Great result for PoliTO student team QubiTO in CERN hackathon

PoliTO QubiTO student team recently earned a second place in the GenQ Hackathon, organized by QAIVentures at CERN, Geneva. The team tackled a challenge focused on improving the classification of breast cancer images from the BreastMNIST dataset using a Quantum Convolutional Neural Network (QCNN). 

The project had a dual focus: first, to systematically benchmark various quantum encoding techniques for mapping classical image data onto qubits, analyzing their impact on quantum resource requirements (qubit count, gate count, circuit depth); and second, to explore pathways for enhancing the model's potential robustness.

Moving beyond a mere analysis of encodings, QubiTO innovated the core QCNN architecture by integrating a parallel, classical pathway for topological data analysis (TDA). This module is designed to extract persistent, noise-resistant features from the images. The outputs from both the quantum and topological streams were then fused and fed into a final classical layer for classification.

The team’s initial investigations into this hybrid model's resilience showed promising signs. When exposed to synthetic noise profiles mimicking artifacts from old or damaged MRI machines, the model's behaviour indicated a potential for improved noise resistance compared to the baseline QCNN. While this hypothesis requires further validation, it suggests a compelling avenue for future research. This work explores a novel paradigm for developing reliable diagnostic tools, which could one day be valuable in challenging environments, such as developing countries, where medical imaging equipment may be prone to producing noisy data.