Please login first
REAL-TIME PROSTATE CANCER SCREENING USING A HYBRID AI-INTEGRATED ELECTROCHEMICAL BIOSENSOR
* 1 , 2 , 3 , 4
1  Department of Computational Intelligence, School of Computing, SRM Institute of Science and Technology, Kattankulathur, 603203, India
2  Symbiosis Institute of Computer Studies and Research (SICSR), Pune, 412115, India
3  Associate Dean- Students, Shiv Nadar University, Delhi-NCR Campus, Noida, 201310, India
4  Department of Computer Science and Engineering, Rajalakshmi Engineering College, Chennai, 602105, India
Academic Editor: Eden Morales-Narváez

Abstract:

The integration of biosensor technology with hybrid artificial intelligence (AI) algorithms has greatly advanced the field of biomedical diagnostics. Prostate cancer is one of the most common types of cancer in men, and prostate-specific antigen (PSA) is a key biomarker for early diagnosis. Using an unprecedented approach that combines the strengths of a range of machine learning and deep learning methods, this methodology shows a balance between accuracy, sensitivity, and reliability in early cancer detection. The biosensor, using an electrochemical platform, shows the dual functionalization of a specific aptamer and gold nanoparticles (AuNPs), enabling PSA detection with high specificity. The hybridization of the artificial intelligence methods used in this study includes convolutional neural networks (CNNs), support vector machines (SVMs), and gradient boosting machines (GBMs), and it allows for the processing, assessment, and classification of biosensor data. The initial step involves collecting the biosensor signal of biomolecular interactions, which are transduced and transcribed into electrical signals. To prepare the raw data set for processing, advanced denoising and normalizing techniques are applied. Then, the CNN is run on these data to encode their features and identify complex patterns. Then, SVMs classify the PSA levels into three groups, namely normal, elevated, and at risk. In contrast, GBMs use CNN and SVM output predictions as inputs for the decision process.

In this manner, this hybrid algorithm methodology balances the interpretability of SVMs, the deep feature learning capacity of CNNs, and the strong prediction power of GBMs. Its second contribution first aims at increasing the amount of data by developing an ERL (ensemble reinforcement learning) framework that dynamically modulates the parameters based on instantaneous rewards. This makes the ensemble reinforcement learning framework unfixed for different biomedical applications but also for different datasets and patient profiles. Herein, we detail the design, development, and evaluation of this transformative system with the potential to fundamentally change cancer diagnostics by enabling scalable, low-cost, and accurate cancer screening. The hybrid algorithm strategy is a paradigm shift in biosensor technology, allowing us to address the limitations of traditional approaches while overcoming issues with achievable data standardization and model optimization.

Following the summary of the acquired optically significant metrics, the output efficiency is estimated to thoroughly review the proposed AI-adapted biosensor framework. Using this approach, the system achieved high accuracy in detecting prostate-specific antigen (PSA) concentrations and classifying the risk for cancer. This can be particularly useful for early-stage detection while reducing the risk of false-negative diagnoses. The specificity will reduce the false-positive results—meaning those whose condition is misidentified—and spare some unnecessary medical interventions. The biosensor also exhibits an ultra-low limit of detection (LOD), which allows for the detection of PSA within the subclinical range. This innovative system takes advantage of machine learning and AI techniques for biosensor technology applications, thus enabling new biomedical diagnostics with high sensitivity and specificity and low cost.

Keywords: AI-enabled biosensors; Prostate cancer detection; Prostate-specific antigen (PSA); Hybrid AI algorithms; Electrochemical biosensor; Early-stage cancer diagnosis; Sensitivity and specificity; Limit of detection (LOD); Real-time analysis; Machine learning m
Top