Will artificial intelligence replace your doctor?

In recent times, news about the potential of artificial intelligence (AI) has grown exponentially. The launch of Chat GPT about a year ago and other similar applications based on AI algorithms has brought into the limelight Benefits and risks of using these technologies in different areas of our livesAnd also about the need for its regulation.

One of the areas where artificial intelligence has begun to be used successfully is medicine, for example, as an aid in diagnosis or for the design of new treatments. still, This technology has limitations and its uncontrolled use can pose a risk to patients. Here we review some of the main current and future medical applications of artificial intelligence.

Artificial intelligence is technology that allows computers to mimic some of the abilities of the human brain such as learning, understanding language, and decision making. Thanks to this technology we can make computers learn from experience and act, solving problems and making decisions like a person, even faster and more efficiently. And all this without constant human intervention: An AI can learn on its own and improve over time. It is not surprising that this ability finds application in the medical field.

AI algorithms are capable Process large amounts of data from patients’ electronic medical records. This skill is not only very useful for biomedical research, but can also help in the diagnosis and prevention of diseases. Using clinical data, AI tools can be trained to recognize specific patterns associated with diseases or even discover new patterns.

The type of medical data that can be processed by AI is very diverse. This includes medical images such as X-rays, CT or MRI scans, microscopic images of tissues, photographs of different parts of the skin or the eye. Also included is data from medical records such as diagnoses, treatments, medications or vital signs, and results of laboratory tests such as routine blood and urine tests. Similarly, genetic data can be processed from continuous monitoring devices (such as heart and glucose monitors) or from connected devices such as smart watches that record physical activity, sleep, and other health-related parameters.

endless applications

in medicine, Artificial intelligence can revolutionize the way diseases are diagnosed and treated. Its ability to increase the accuracy and speed of diagnosis, predict the development of diseases, and early detect changes in patients’ health is being studied; Predicting a patient’s response to a specific treatment and thus being able to personalize treatments. AI can also play an important role in accelerating the turnaround time of biomedical and pharmacological research, for example, when applied to the discovery of new therapeutic targets and the development of new drugs. And in hospitals, it can be used to improve the management and administration processes of health systems and as a virtual assistance tool in areas such as telemedicine.

Despite his ability, The use of AI in medicine presents a series of methodological and ethical challenges. Concerns that this technology raises include transparency of algorithms and their biases, data privacy, and equity in access to the technology.

Many artificial intelligence models are difficult to explain, especially those based on so-called deep neural networks. These algorithms lack transparency because We don’t know how they make decisions and this can lead to mistrust among health professionals and patients. To gain acceptance and adoption of artificial intelligence tools, it is important to be able to explain the reasons behind the diagnosis. Ultimately, these devices, like any other medical device, must undergo rigorous clinical trials and demonstrate their effectiveness and safety before they can be widely adopted.

Algorithms that don’t reflect reality

The quality of AI results largely depends on the quality and representativeness of the data with which it has been trained. If certain groups of patients are underrepresented because of their origin, gender, age, or socioeconomic status, these biases will be reflected in the AI’s behavior. For example, If the data comes primarily from white men of a certain age, the model may not apply to other patients., where it will not be as accurate and may lead to misdiagnosis and inappropriate recommendations. This is the case of some algorithms designed to identify skin cancer from the analysis of biopsies, which identify lesions worse in darker-skinned patients.

Algorithms may also be affected by bias if the data set used does not well reflect all the characteristics of the disease and its subtypes. In these cases, the model may not generalize to what happens in the real world and new algorithms trained with more representative data sets will have to be developed. AI often does not understand the full clinical context of the patient: their prior medical history and other relevant factors that doctors take into account to make an accurate diagnosis. And if forms or unusual forms of disease emerge that were not present during training, the algorithm will also make mistakes.

disease diagnosis

Applications of AI are being studied as a diagnostic tool In various fields of medicine such as cardiology, ophthalmology, dermatology, oncology or pathological anatomy. The latter studies the causes, development and effects of diseases based on the structural changes that occur in the cells and tissues of our body, and which are detected, for example, with a microscope in a biopsy. Cytological (cell) or histological (tissue) preparations can be scanned and digitized and, therefore, are susceptible to analysis with AI.

in pathological anatomy AI tools developed to diagnose diseases like hepatitis B And its utility has been explored in oncology to detect certain cancers such as breast, prostate, stomach or colon.

In 2020, researchers in the United Kingdom and the United States, in collaboration with Google, developed an AI system based on Deep Mind algorithms that outperformed expert radiologists in detecting breast cancer from mammograms in the early stages of the disease. Was capable. Treatments may be more effective. That same year, researchers at the University of Pittsburgh developed an algorithm Prostate tumors were identified with high precision from biopsy images.

In the field of ophthalmology, AI has been used to diagnose diabetic retinopathyAn eye complication that is not detected in time can lead to vision loss and blindness in people with diabetes.

In 2018, the US Drug Administration approved the first AI-based autonomous diagnostic tool. This IDx-DR (currently Luminetix Core) is a program Detects the presence of diabetic retinopathy in retinal images, The deep learning algorithm was conceived by Michael Abramoff, an ophthalmologist and artificial intelligence expert at the University of Iowa, who designed it to be used in primary care environments by technicians with minimal training in ophthalmology who scan patients’ retinas. Are responsible for. If the result is positive, the patient is referred to an ophthalmologist’s office for a thorough clinical evaluation, while if it is negative, the test is repeated after a year.

In ophthalmology, images are essential to diagnose diseases and follow their evolution. For this reason, multiple AI applications for diagnosing cataracts, glaucoma and other eye diseases like macular degeneration Age related.

However, not all mountains are oregano. In 2021, researchers at the University of Washington in Seattle examined seven AI algorithms designed to detect diabetic retinopathy and found that, under specific conditions of use, only one passed doctors’ judgments. Such results indicate that There is still much work to be done for deep learning algorithms to diagnose eye diseases with performance on par with clinical experts.

Medicines designed with AI

On average, the development of a new drug takes more than a decade to complete. This time, Artificial intelligence is used to reduce this time and large amounts of moneyOr that laboratories allocate this work. Although it will still take time for them to reach the market, the first drugs developed with AI are already being studied in clinical trials to determine whether they work and are safe.

For every new drug that comes to market, hundreds of molecules must first be evaluated in the laboratory, most of which are unsuccessful. AI can help find the most suitable therapeutic targets in our bodiesDesign drugs that interact with them and determine subgroups of patients who will respond best to a specific drug.

For example, natural language processing is used to extract data from extensive archives of scientific publications and find relationships between them that allow identifying potential targets for treating a disease. and machine learning algorithms are used Analyze massive amounts of chemical data and create models that predict how a certain drug will behave in our bodies. In this way, AI helps to select the best candidates from countless compounds and only the most promising compounds are tested in the laboratory. It is also used to design new molecules or modify existing molecules so that they fit like a glove with their target: for example in the case of antibodies that block proteins responsible for cancer. . All this work, now done by computers, previously took months and years in the laboratory and involved large numbers of researchers.

-Go to Third Millennium Supplement

Sign up and receive the Science newsletter in your email every week