Machine Learning Fails In The diagnosis of Rare Diseases
A Call to Action for Quantum Machine Learning
At first glance, machine learning seems to be the perfect solution for medical diagnosis. Large-scale pattern recognition, relentless analysis, and ever-improving models. But when it comes to rare diseases, most machine learning systems reach their limits.
On average, patients with a rare disease wait 5 to 7 years before receiving a correct diagnosis. In that time, they may see multiple specialists, undergo dozens of inconclusive tests and receive misdiagnoses that lead to unnecessary or even harmful treatments.
This is the introduction into a real-world project that has been reimagined with an open-source dataset to respect confidentiality agreements. While the data is public, the methods and algorithms are largely the same as those used in the original, confidential work.
The intersection between quantum computing and clinical diagnosis is a fascinating and complex area. This series is an attempt to explore how Quantum Bayesian Networks could push the boundaries of rare disease diagnosis.
But to be clear: this is not another overhyped claim about the magical potential of quantum computing. We are taking a critical, grounded approach, diving deep into the mechanics of the algorithms and rigorously evaluating every step.
At this stage, we do not know whether our work will advance the state of the art in rare disease diagnosis. However, we are confident that, success or failure, this project will serve as a transparent blueprint for how to design, develop and evaluate a machine learning solution from start to finish.