Researchers Unveil AI Capable of Forecasting Medical Diagnoses Years in Advance
Scientists announced on Wednesday that they have created an artificial intelligence (AI) model capable of predicting medical diagnoses years ahead, using the same technology behind OpenAI’s ChatGPT.
Delphi-2M: An AI Model for Early Disease Prediction
Based on a patient’s clinical history, the Delphi-2M AI “predicts the prevalence of over 1,000 diseases” years in advance, according to a study published in the journal Nature.
Collaborative Effort Across Borders
Researchers from the UK, Denmark, Germany, and Switzerland utilized medical records of approximately half a million patients from the UK Biobank biomedical database to train the AI.
They employed a technology similar to ChatGPT, an initial language model based on textual content.
Learning Medical Patterns
Moritz Gerstung, an AI specialist at the German Cancer Research Center, explained to the press that understanding a sequence of medical diagnoses is like “learning the grammar of a text.”
Through its training, Delphi-2M learns to detect patterns in health data before diagnoses are made, identifying combinations and sequences that lead to specific conditions. This capability allows for “highly significant and relevant health predictions,” Gerstung added.
Identifying High-Risk Individuals
Gerstung presented diagrams suggesting that Delphi-2M could identify individuals with a higher or lower risk of conditions like heart attacks than factors such as age might predict.
Testing and Future Applications
The AI’s capabilities were tested using health data from nearly two million people in Denmark.
In the future, models like Delphi-2M could help guide patient monitoring and potentially lead to earlier clinical interventions in preventive medicine, Gerstung stated.
However, research teams noted that the AI model requires further testing and is not yet ready for practical use.
Resource Optimization in Healthcare Systems
Tom Fitzgerald from the European Molecular Biology Laboratory and co-author of the study emphasized that such tools could help “optimize resources in a strained healthcare system.”
Currently, many countries use computer devices to predict disease risks, such as the QRISK3 program used by British doctors to assess stroke and heart attack risk.
Delphi-2M, however, can manage multiple diseases simultaneously over extended periods, according to Ewan Birney, another study author.
Significant Advance in Medical AI
Gustavo Sudre, a King’s College London professor specializing in medical AI, described this development as “a significant step toward scalable, interpretable, and ethically responsible predictive modeling in medicine.”
Key Questions and Answers
- What is the AI model called? Delphi-2M
- What does the AI model predict? The prevalence of over 1,000 diseases years in advance based on a patient’s clinical history.
- Which technology did researchers use to develop Delphi-2M? A technology similar to ChatGPT, an initial language model based on textual content.
- What databases were used to train the AI? Medical records of approximately half a million patients from the UK Biobank biomedical database.
- What are the potential future applications of Delphi-2M? Guiding patient monitoring and potentially leading to earlier clinical interventions in preventive medicine.
- Is Delphi-2M ready for practical use? No, the AI model requires further testing before it can be implemented.
- How could Delphi-2M benefit healthcare systems? It could help optimize resources in strained healthcare systems by predicting multiple diseases simultaneously over extended periods.
