MDLinxArtificial Intelligence Detects Heart Defects in Newborns

AI-Enhanced Diagnostics: Pioneering the Early Detection of Neonatal Pulmonary Hypertension

In a significant advancement in neonatal care, researchers have collaborated to develop an AI-driven model that aids in the early detection of pulmonary hypertension in newborns. This development promises to enhance diagnostic accuracy and expedite treatment initiation, crucial in improving outcomes for affected infants.

Key Points:

• Pulmonary hypertension in newborns is a critical condition where the arteries to the lungs remain narrowed, impairing oxygenation.
• Early detection and treatment are vital, yet diagnosis is complex, traditionally requiring highly experienced pediatric cardiologists.
• A collaboration between ETH Zurich and KUNO Klinik St. Hedwig has led to the creation of an AI model to assist in diagnosing this condition in infants.
• The AI was trained on video recordings from ultrasound examinations of 192 newborns, encompassing various heart angles and disease severity levels.
• Validation on a separate dataset of 78 infants showed the AI model’s diagnosis accuracy ranged between 80% to 90%, with severity assessment accuracy between 65% to 85%.
• The AI model enhances transparency by highlighting image areas used for its diagnostic decisions, aligning with human experts’ focus points.
• This technology could potentially extend to diagnosing other cardiac conditions and aid regions lacking specialist access.
• The approach underlines the complementary role of AI in healthcare, emphasizing that final diagnostic decisions should remain with human physicians.

“The key to using a machine-learning model in a medical context is not just the prediction accuracy, but also whether humans are able to understand the criteria the model uses to make decisions.”
– Julia Vogt

More on Neonatology