2 min read
The idea of artificial intelligence captured my attention in the early 2000's when Will Smith graced us with the cinematic thriller "I, Robot." The movie is set in 2035, where humans are building smart robots using AI, and Smith's character has to fight them from taking over, and blah blah blah. You've seen it 100 times in a million movies.
Enter real life... artificial intelligence captured the world when the IBM super computer, known as Watson, won the TV show Jeopardy!, beating the reigning Jeopardy! champion, Ken Jennings. Now, the supercomputer, branded for healthcare as WATSON HEALTH AI, is found in at least 16 cancer treatment centers across the United States, helping to diagnose and treat cancer patients.
Google has also entered the AI realm. Great Britain’s National Health Service is working in partnership with Google’s DeepMind (DeepMind Health), whose goal is to create new apps for healthcare professionals, assisting them with patient emergency alerts and risks associated with potential treatment options. Already, the system is being loaded with over one million digital eye scans, and it has demonstrated the capability to identify sight-destroying conditions that equal the accuracy of ophthalmologists. Quite the lab assistant they are building.
Deep Learning is Invaluable to Medicine’s Use of AI
So, what is deep learning, and why is it so important to AI in medicine? Deep learning is a sub-field of machine learning. It is primarily concerned with algorithms that are influenced by the structure and the function of the human brain. These algorithms are called artificial neural networks.
Using Deep Learning, scientists can, using the same amount of data for forming older learning algorithms, create Deep Learning algorithms that perform with greater accuracy than their older versions. The neural networks have another important characteristic: their performance improves with more data as they create bigger models.
Why is This Important?
In March 2017, Google published a paper titled "Detecting Cancer Metastases on Gigapixel Pathology Images." This paper shows that by using GoogleNet AI, the algorithm was loaded with thousands of medical images given by a university in Holland. The program correctly identified malignant tumors in breast cancer images with an accuracy rate of 89 percent. Here's how the Google researchers explained these results:
“Pathologists are responsible for reviewing all the biological tissues visible on a slide. However, there can be many slides per patient, each of which is 10+ gigapixels when digitized at 40X magnification. Imagine having to go through a thousand 10-megapixel (MP) photos, and having to be responsible for every pixel. Needless to say, this is a lot of data to cover, and often time is limited.”
The program is blindingly fast and accurate, and its ability to report quickly and accurately helps pathologists finalize reports and transmit them to ordering physicians.
Some may think pathology is a static science; some are wrong. Pathology has made great strides in the last five decades. With advancements in technology like AI, digital pathology, speech recognition, and smart reporting, those in the pathology field will be able to avoid loss of efficiency due to budget cuts and shortage of support staff. This will keep pathologists productive in diagnosing illnesses quickly and accurately so that disease-ending treatment can be started sooner, and more lives will be saved.