In a fusion of 19th-century medical technology and 21st-century artificial intelligence, cardiologists have unlocked a new method to foresee patient survival rates following heart surgery.
Researchers from US hospitals – Cedars-Sinai, Stanford University, and Columbia University – extensively analysed nearly 46,000 patients’ electrocardiograms (ECGs), a tool dating back to 1895.
By leveraging state-of-the-art AI algorithms, they discovered a significant advancement in predicting post-operative mortality rates.
The study, detailed in The Lancet Digital Health journal, revealed AI’s striking 83% accuracy in forecasting which patients would survive 30 days post-surgery.
This triumphed over the conventional method, the Revised Cardiac Risk Index, which demonstrated 67% accuracy in similar predictions.
Dr Da Ouyang, a cardiologist from Cedars-Sinai's Smidt Heart Institute and a study co-author, highlighted the algorithm’s potential impact.
‘This is the first electrocardiogram-based AI algorithm that predicts postoperative mortality,’ Dr Ouyang said, emphasising the pivotal role of AI in guiding surgical decisions.
The AI algorithm effectively identified high-risk individuals by analysing pre-surgical ECG data against patients’ 30-day post-surgery outcomes.
Those identified as high-risk by the AI exhibited a staggering nine-fold increased risk of mortality within the first month after surgery.
Expressing the inadequacy of current clinical risk prediction tools, Dr Ouyang underlined the AI model’s potential to discern the ideal candidates for surgery.
‘This AI model could be used to determine exactly which patients should undergo an intervention and which patients might be too sick,’ he said.
The authors write: ’In summary, our findings demonstrate how a novel deep-learning algorithm, applied to a single preoperative ECG, can improve discrimination of postoperative adverse outcomes while running efficiently on a standard clinical workstation.
‘Recognising that clinicians have limited time for making clinical assessments and decisions around potential post-procedural outcomes, conventional risk calculators using easily accessible information have been recommended by professional society practice guidelines to aid in perioperative risk stratification. The opportunity to implement potentially more informative and easier-to-use prediction algorithms that integrate with existing clinical workflows offers a potential path towards improving postoperative outcomes. These promising results warrant further studies to establish the prospective validity of deep-learning algorithms for prognosticating post-procedural risk.’
Researchers are exploring the possibility of integrating this AI technology into accessible platforms.
This move could empower medical practitioners worldwide by providing easily accessible predictive tools for assessing surgical risks.
The fusion of century-old ECG technology with cutting-edge AI showcases a promising leap in predictive medicine.
It could revolutionise how doctors assess and manage risks associated with heart surgeries, ultimately enhancing patient care and outcomes.
