Machine learning is indeed extremely good at pattern recognition, but I wouldn't trust an LLM to reliably identify patterns, especially in a medical context. As other commenters have said, this article is evidence of classical methods continuing to be useful.