A quantum-optimized approach for breast cancer detection using SqueezeNet-SVM

Journal: Scientific Reports

Year: 2025

Citations: 57

DOI: 10.1038/s41598-025-86671-y

Abstract

Breast cancer is one of the most aggressive types of cancer, and its early diagnosis is crucial for reducing mortality rates and ensuring timely treatment. Computer-aided diagnosis systems provide automated mammography image processing, interpretation, and grading. However, since the currently existing methods suffer from such issues as overfitting, lack of adaptability, and dependence on massive annotated datasets, the present work introduces a hybrid approach to enhance breast cancer classification accuracy. The proposed Q-BGWO-SQSVM approach utilizes an improved quantum-inspired binary Grey Wolf Optimizer and combines it with SqueezeNet and Support Vector Machines to exhibit sophisticated performance. SqueezeNet's fire modules and complex bypass mechanisms extract distinct features from mammography images. Then, these features are optimized by the Q-BGWO for determining the best SVM parameters. Since the current CAD system is more reliable, accurate, and sensitive, its application is advantageous for healthcare. The proposed Q-BGWO-SQSVM was evaluated using diverse databases: MIAS, INbreast, DDSM, and CBIS-DDSM, analyzing its performance regarding accuracy, sensitivity, specificity, precision, F1 score, and MCC. Notably, on the CBIS-DDSM dataset, the Q-BGWO-SQSVM achieved remarkable results at 99% accuracy, 98% sensitivity, and 100% specificity in 15-fold cross-validation. Finally, it can be observed that the performance of the designed Q-BGWO-SQSVM model is excellent, and its potential realization in other datasets and imaging conditions is promising. The novel Q-BGWO-SQSVM model outperforms the state-of-the-art classification methods and offers accurate and reliable early breast cancer detection, which is essential for further healthcare development.

Summary

This paper presents a computer-aided diagnosis pipeline for classifying mammograms as benign or malignant. Images are first cleaned with adaptive median filtering, morphological operations and CLAHE contrast enhancement, then augmented with rotations, flips, shears and blur, and segmented with Otsu thresholding to isolate regions of interest. SqueezeNet, a compact convolutional network with only 1.2 million parameters, extracts features from those regions. A support vector machine then performs the final classification, with its penalty parameter, kernel choice and feature subset tuned by a quantum-inspired binary Grey Wolf Optimizer (Q-BGWO).

The optimizer is the paper's main methodological contribution. Standard Grey Wolf Optimization moves candidate solutions toward the three best "wolves" in the population. The quantum-inspired variant represents each wolf's position as a qubit vector and updates it with rotation gates, so the search explores the parameter space more broadly and is less likely to stall in a local optimum. On 23 standard benchmark functions, Q-BGWO matched or beat binary GWO, plain GWO and particle swarm optimization on most tests.

The classifier was evaluated on four public mammography collections (MIAS, INbreast, DDSM and CBIS-DDSM) with 5-, 10- and 15-fold cross-validation, and compared with random forest, k-nearest neighbours, decision trees, naive Bayes, logistic regression, AdaBoost, gradient boosting, a plain SVM and a non-quantum BGWO-SQSVM. Q-BGWO-SQSVM led on every metric in every setting, reaching 99% accuracy, 98% sensitivity and 100% specificity on CBIS-DDSM under 15-fold cross-validation, and it also topped a table of 22 previously published methods. The authors note that validation on independent clinical datasets and collaboration with radiologists are needed before deployment.

Main Takeaways

• Quantum-Inspired Optimization Improves SVM Tuning: Replacing the standard Grey Wolf update with qubit vectors and rotation gates gave a consistently better search on benchmark functions and translated into higher classification accuracy than the non-quantum version of the same pipeline.
• A Lightweight Backbone Is Sufficient: SqueezeNet's 1.2 million parameters (versus 138 million for VGG-16) extracted features that were discriminative enough to reach near-perfect scores, which matters for clinics with limited computing hardware.
• Consistent Results Across Four Datasets: The model's lead over nine comparison classifiers held on MIAS, INbreast, DDSM and CBIS-DDSM, and performance rose slightly as the number of cross-validation folds increased, suggesting limited overfitting.
• 100% Specificity on CBIS-DDSM: The authors emphasize this figure because false positives drive unnecessary biopsies and patient anxiety. They stop short of clinical claims and call for testing on larger real-world data.