Artificial intelligence (AI) in healthcare is only getting started. These are five areas where AI shows great promise for future developments.
Accelerating the Development of Novel Drugs: Finding a molecular target, creating a medication to target it, and testing the medication for efficacy and safety are the steps in the drug discovery process. This calls for the in-lab screening of a sizable number of possible chemicals. Nine out of ten drugs that are selected for human clinical trials eventually fail. The procedure can cost billions and take ten to fifteen years.
Drug discovery is now being accelerated by AI. The first Phase I human trial for an anti-fibrotic medication for idiopathic pulmonary fibrosis, a condition that stiffens the lungs and makes breathing difficult, was started in February 2022 by Insilico Medicine. The chemical targets a recently discovered biological mechanism that was also found by AI.
Compared to conventional drug discovery, the 30-month procedure was far faster. Although encouraging, the Food and Drug Administration has not yet authorized any AI-discovered medications.
Assisting Physicians with Complicated Diagnoses: Many diagnosis in clinical medicine, like skin infections, are very specific. However, certain cases baffle medical professionals. Researchers examined seven complex cases in a JAMA study to see if ChatGPT-4, an AI platform, could diagnose the condition based solely on the test data and case narrative. In 39% of cases, the AI properly identified the case, and in 64% of cases, it indicated the correct diagnosis as a possibility but not the top diagnosis. Patients might not find comfort in those numbers. However, it demonstrates to medical professionals the ability of AI to copilot confusing instances and support difficult medical decision-making.
Improving Diagnostic Image Interpretation: Sometimes AI sees things that human radiologists are unable to. An AI model was used in a Nature Medicine article to screen for lung cancer using CT scans. High false-positive rates—those receiving biopsies that turn out to be normal—and high false-negative rates—those whose CT scans appear normal but really contain cancer—are frequent issues with lung CTs. Not only did the AI model outperform radiologists in terms of early cancer detection, it was also incredibly accurate.
A study published in the Lancet Digital Health evaluated the accuracy of AI and human radiologists in detecting breast cancer through mammography. 269—or 0.5%—of the almost 55,000 women between the ages of 40 and 74 who had mammography received a breast cancer diagnosis. A radiologist and AI jointly read the mammograms, and 261 cases were found. Two doctors discovered 250 cases when they manually reviewed the mammograms. Just the AI found 246 cases. The study found that, at least for the time being, mammography reading still requires human radiologists.
Letting Physicians Concentrate on Patients: Some physicians take notes during clinical visits with patients and enter them into electronic health records. Patients may find it difficult to connect with their doctors as a result of this distraction. Alternatively, following their shift, doctors can compose their notes at home.
A new technology called ambient AI has the ability to record doctor-patient conversations and generate preliminary notes. This facilitates a more meaningful interaction between the doctor and patient by freeing them up to chat instead of write notes. Additionally, doctors spend less time writing notes. The first 10 weeks of ambient AI use at The Permanente Medical Group in California were examined in a study published in NEJM Catalyst. During this time, 3,442 physicians from a variety of specialties and locales employed the technology in 303,266 visits. For a new technology, this adoption rate is exceptionally high.
Allowing Patients to Feel Included: When people visit a doctor, they want to “feel heard.” A startling study published in the Proceedings of the National Academy of Sciences revealed that reading messages produced by AI made people feel more heard than when they read messages created by humans. This is as a result of AI’s improved ability to recognize emotions and offer assistance. In comparison to humans, the AI provided fewer useful recommendations, which may lessen people’s sense of being heard when recommendations get overly specific.
This feature demonstrates how AI may enhance patient communication. This might be used for a variety of AI use cases, such as creating more compassionate discharge instructions and interacting with patients via AI chatbots. People felt less heard, though, when they realized their words were generated by AI. This suggests that AI would work best as a copilot to assist doctors in having compassionate and understanding conversations with their patients.
Over the past few years, there has been a proliferation of new healthcare AI startups. Some might turn into unicorns. Some won’t make it. It’s unknown which use cases will be implemented in the end, but one thing is certain: as AI develops and becomes more integrated into healthcare, it will have a revolutionary impact.
