The hybrid quantum-classical neural network for analysing brain MRI scans is a development by researchers of Lobachevsky State University of Nizhny Novgorod, presented in September 2026. The system combines a classical neural network architecture with quantum circuits and is designed to detect brain tumours automatically on medical images. In tests on open databases the network identified the presence of a tumour with accuracy of up to 98 per cent.[1][2]

Development and architecture

The network was created by specialists of the research centre for artificial intelligence at the university's Institute of Information Technology, Mathematics and Mechanics, in particular at the laboratory "Artificial Intelligence in Preventive Medicine".[1] The key feature of the architecture is a quantum layer working with a compressed set of image features: this allows deeper analysis with far fewer trainable parameters than traditional neural network models, which may run to billions of such parameters and demand enormous computing resources.[1][3]

The author of the study, senior researcher at the laboratory and candidate of physical and mathematical sciences Marina Bastrakova, explained that the system's effectiveness comes chiefly from the interaction of the classical and quantum parts of the model. Experiments showed that the structure of the quantum layer substantially affects the result: models with more elaborate repeating quantum blocks retained accuracy better on distorted images. The mechanism behind this effect has yet to be studied and confirmed on real quantum hardware.[1]

Noise resistance and test results

One important result was the system's resistance to low-quality MRI scans — those with noise and defects arising from patient movement, equipment characteristics and other factors. Under strong Gaussian noise the quantum-classical model achieved 96.6 per cent accuracy, while a classical neural network under the same conditions gave 92.2 per cent.[1][2]

The laboratory's chief researcher, doctor of physical and mathematical sciences Mikhail Ivanchenko, stressed that such systems are not meant to replace the physician: their task is to help the specialist in the first analysis of images, draw attention to important signs and reduce the chance of missing pathological changes.[1][2]

Publication and next steps

The research was carried out under the "Priority-2030" programme and its results were published in the international journal The European Physical Journal Special Topics.[2] As the next stage the developers plan to test the technology on new medical datasets using Russian quantum computers, to train the system to describe scans in more detail and to broaden the classification of images.[3][2]

Sources

  1. NIA Nizhny Novgorod — University researchers teach a quantum neural network to analyse MRI scans more accurately
  2. Newsroom24 — Neural network of university researchers identifies brain tumours on MRI with 98% accuracy
  3. Zhivyom v Nizhnem — Researchers combine quantum technology with AI for MRI analysis (archived)
  4. Otkrytiy Nizhny — Researchers teach AI to find tumours on MRI even on noisy scans (archived)
  5. Vremya N — University researchers present a new development