FAU Engineering Study Takes a 'Quantum Leap' to Detect Kidney Disease

Researchers explored how machine learning and quantum computing can be used to improve early detection of chronic kidney disease, aiming to develop faster, more accurate diagnostic tools for clinicians.

Chronic kidney disease (CKD) is one such condition - a progressive illness that damages the kidneys over time, eventually leading to kidney failure if left untreated. Because CKD develops gradually and often shows few symptoms in its early stages, timely diagnosis is a major clinical challenge.

Globally, an estimated 850 million people are living with some form of kidney disease. Among them, as many as 10 million require dialysis or kidney transplantation to survive. Despite the scale of this problem, CKD frequently goes undetected until it reaches an advanced stage. Early diagnosis is essential not only to slow disease progression but also to improve quality of life and survival rates.

To help address this widespread issue, researchers are increasingly turning to artificial intelligence and machine learning (ML) to build automated tools that can detect CKD more efficiently and accurately. ML algorithms can recognize subtle patterns in complex medical data - patterns that might otherwise go unnoticed by clinicians.

Research from the College of Engineering and Computer Science at Florida Atlantic University is taking this concept further by exploring how quantum computing could enhance the accuracy and performance of ML-driven CKD diagnosis systems.

Arslan Munir, Ph.D., senior author and an associate professor in the FAU Department of Electrical Engineering and Computer Science, and his colleagues from Bangladesh, developed and compared two automated systems for CKD diagnosis: the Classical Support Vector Machine (CSVM) and the Quantum Support Vector Machine (QSVM). The goal of their study was to evaluate the efficiency and diagnostic accuracy of both approaches, and to better understand how emerging quantum machine learning techniques could eventually revolutionize real-world medical diagnostics.

The team began by preparing and refining a CKD dataset, applying comprehensive data preprocessing to ensure the reliability of results. They then used two advanced data optimization methods: Principal Component Analysis (PCA) and Singular Value Decomposition (SVD) to reduce noise and improve computational efficiency. Each optimized dataset was subsequently analyzed using both CSVM and QSVM algorithms, allowing for a detailed comparison of the two methods' predictive capabilities.

The results of the study, published in the journal Informatics and Health, revealed clear differences. When paired with PCA, CSVM achieved a remarkable 98.75% accuracy, while QSVM reached 87.5%. Using SVD, the CSVM achieved 96.25%, compared to 60% for the QSVM. The classical SVM also proved far faster: up to 42 times quicker than the QSVM in certain experimental settings. The outcome indicates that, under current hardware conditions, the classical approach remains superior in both accuracy and time efficiency.

However, Munir and his colleagues emphasized that QSVM's underperformance is primarily a reflection of today's computational limitations rather than the potential of quantum algorithms themselves. Even within the constraints of classical hardware, the QSVM still achieved competitive performance - its 87.5% accuracy using PCA surpasses that of several existing classical SVM methods reported in prior studies. This suggests that hybrid quantum-classical systems could play an increasingly important role in the near term, combining the strengths of both paradigms to improve diagnostic precision while managing current technological challenges.

"What makes our work unique is that we didn't just apply classical machine learning to detect chronic kidney disease - we also tested a quantum version under the same conditions," said Munir. "By directly comparing classical and quantum models, and using two different optimization methods, we gained valuable insight into where the technology stands today and how quantum computing could help shape the future of health care analytics."

Looking ahead, the research team plans to extend their work by exploring additional quantum ML algorithms beyond QSVM and testing their methods on larger, more diverse medical datasets. They also intend to focus on optimizing feature selection techniques to ensure scalability and adaptability across a wide range of diagnostic applications. Ultimately, the goal is to create more reliable, efficient and accessible AI-driven diagnostic tools that can assist clinicians in making faster, more accurate medical decisions.

"This research is an important leap toward bringing quantum computing into health care - an emerging field with the power to transform how we detect and treat complex diseases," said Stella Batalama, Ph.D., dean of the College of Engineering and Computer Science. "By combining machine learning with next-generation quantum technologies, this work offers real hope for earlier, faster and more accurate diagnosis of chronic kidney disease, ultimately improving outcomes and saving lives."

-FAU-