5 Questions, 5 Answers on Medical AI

Avi Tsur, Head of the High-Risk Pregnancy Unit at Sheba Medical Center, explained that AI is already being used across the entire pregnancy process. In early pregnancy, it predicts risks of preeclampsia, fetal growth restriction, and preterm birth, and during hospitalization, it helps patients understand their records and communicate with medical staff.

• At the delivery stage, it is used to predict complications, including the choice between vaginal delivery and C-section.
• Postpartum, its scope can expand to predicting complications and long-term health outcomes.

In his view, the biggest change is in clinical reasoning. Medical staff will move away from looking at just two or three key variables and instead read hundreds of risk factors, subtle patient complaints, and imaging and AI-based examination data together. AI will take on operational tasks, surgical assistance, and clinical decision-making, but explaining a patient's condition and reading their emotions will remain a human role for the time being.

The remaining bottlenecks are integration with electronic medical records and workflows, and transparency. He pointed out that AI fails to properly present the grounds and references for its recommendations, which conflicts with how doctors think, and that AI hallucination is especially notable in reference accuracy. Ultimately, even as technical tasks are replaced, empathy—explaining the situation to patients and offering comfort—will remain the domain humans hold onto the longest.