultrasounds, x-rays, MRIs, and CT scans, artificial intelligence has the potential to speed up and improve diagnosis. While every AI model works slightly differently, the most common and successful ones use machine learning. An AI program might analyze thousands of images from an MRI chest scan, for example, of someone with lung cancer, and then compare them to the MRI chest scans of a healthy person who's cancer-free. The model can then use image features it learned to classify the likelihood of lung cancer in new images with incredible accuracy and speed. And improvements in medical imaging over the past few decades also make it a lot easier.
Hmm, if it isn't a piece of candy stuck in Junior's throat. X-ray gives the doctor the inside information, and now he can assure Mother that the candy will soon dissolve and Junior will be as good as ever. Today, our medical images have a much greater level of detail and higher resolution. Thanks to better quality photos, AI programs are able to achieve diagnostic results with more reliability and accuracy. So how good are these programs really? Using these special languages allows the computer to take over much of the painstaking and detailed work that once had to be done by the human programmer.
A 2021 Tulane study found that AI can detect colorectal cancer based on tissue scans better than medical experts themselves. Their program analyzed over 13,000 images of colorectal cancer in over 8,800 people across 13 independent cancer studies in China, Germany, and the United States. While pathologists were on average able to diagnose the disease with 96.9% accuracy, the AI computer program had a 98% accuracy rate. Radiologists are often under pressure to work faster and see more patients. That's where AI diagnostic support comes in handy. It can make their job easier by helping them read scans faster and decide which cases to focus on first. This is especially helpful in situations where time is critical. Additionally, this technology can be a big help in places where there aren't as many resources or specialists available to help out. Think of it as an extra set of eyes to help speed up the process.
It's not just pictures of your insides from radiology. AI can also scan your outsides from your phone. A new smartphone application called Fast AI could soon help diagnose strokes in real time, according to preliminary research. The application uses the smartphone's video function to examine patients' facial features, arm movement, and speech patterns. Early findings suggest that the app could be as accurate at diagnosing stroke as a neurologist and would save a critical amount of time for both patients and doctors.
Artificial intelligence can also aid doctors in caring for those with chronic diseases by simply analyzing health history data to identify high-risk patients. Techniques like machine learning algorithms and neural networks can track and diagnose chronic diseases such as chronic obstructive pulmonary disease and chronic kidney disease by finding patterns in medical data. Basically, the AI keeps track of your health and looks for patterns and alerts your doctor if something seems off. AI-supported medical diagnostics could not only improve patient outcomes, but also save money. One study by Clare Medical found that earlier diagnosis and treatment of cancers could reduce hospital admissions and the associated costs by more than 50%.
AI experts say that it's ethics, not the technology itself, that's one of the main problems for using AI in diagnosing illnesses. In a debate held in November 2020 about AI in healthcare, Jim Stoltz, who starts tech companies, and Nell Watson, a machine intelligence engineer, talked about this issue. A major takeaway was that we need to make sure medical information is gathered in a safe, dependable, honest, clear, and well-regulated way. Hospitals and companies that do medical testing have a lot of pictures and records about people's health because people might not trust AI to help them if they think their privacy and safety could be at risk. And there are actual security issues when it comes to patient information. Back in 2019, ProPublica found 187 servers in the United States storing patient x-rays and MRI data that were unprotected by passwords or basic security precautions. Jackie Singh, a cybersecurity researcher, told ProPublica, it's not even hacking, it's walking into an open door.
Anonymizing images is one technique that's becoming an essential practice. Using a pseudonym or a tracking number instead of a patient's name can protect their identity while giving researchers the ability to analyze and extract valuable insights from medical images. But medical images are not the only thing medical researchers are training algorithms on. AI diagnostic results improve when they learn other data points like age and health-related behaviors such as smoking. Providing access to all this patient data makes privacy a particular concern. According to AI legal experts, there are a few ways to protect patient privacy. Like we mentioned earlier with images, data can be de-identified before being used. Companies should also do a thorough check of anyone they share patient data with to limit legal and monetary consequences. Finally, companies should use security measures like enhanced monogamy and data encryption to keep patient data safe and build trust in the technology.
We care about keeping our health information safe and making sure we get good healthcare. At the same time, if we can use AI to save lives, we might need to share some of our information with healthcare companies. Soon, AI might be a standard part of a doctor's diagnostic toolkit, but it will take some time to implement. It's important to think carefully about the benefits and risks of using AI in healthcare and to make sure we're using it in a safe and responsible way. By doing so, we can work towards a future where AI and human doctors work together to provide the best possible care for everyone. Would you want AI helping your doctor or would you prefer to keep things the old-fashioned way? Let us know in the comments.