Universal tests, prevention instead of treatment, and big data-based diagnostics: will there be room for human doctors among new technologies?
Nowadays medicine is more than only an interaction between doctors and their patients that results in making a diagnosis or prescribing treatment. At the present time, it aims at preventing and early determining diseases, accurate diagnostics, and selecting the proper therapy that will be suitable for the patients and will not affect their life quality.
Diseases diagnostics have become more sophisticated. Doctors go on relying on their (or their colleagues’) expertise. At the same time, they also access the latest scientific papers, participate in telemedical meetings with experts from other countries, and are actively assisted by artificial intelligence (AI) that can analyze histological sections or MRI scans as any expert can.
However, modern medicine is not limited by this: diagnostics is going to see even more inventions and discoveries. This article focuses on its most perspective tendencies.
1. Preventive medicine: it is better to prevent than to treat
Life expectancy in the modern world is gradually increasing. This is mainly because medicine learned how to cope with many diseases that used to be considered incurable. Even if some diseases cannot be still fully treated, anyway, they changed from fatal to chronic and can be properly managed.
At the same time, longer life expectancy became the reason of other problems, because now people have a risk to have more diseases than their ancestors. Therefore, timely prevention is the best strategy in this case. We consider prevention as the identification of risk factors of certain diseases and early diagnostics.
One of such strategies is screening since it is meant to detect whether people belong to certain risk groups and determine a disease at its early stages. It is known that the sooner the disease is identified, the easier, less expensive, and more effective its treatment is.
On the other hand, patients often hesitate to take a preventive examination or a screening. According to the studies, a lot of people even having some symptoms and understanding they have some health issues refuse from visiting a doctor for a fear to hear bad news. At the same time, healthy patients do not intend to see a therapist either, because they do not have any vivid health concerns.
It is possible to solve this problem, but only partially. The solution is to implement checkup programs that include several examinations to assess the health status. Although this solution is not suitable for everyone, a lot of people would prefer taking a single test to obtain detailed information on their health.
There is another solution that will be highly demanded. This is the introduction of technologies based on AI to clinical practice. Those technologies can process large amounts of data and based on their analysis results or scans, e.g., X-ray images, make conclusions on the status of patients’ health. Besides, the majority of AI-based systems are still being developed and are not available for regular clinical use.
Alternatively, the liquid biopsy that requires a simple blood test rather than collecting tissue samples is no longer thought to be improbable and is widely introduced to clinical practice.
Nowadays, it is possible to identify several unique DNA sequences that are associated with certain cancer types in the patient’s blood. Those sequences act as a sort of tumor marker helping make the correct diagnosis.
The involvement of AI helps considerably simplify and speed up the detection of tumor patterns in test samples. Such approaches have already been adopted and are widely used to diagnose colorectal malignancies, lung cancers, and other cancer pathologies. GRAIL is based on a universal test applying liquid biopsy methods that can identify 50 cancer types.
2. One test for revealing everything
A number of research groups are making an attempt to create something similar not only for detecting cancer but also for making general assessment of health. For instance, SomaLogic, the US company, is working hard to develop a test that can uncover potential health risks and identify those diseases patients have already got. Everything required for this is a blood sample; the thorough analysis can measure about 5,000 various proteins. Using specially developed AI, it is possible to make conclusions on the status of health, as well as risks of certain diseases including type 2 diabetes and cardiovascular disorders.
The prospects of having only one test to address most, if not all, health problems fascinate scientists. A lot of areas of medicine are actively developing this method. Prenatal diagnostics has been demonstrating the most outstanding progress over the past 10-15 years.
The prenatal medicine is actively using ultrasound tests and screenings made during certain weeks of gestation. At the same time non-invasive tests have been under development and are actively applied. Such tests have become extremely popular. They help reduce the number of invasive interventions, since all they require is a blood sample of a pregnant woman. This blood sample is provided to a lab, where the fetus’ DNA sequence taken from the mother’s blood is analyzed.
Non-invasive tests are also applied to identify a lot of possible disorders before the child is born. They can also assess the risks of pathological development by using bioinformatics. Such utmost accuracy has become possible due to machine learning and the possibility to process large amounts of data. The databases that contain multiple patterns make it possible for the software to evaluate the likelihood of certain pathologies.
Modern tests can identify dozens of disorders with high accuracy. In spite of this, diagnostic methods need more improvements. For example, the majority of testing systems can detect the most common trisomies related to 13, 18, and 21 chromosomes, but at the same time, they fail to deliver any tangible results for trisomies that are caused by other chromosomes. In addition, today’s tests are not sufficiently accurate in case of multiple pregnancy. Although it is possible to detect a high risk of pathologies, it is still impossible to define which fetus is exposed.
3. Diagnosed by the artificial intelligence
In terms of modern medical diagnostics, the majority of scientists, doctors, and patients have high hopes for big data analysis, machine learning, and artificial intelligence.
Such opportunities were mentioned for the first time in 1972 when specialists from the Stanford University developed MYCIN, the system that can analyze causes of blood infections and predict the most effective therapy. This system has never been used in practice and remained a generalized concept, it still proved to be extremely effective. In the majority of cases, MYCIN offered a better therapy than highly qualified specialists did. At the same time, it was impossible to apply it clinically because of the unavailability of Internet and high time consumption of the analysis itself. In spite of this, developers of MYICN showed and proved that artificial intelligence could be effectively used for diagnostic purposes.
At the present time, artificial intelligence and machine learning can boast even a broader scope of application: cloud services and supercomputers enable specialists to detect discrepancies, find common patterns, analyze vast arrays of data, as well as make conclusions based on MRI and CT scans, photos of skin neoplasms, and histological samples.
The completion of the analysis takes only a few minutes. Such results cannot be made so quickly even by the best experts. The study that was conducted in 2015 states that the radiologists’ workload is so extensive that they have to interpret MRI or CT scans every 3-4 seconds throughout their 8-hour working day. Such overload inevitably causes human factors mistakes, while computer software is devoid of such constraints, and their use is less expensive than an average day of a medical specialist.
Perspective developments in this area also include the PAIGE technology diagnose cancer. The AI trained on the oncopathology images database can make a faster and more detailed diagnosis than any expert. Besides, it can identify an anomaly that would be missed by even highly skilled specialists.
In 2021, Paige Prostate, one of the systems that applies this technology, became the first digital pathology solution that got an FDA clearance. Its use reduced false positive and false negative results of prostate cancer diagnostics by 24% and 70%, respectively. Paige Breast Lymph Node, another product introduced in 2022, aimed at detecting breast cancer metastases in lymph nodes. This system has not been used clinically yet, but it is going to be introduced for improving diagnostic results.
Exhaled breath biopsy is another emerging trend. The development of treathomics, the scientific discipline analyzing volatile organic compounds present in the exhaled air, enables to effectively and rapidly identify a lot of diseases, including those that are hard for diagnosis. The Owlstone Medical company, the pioneer of this industry, is currently working on biotechnologies that can identify several disorders, e.g. mesothelioma, a rare cancer type progressing for decades without any vivid symptoms.
It is clear that computers cannot replace human specialists. Artificial intelligence can only assist medical workers staying in charge of making the final diagnosis. It is also necessary to keep in mind that a human can detect an image defect and differentiate it from actual pathologies, while artificial intelligence uses its algorithms and, therefore, can misdiagnose.
4. What about telemedicine without human doctors?
Telemedicine is one more trend that has been gaining its popularity since the late 20th century. It is becoming even more popular with the wide spread of the Internet and the 2020 COVID-19 pandemic. Just in a few months the whole world had to self-isolate and medical workers had to cope with the heavy burden of saving patients’ lives without any effective protocols.
The coronavirus has definitely become the most pressing health issue, but other illnesses have not disappeared either. Patients went on getting sick and needed getting a diagnosis or adjusting their therapy. Telemedical consulting (when patients communicate with their doctors on a remote basis by using a messenger or application instead of visiting the clinic) happened to be able to solve a lot of problems, make the doctor and patient interaction simple, and save time for every specialist.
Telemedicine has evident advantages, but at the same time, it has a downside – in most cases, it can be effectively applied to recurring appointments only. It is impossible to carry out initial examinations by some specialists, e.g., ophthalmologists or OB-GYNs in this way.
Moreover, some people find it difficult to trust artificial intelligence in terms of making a diagnosis, regardless of how many large databases were applied to train it. At the same time, Chine introduced special terminals for online consulting. They are developed and launched by Ping An company specializing in digital medical technologies.
Such terminals are a bit bigger than a phone booth and are called “One-minute Clinic”. People using them can get help as soon as possible: artificial intelligence can make online diagnostics and give recommendations that comply with international medical standards. Here patients can even buy medicines that are stored in this terminal by following all the required regulations. If there is no required medicine, it can be easily ordered online and picked up from the nearest drugstore within an hour.
Of course, this technology is an exception rather than a rule – the majority of patients would prefer to talk to a real doctor instead of a virtual assistant, and, in fact, quite few are ready to trust a machine in the issues relating to their health. However, according to experts, telemedicine is going to improve and develop. For example, it is expected that scientists will continue working at developing AI technologies to manage routine tasks, e.g., collecting medical histories or test results for providing already processed information to the doctors for them to spend more time with the patient and select the most efficient therapy.
What to expect?
The demand for new diagnostic technologies will continue growing. Patients require quick and innovative solutions that need minimum time and efforts for getting an accurate diagnosis, and doctors need reliable supporting tools for taking over their routine activities. Machine learning will definitely become more sophisticated, supercomputers are going to process more data, and artificial intelligence is sure to become smarter and make less mistakes. Solutions are going to be more “human-like” without any speed and accuracy delay.
About the Author
Rustam Gilfanov is an IT businessman, an IT company founder, and the LongeVC Fund venture partner.
