In the early 1970s, scientists from Stanford University developed MYCIN, an AI-based healthcare solution. A smart algorithm asked a doctor yes/no questions. Based on symptoms, it generated a list of possible bacteria causing serious illnesses. AI offered an optimal treatment regimen with a dosage adjusted to a patient’s body weight. It was one of the first medical AI platforms; however, it was not used in practice. A committee assessed the correctness of treatment at 65% and there were ethical and legal obstacles to introducing AI. Half a century later, people learned to overcome the barriers associated with this technology. Thanks to technical progress, smart algorithms have proven efficient in saving people. We’ll consider seven examples of that in more detail.
Prediction and prevention of heart disease
According to the WHO data for 2020, coronary heart disease is still the leading cause of death (16% of deaths recorded worldwide each year). Heart disease treatment and prevention account for 7%-21% of national healthcare spending. This is about $219 billion a year for the US.
Researchers hope that AI can make a difference and help prevent the onset and progression of heart disease. A Californian company has developed a non-invasive method for predicting and diagnosing coronary heart disease. HeartFlow is an AI-based healthcare project. Doctors usually need to do stress tests (SPECT and stress echo) and use coronary computed tomography angiography (CCTA) to see how the heart is working. But this does not always give accurate results.
After a patient passes the HeartFlow analysis, they get a 3D model of their coronary arteries in their own unique color. The model is a combination of anatomical and physiological information about the blood flow and vascular plaques. Doctors receive more details about the heart’s state than when doing conventional tests. Thus, they can prescribe the best treatment. The visual model obtained during the analysis also helps to assess the risk of developing cardiovascular disease.
It took five years of detailed research and more than 100,000 items of patient data to build this medical AI platform. As a result, the algorithm has achieved an extremely high level of accuracy. It helps to detect the disease two times more often than using other methods.
Early diagnosis of neurological diseases
Alzheimer’s disease is considered the most dangerous neurological disorder. It ranks 7th in the number of deaths globally. But this statistic does not diminish the severity of other neurological diseases, such as ALS, multiple sclerosis, Parkinson’s disease, epilepsy, and so on. They are extremely devastating.
To detect early signs and slow down deterioration processes, a British startup created an AI platform for diagnosing neurological diseases. The team taught a smart algorithm to scan medical images (MRI, CT, X-ray, PET) and other non-visual data sources (EHR, information from IoT devices, and genome storage). It processes the information received, diagnoses neurological disorders, and predicts the rate of deterioration of memory and cognitive functions.
With this AI-based healthcare solution, doctors can detect early signs of neurological disorders. The system automatically uploads the received data to hospital PACS and EMR. This way, healthcare professionals can control many incurable neurological diseases and delay cognitive decline.
Given that AI can process and structure a lot of data coming from different sources, this technology can be used in any business related to prediction.
MGH and Broad researchers developed an AI model that processes gigantic data and predicts the statistics of the spread of coronavirus strains. So, a smart algorithm determined that the Omicron’s BA.2 subvariant would be more contagious than BA.1, BA.4, or BA.5. Smart prediction tells people how the virus will adapt and mutate. This information can be used to prepare a response to COVID-19.
Scientists from the University of Glasgow used AI to predict the transmission of animal viruses to humans. In theory, approximately half of the 1.67 million viruses identified in animals and birds can be transmitted to humans. AI analyzes the genomes of viruses to determine which ones can cause human infection. The researchers plan to improve the algorithm to detect their virulence and the ability to transmit between people.
AI-powered robotic surgery
Without AI, it would be impossible to implement another idea and build robotic surgeons. A standard surgical robot includes a camera, mechanical arms with surgical instruments, hardware for a human surgeon, and control software. An AI-powered surgical robot by Andersen will soon operate on people on its own, under human supervision.
The first steps towards autonomy were taken in the winter of 2022. A robot successfully performed laparoscopic surgery on the intestines of a pig without medical supervision. According to engineers at Johns Hopkins University, the machine carried out the procedure better than a human surgeon could do. This operation is considered one of the most difficult and requires high precision and attentiveness. After all, an incorrectly applied seam can cause complications in the future.
Perhaps in the coming years, people will delegate most types of operations to robots. Such a move will solve the problem of a shortage of doctors and waiting lists for operations. These advantages are driving the growth of the global market for surgical robots. In 2021, it was estimated at 3.6 billion US dollars. In 2022-2030, it will increase by 19.3% annually.
Prompt classification of tumors in radiology
The ability of AI to analyze medical images can be used not only for diagnosing diseases, but also for narrower tasks. For example, to classify tumors found in humans. Selecting the appropriate treatment depends on the accurate detection of the tumor type.
According to the 2020 NCI study, a smart algorithm can identify a brain tumor type three times faster than traditional methods. By implementing AI-based analytics, scientists enhanced the capabilities of SRH, a specialized form of microscopy. SRH visualizes tissue samples, and a smart algorithm analyzes images to detect tumors and non-tumor tissues.
AI was trained on 2.5 million tumor images from 415 patients. As a result, the algorithm diagnoses neoplasms in the brain with an accuracy of 94.6%. For comparison, a standard pathologicoanatomic analysis has an accuracy of 93.9%. Scientists suggest that pathologists collaborate with AI to cross-check each other and improve diagnostic accuracy.
AI-based healthcare project for laboratory testing
At the beginning of the pandemic, laboratories were under enormous pressure. Between February 2020 and January 2022, over 910 million COVID-19 tests were performed in the US alone. Laboratories had to hire more staff to cope with numerous orders and work around the clock.
European developers used AI to solve these problems and created an electronic platform for analyzing the results of PCR tests. The system downloads raw data from laboratory machines and processes it thanks to a smart algorithm. Together with generated reports, results are displayed on a dashboard as graphs or tables.
All a laboratory assistant needs to do is review the results and approve them. If there is a reason for doubt, the specialist rechecks the test manually. The algorithm has been trained on tens of thousands of tests; therefore, inaccuracies are rather exceptions. This medical AI platform analyzes up to 96 tests per minute, enabling labs to increase throughput without sacrificing quality.
AI for early detection and treatment of cancer
According to the WHO, cancer is still the leading cause of death. Every sixth inhabitant of the planet dies from this disease every year. The organization also predicts that the number of new cancer cases will increase by 70% within two decades. Therefore, doctors and engineers are working hard to develop methods for diagnosing, preventing, and treating cancer. AI is one of the extremely promising technologies for these purposes.
For example, Cyrcadia created an intelligent platform for the early detection of breast cancer. This type is one of the most common: about 2 million cases are recorded annually. A woman puts on breast pads recording circadian temperature changes in her chest and transmits data via smartphone to the laboratory platform. AI analyzes the information and reports the results. If necessary, the result is copied for doctors, insurers, or family members. The platform makes it possible to detect cancer earlier, which increases the woman’s chances of recovery. The doctor contacts the patient to prescribe treatment.
Another startup, SkinVision, uses AI to evaluate photos of skin blemishes and diagnose skin cancer. The algorithm has been trained on 2.9 million images and is familiar with risk labels issued by dermatologists. This allows AI to make predictions with 95% accuracy while a patient stays at home and uses a camera phone.
We have mentioned only a few AI-based healthcare projects as an example of the massive potential of this technology. What seemed science fiction once is now a useful tool for healthcare, banking, finance, and any other area working with data. By analyzing data streams, smart algorithms offer the best solutions to problems and effective options for diagnosing and treating diseases. We are confident that there will be more ways to apply AI soon. There will be more medical AI platforms on the market.