Predicting the future, the big challenge of AI in medicine

Simply put, artificial intelligence (AI) is a branch of knowledge that investigates how to make machines make decisions autonomously, just like we do. This objective, raised in the 1950s, has reached its peak in the last five years.

Currently, AI is capable of simulating human behavior in many fields, many of which are essential to our lives, such as medicine. In this area, it is already fundamentally changing the daily work of specialists: it can perform repetitive tasks that would require too much time and visual fatigue for a medical specialist; Generate medical reports automatically; improving the training of health experts; Provide very accurate tools for diagnosing diseases, and even help in the design of treatments and therapies.

Its applications in health are becoming increasingly numerous and, above all, more ambitious. It is no longer enough to help a doctor diagnose a disease; It seems that test has already been passed. Now a more far-reaching objective is being pursued: predicting the future. Can you predict that an apparently healthy person will suffer from some disease in the near future? Is it possible to know in advance the progress of a patient’s pathology? What if he survives it?

Although they may seem like questions taken from a science fiction movie, many works published in scientific media provide some answers.

Diagnose before symptoms appear

With regard to predicting whether a person will suffer from a disease in the future, researchers from Massachusetts General Hospital and MIT in Boston (USA) have presented an AI system that is capable of predicting whether a currently healthy woman Whether or not you will suffer from breast cancer in the next five years. About 70% accuracy.

The system was trained with thousands of mammograms, identifying patients who developed cancer over that period. They thus demonstrated that the model could detect subtle changes in breast tissue that are precursors to disease and which until now have not been easily detected by medical experts.

Similarly, several researchers led by the Hong Kong University of Science and Technology published in the journal communication therapy (nature) A model that is able to predict whether a person will suffer from Alzheimer’s even before symptoms appear. To do this, they trained the system with thousands of genomic data, and demonstrated that its ability to detect genomic variations associated with neurodegenerative disease was superior to traditional statistical methods.

Machines that predict disease development

The second issue, predicting the development of a disease, may be very relevant for some pathologies. This is the case of glaucoma, a disease that affects the optic nerve and can cause blindness. In order for a patient to receive personalized treatment, it is important to predict how quickly their pathology will progress.

This is the goal of the AI ​​model proposed by the Singapore scientific team. Their project is able to predict changes in a patient’s visual field within a 12-month period from the first visit to the doctor with approximately 80% accuracy. Its parameters were adjusted using medical images, visual fields, and demographic and clinical data of patients who had five seizures over 12 months.

Spanish researchers are also not lagging behind. An international multicenter team led by the Polytechnic University of Madrid presented an artificial intelligence system that automatically identifies, quantifies, and evaluates disease severity and predicts patient mortality and admission to intensive care units. And Covid-19 is able to characterize the pneumonia pattern. (ICU) and require mechanical ventilation with high precision. In this case, the model was trained with computed axial tomography (CAT) images of 103 patients.

survival estimates

The last question is, perhaps, the most delicate: Is an AI program able to reliably predict whether a patient suffering from a serious illness will survive? Well, that’s the aim of some recent work, which uses artificial intelligence to predict whether a sick person can wake up from a cerebral coma or the survival of patients with different types of cancer: larynx, colon and brain tumors. Can guess.

a revolution is in the making

With complete certainty, this new direction in research will bring revolutionary changes in the following areas:

• preventive Medicine. AI can help identify patients at risk of developing a disease long before it manifests.

• Personalized medicine. These systems are able to predict the progression of pathology for a specific patient and thus allow the doctor to design the most appropriate treatment.

• Palliative medicine. Predicting a patient’s survival can help a doctor properly design and implement the necessary treatments.

Other areas of knowledge that should play a relevant role in this revolution are the areas of law and ethics; Establishing appropriate rules and boundaries is essential. Are AI inferences reliable enough for these programs to be incorporated into daily clinical practice? Is it ethical and/or legal to use your predictions? Who is responsible if the system makes a mistake?

We don’t have answers to these questions yet. Only time will tell.