Related Experiment Video
Updated: Apr 29, 2026
06:08
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
2.1K
Prostate Cancer Classification Using Quantum Machine Learning on Multi-parametric MRI
Peng Chen1,2, Mojtaba Safari3, Rowan Barker-Clarke1
1Genomic Sciences and Systems Biology, Cleveland Clinic Research, Cleveland, OH 44106, USA.
Summary
Quantum machine learning, specifically amplitude-encoded quantum support vector machines (QSVM), shows promise for classifying prostate cancer lesions. This approach outperformed classical methods in a study using radiomics features from MRI scans.
Area of Science:
- Medical Imaging and Machine Learning
- Quantum Computing Applications
Background:
- Accurate prostate cancer lesion classification is crucial for patient management.
- Multiparametric MRI (T2W and ADC) radiomics features aid diagnosis.
- Classical machine learning models (SVM, RF, XGBoost) show promise but may miss complex patterns.
Purpose of the Study:
- To compare classical machine learning classifiers with quantum support vector machine (QSVM) variants for prostate cancer lesion classification.
- To evaluate the performance of different QSVM encoding strategies (amplitude, angle, projected quantum kernel) in a clinical imaging context.
Main Methods:
- Radiomics features were extracted from T2-weighted (T2W) and apparent diffusion coefficient (ADC) MRI images of 299 prostate lesions.
- Classifiers compared included SVM-RBF, Random Forests (RF), XGBoost, and three QSVM variants (amplitude encoding, angle encoding, angle encoding with projected quantum kernel).
- A nested stratified cross-validation pipeline with feature selection and hyperparameter optimization was utilized.
Main Results:
- Amplitude-encoded QSVM achieved the highest mean AUC (0.799 ± 0.082), surpassing SVM-RBF (0.608 ± 0.244).
- QSVM performance matched or exceeded RF (0.728 ± 0.083) and XGBoost (0.720 ± 0.065) while offering improved sensitivity at similar specificity.
- Quantum machine learning demonstrated competitive or superior performance in this small-sample, low-dimensional clinical imaging dataset.
Conclusions:
- Qubit-efficient QSVMs, particularly amplitude encoding, show significant potential for enhancing prostate cancer lesion classification accuracy.
- Quantum machine learning offers a valuable alternative for uncovering complex patterns in clinical imaging data that may be missed by classical methods.
- These findings support the future integration of quantum computing in diagnostic radiology for improved patient outcomes.
