Cerrar

The word “diagnosis” is a common denominator among many health professionals. Every day, doctors evaluate symptoms, medical histories, and the results of laboratory tests and procedures such as biopsies or imaging tests to determine whether patients have a condition or disease. This health care process is going through a dramatic shift thanks to a non-human presence: artificial intelligence (AI).

A recent study raised some of the many questions this new challenge elicits in the medical field. Can a computer algorithm be just as or more precise as the human eye in detecting a millimetric tumor on a radiographic image? Can we generate advanced diagnostic models? Can the precision of data development and analysis change patients’ health outcomes? Make them healthier... or even cure them?

Many people are trying to answer these questions with the AI already present in many health systems.

According to the US White House’s analysis on AI, each year, hospitals and health centers produce around 3.6 billion images worldwide from preventive and follow-up exams. The report says that “AI is helping doctors analyze images more quickly and effectively, seeking signs of breast cancer, lung nodules, and many other conditions to reach more people with early detection than has previously been possible. Today, developing new drugs takes years and costs over $2 billion on average. AI is streamlining development with its ability to match drug targets with new molecules that can treat and cure diseases, saving time and money—and translating to cheaper, better care for patients.”

The report stresses that clinician burnout is another major challenge. On average, for every patient they attend, hospital staff must complete over a dozen forms. New AI applications can extract data from patients’ medical records, immediately fill in forms, record notes from sessions, and speed up and improve communications with patients.

During the December AI + HEALTH conference organized by Stanford University, a series of articles were presented that sought to answer key questions on AI in clinical practice and how to incorporate its algorithms as a “noninvasive” way of speeding up and pinpointing a diagnosis.

At the 2023 annual meeting of the American Society of Anesthesiologists, researchers presented an automated pain recognition system using artificial intelligence as an unbiased method to detect pain in patients before, during, and after surgery. Experts said that this system would prevent the prescription and substance abuse that has become a public health crisis in the United States.

It is interesting that this model emphasizes the “impartiality” of AI as a “prejudice-free” system, when the ones who created this intelligence are humans.

Although people are now discussing its ethical implications—in fact, part of the debate is the bias that automatization may have; for example, its inability to operate at the same level in different languages and dialects—AI is already being used in day-to-day health care. For example:

• Mammogram facilities in the United States are offering an “AI-supported reading” option for an additional $40 that the patient must pay out-of-pocket.
• The New York State Office of Aging offers a free artificial intelligence device that provides companionship to older adults after a study showed this software’s ability to reduce feelings of loneliness in seniors by up to 95%. The AI tells jokes, reads stories or novels, organizes the week’s menu, gives reminders for medical appointments, and more.
• At Cedars-Sinai Medical Center in Los Angeles, doctors have used AI to identify early signs of pancreatic cancer, one of the most difficult cancers to diagnose in its early stages. The hospital also uses artificial intelligence to predict common heart conditions, Alzheimer’s diagnosis, and liver disease.

Although we have been debating on the use of AI since the 90s, it is now, as AI is starting to be used in health centers, that discussion is opening up on its consequences on health care, the economy, and, ultimately, public health.

What most health professionals do agree on is that AI engineering and software still need to be perfected and finetuned to guarantee precision and quality. Researchers at Stanford University studied an AI program’s responses in 64 clinical scenarios and found that 6% of answers were hallucinated citations.

The development and implementation of AI in medical diagnosis is still in its early stages and many technical, regulatory, and ethical challenges have yet to be overcome so that this technology can reach its full potential. Patient confidentiality is one of the most critical concerns.

This story was produced using content from original studies or reports, as well as other medical research and health and public health sources cited in links throughout the article.