What Is Machine Learning? Are We Ready for Artificial Intelligence in Healthcare?

With the invention of computers and their becoming indispensable in our lives, people began asking: How intelligent can a computer be? Is artificial intelligence possible? How can we enable computers to learn by themselves? This led to the question: What Does Learning Mean for a Machine? In 1959, Arthur Samuel said that computers could acquire learning abilities independently of programming and developed a checkers game. What distinguished it was its ability to improve by learning from previous moves.

What Is Machine Learning?

Machine learning brings together disciplines such as statistics, information theory, algorithms and functional analysis, enabling computers or developed AI devices to draw inferences from existing data. Today it is used in areas such as speech recognition, facial recognition, search analyses in autonomous vehicles, data mining and medical applications.

What Stage Has Machine Learning Reached?

In the Gartner research group’s analyses of trending technologies, the hype cycle for emerging technologies shows machine learning at “The peak of expectations.” We see that 3-dimensional printers, once at this point, are now easily accessible and developing rapidly. Similarly, we will encounter machine learning more often in the near future.

We are all now familiar with data focused on healthcare transformation, ranging from personal genomics and biosensors to electronic health records (EHR) and health insurance claims, forming big data (big data). By using these data, machine learning can offer analyses and make supportive predictions. In healthcare, these are called decision support systems . However, making predictions in healthcare is not new (for example, APACHE classification for predicting intensive care patients’ course and mortality risk, or the ASCVD risk calculator for predicting the need for cholesterol-lowering drugs). Using machine learning and clinical data, predictive models can be created for thousands of similar clinical problems.

Artificial Intelligence Does Not Have an Easy Task in Healthcare

The project aimed at predicting and monitoring epidemics worldwide, Google Flu , was found to have failed mainly because of disappointments in data analysis. There are indeed serious problems with transferring and processing health data today. For example, while traditional patient registration methods generated comprehensive data, current patient data sources are more restricted (difficulty conforming to certain patterns).

Machine learning applications in healthcare range from diagnosis and personalized treatment to drug discovery and radiology. The greatest obstacle to its development is data standardization and management. According to the hype cycle, this technology trend is at the peak of expectations; to move faster and more effectively from phase 3, the “trough of disillusionment,” to phase 4, the “slope of enlightenment,” past mistakes must be recognized. For healthcare, electronic health record applications need to become more efficient and effective for machine learning.