麻豆淫院

Quantum computing is still in its early stages of development, but researchers have extensively explored its potential uses. A recent study conducted at S茫o Paulo State University (UNESP) in Brazil proposed a hybrid quantum-classical model to support breast cancer diagnosis from medical images.

The work was as part of the 2025 IEEE 38th International Symposium on Computer-Based Medical Systems (CBMS), organized by the Institute of Electrical and Electronics Engineers (IEEE). In the publication, the authors describe a hybrid neural network that combines quantum and classical layers using an approach known as a quanvolutional neural network (QNN). They applied the model to mammography and ultrasound images to classify lesions as benign or malignant.

"What we wanted to bring to this work was a very basic architecture that used quantum computing but contained a minimum of quantum and classical devices," says Yasmin Rodrigues, the first author of the study. The work is part of her scientific initiation project, supervised by Jo茫o Paulo Papa, full professor in the Department of Computing at the Bauru campus of UNESP. Papa also co-authored the article.

They chose breast cancer as the target for testing the computational model because it is the most common type among women worldwide. In 2022, there were 2.3 million cases and 670,000 deaths recorded. Early detection is crucial to increasing the chances of cure and survival. However, traditional methods, such as mammography, rely heavily on human interpretation, which can lead to variations in diagnosis.

"Although theoretically simple to follow, mammography is still an exam whose interpretation depends heavily on the professional performing the procedure," says Papa.

What distinguishes UNESP's work from other artificial intelligence initiatives in health is its use of a quantum convolution layer alongside a classical layer.

"Like classical convolution, the goal of quantum convolution is to extract local features from structured data, such as images. But it does so by taking advantage of unique properties of quantum systems, such as superposition and entanglement, which make it possible to process information much more efficiently and quickly," says Papa. In the study, the quantum layer, composed of four qubits (quantum bits), replaced the traditional process of extracting features from images.

"What we did, basically, was to pass the images through this four-qubit quantum circuit, with rotation operations and logic gates. This enabled us to obtain the necessary measurements. Then, the images went to simple classical layers, which delivered the final classification," explains Rodrigues.

The study did not use a true quantum processor, but rather a classical platform such as the PennyLane framework, which reproduces the ideal behavior of a quantum circuit without environmental noise.

There are few true quantum computers in the world. They are all in the experimental phase and have a limited number of qubits, ranging from a few dozen to just over a thousand. They require impeccably clean rooms, vibration isolation, electromagnetic shielding, and in most cases, cooling close to absolute zero (-273 掳C). Therefore, when made available to customers, their use is disproportionately expensive.

"Simulators like the ones we use work entirely on classical platforms, don't use real qubits, but give an idea of how circuits would behave in the ideal quantum world. They're error-free, free from environmental variations that greatly affect current quantum computers," explains Rodrigues.

According to the researcher, although extremely simple, the simulated quantum circuit has already shown promising results.

"The best-performing classical network had 11 million parameters. Ours, with the quantum layer, had about 5,000. That changes everything," she compares.

A fundamental physics concept behind the model is superposition. Superposition differentiates the qubit from the classical bit.

Rodrigues explains, "To understand superposition, it helps to refer to a representation known as the Bloch sphere. We can imagine this sphere as a soccer ball, where each point on the surface represents a possible quantum state. At the top of the sphere, the North Pole, we have state 0. At the South Pole, state 1. All other points on the surface correspond to states of superposition between 0 and 1.

"When the system is exactly at the North or South Pole, we're 100% sure that the qubit is in state 0 or 1, respectively. But at any other point on the sphere, we enter probabilistic territory: The qubit is in a superposition with varying probabilities of being measured as 0 or 1."

There are many possible superposition states. For instance, a qubit could be in a state with a 70% probability of being a 0 and a 30% probability of being a 1, or any other combination. Due to these superimposed states, qubits can process much more information than classical bits. This is one reason quantum computing has so much potential to surpass classical computing. It has incomparably greater processing capacity and speed. However, many technological challenges must be overcome before this potential can be realized on a large scale. =

"The expectation is that people will have quantum computers at home, just as they have conventional computers today," Papa points out.

In the study, the information encoded in the qubits was pixels from mammography and ultrasound images. Sometimes it was just one pixel, and sometimes it was more than one. The model was tested with two databases: BreastMNIST (with ultrasound images) and BCDR (with segmented mammograms). Even with a circuit of only four qubits, the hybrid network performed competitively. In the best case, it achieved 87.2% accuracy in the test set and 86.1% in the validation set.

"The idea was to create an architecture that could be used and further developed in other studies," Rodrigues comments.

Other applications

Although the study focused on breast cancer, the authors point out that the developed architecture can be applied to other areas. For example, it can be used to analyze brain lesions or classify tissues in microscopy images.

"We're taking the first step toward a new computing paradigm for medical diagnosis. It's a promising field likely to grow significantly in the coming years," the researcher concludes.