Quantum kernel classifiers never beat a plain classical SVM on medical images.

Scientists often suggest that quantum computers might classify medical scans better than normal computers. Quantum methods can project image information into vast mathematical spaces to find hard-to-see patterns. To see if this actually helps, researchers ran a fair test. They took medical pictures from brain and breast datasets and shrank them into smaller digital features. Then, they tested two simulated quantum tools against regular computer models, including standard classical tools called support vector machines. The quantum models did not win a single time across 24 matched comparisons. In fact, standard classical models beat them in every test and scored higher overall. On top of that, the quantum methods were roughly 50 to over 220 times slower to run.

Why it matters

Hospitals and researchers deciding whether to invest in quantum machine learning need evidence, not hype. This benchmark shows that on compressed medical image features, quantum kernel SVMs gave no accuracy gain while costing 50 to 220 times more compute. That shifts the burden of proof onto quantum advocates.

Questions answered