Reasons to Implement Machine Learning in Healthcare: Case Studies of Those Who Did

AI technologies, particularly ML, are becoming increasingly popular in all spheres of life. Essentially, with the help of ML, AI performs functions that were previously assigned to people but faster, more efficiently, and at a lower cost. According to Refinitiv AI/ML, 46% of organizations use ML technology in several business areas and simultaneously consider it a priority for development.

Machine Learning has a wide range of capabilities:

• recognizing human speech and turning it into understandable text (ASR);
• computer vision technology – recognition of real objects;
• analysis of a large amount of data;
• learning based on the results of human activity (supervised learning) and program learning from its mistakes (reinforcement learning).

Thanks to all of the above, ML is used in various industries, including the financial sector (stock trading, e-commerce), customer support services (chatbots), automotive, cybersecurity, and, of course, healthcare. In all these areas, the latest technology guarantees more accurate results in a shorter time.

Let’s consider how ML can help healthcare.

ML in Healthcare: Opportunities and Prospects

Healthcare professionals need more and more time to process patient information, review patient records, and make diagnoses. The global pandemic was not the last role in this, which caused an uncontrolled increase in the incidence and, accordingly, calls to medical institutions. Scientists from Johns Hopkins University claim that more than 500 million people worldwide have been infected with the SARS-CoV-2 virus so far. An impressive number, isn’t it? But we should not forget about other, no less dangerous diseases, which also require urgent medical attention.

So what to do? After all, human resources are limited by individuals’ physical and mental capabilities – in other words, doctors cannot be at the workplace 24/7, they tend to get sick, and it is vital to rest. And a solution was found. The introduction of ML technologies into healthcare can significantly simplify physicians’ work and allow providing high-quality services to a larger number of patients.

Tasks that Machine Learning helps to solve in medicine

The vast possibilities of artificial intelligence, implemented through machine learning methods, can solve many problems. If we talk about healthcare, these tasks include the following:

• Systematization of input data, ML simplifies diagnosis based on the existing symptoms.
• Grouping data based on their characteristics – ML can group similar medical cases into groups, analyze them, identify patterns and use them for future research work.
• Recommending the right information based on input data analysis – the search for professional information no longer requires long-term active searches from physicians.
• Predicting the course of events, the analysis of past cases allows an app based on ML technology to predict the future development of events in the current situation.
• Detecting abnormalities – with ML, medical software can analyze input data, check if each parameter is normal, and detect abnormal ones among them.
• Automation of activities – some secondary, routine duties (reporting, scheduling patient appointments, recording patients, etc.) distract doctors from their direct duties, although ML-based software can be performed.
• Prioritization – by analyzing previous experience, ML easily arranges information blocks depending on their importance and relevance.

Still, doubting that artificial intelligence can be trusted with the life and health of patients? Then pay attention to a study conducted in Korea; according to its results, ML-based software can predict the outcome of Covid19 with over 90% accuracy.

Opportunities in medicine that open up thanks to ML

Based on what tasks software developed using Machine Learning technologies can solve in the medical field, let’s consider the benefits of machine learning in healthcare.

• Increased time available for patients. Thanks to ML technologies, the doctor can relieve responsibilities for performing secondary tasks, such as maintaining a schedule of patient appointments, filling out documentation, and searching for information.
• Improving the accuracy of diagnosis. Studies published in Nature show that physicians are 71.4% more accurate in their diagnoses. This is up to 77.3% for ML software. The 6% difference is significant, especially for serious diseases when an error in diagnosis can cost a patient’s life.
• Help in choosing the optimal treatment plan. Due to the uniqueness of each case history, all patients require a special approach and an individualized treatment plan. ML can analyze the available information and help make the treatment as effective as possible.
• Decreased decision-making time. In most cases, the medical opinion based on the list of symptoms and laboratory data matches the diagnosis made by the ML software. But the machine will take much less time to do this.
• Reducing the human factor. Even the most highly qualified specialist remains, first of all, a person who tends to make mistakes. Automation of routine procedures, available thanks to ML, helps to eliminate many errors and avoid, for example, overlaps in scheduling patient appointments.
• Processing a large amount of data. Medicine is associated with much patient information, and Machine Learning was just created to analyze large datasets.

According to forecasts, artificial intelligence technologies could cause the loss of 375 million jobs over the next 10 years. But not everything is so simple in the healthcare sector. Doctors note that such widespread use of software based on Machine Learning does not mean that artificial intelligence should completely replace humans in medicine. The new tools are expected to increase the coverage and productivity of the existing healthcare system. This opinion is held by Dr. Tejal Patel, one of the AI/ML research participants.

10 Impressive Machine Learning Use Cases

Clinical Decision Support System (CDSS)

Medical decision support systems are tools based on machine learning technology that helps medical staff make decisions regarding various medical cases. Conclusions regarding certain appointments are based on clinical studies and information about the patient.

One of the first such systems, Leeds Abdominal Pain System, showed excellent results in 1971. It helped establish the correct diagnosis in 91.8% of cases, while doctors were right only in 79.6% of situations.

Optical Character Recognition (OCR) from IBM

The completeness of patient data plays a decisive role in making a correct diagnosis. But doctors often simply do not have time to enter constantly updated information manually. It is where optical character recognition technology can come to the rescue, allowing you to recognize handwritten or printed text and convert it into machine-readable files in bmp, jpeg, pdf, and other formats.

InnerEye by Microsoft

The analog format of medical records, including x-rays, in the recent past has complicated the process of studying, analyzing, and conducting new research. But the digitalization of medicine and the use of ML has made it possible to work with this data more efficiently, to identify similar cases, identify anomalies, and select the optimal treatment.

InnerEye software is an example of machine learning in healthcare that allows you to differentiate healthy and tumor cells and plan the course of radiation therapy needed for cancer patients 13 times faster.

Watson Oncology from IBM

Several severe diseases in a patient often complicate diagnosis and treatment. In such a situation, it is essential to consider the compatibility of prescribed drugs and try to minimize their side effects. ML does a great job with such tasks. Thus, the Watson Oncology system not only selects the optimal course of treatment but also offers several options to choose from.

Remote Patient Monitoring by Somatix