Deep Learning-Enhanced Plasmonic Biosensor for High-Sensitivity Hepatitis B Detection: A Portable Diagnostic Platform for Point-of-Care Use

Detecting hepatitis B surface antigen, or HBsAg, has been an important focus in medical diagnostics because of the significant health impact of hepatitis B worldwide because it serves as a marker of infection and is the first sign that hepatitis B is present in the body. The sooner HBsAg can be detected, the more effectively healthcare providers can manage the disease and limit its spread. But current diagnostic tools face big obstacles: they’re often not accurate, fast, or convenient enough for effective early screening. Many of today’s commonly used methods, like the enzyme-linked immunosorbent assay (ELISA), chemiluminescent immunoassay (CLIA), and lateral flow assay (LFA), have their own limitations. ELISA, for instance, is very precise but requires specialized lab equipment and takes time, which doesn’t make it a practical choice when you need quick results in point-of-care settings. CLIA is sensitive but tends to be expensive and hard to access in areas without advanced facilities. Meanwhile, the more portable LFA has limited sensitivity. What’s really needed is a diagnostic approach that brings together high sensitivity, quick turnaround, and ease of use, especially in busy or low-resource healthcare settings. To overcome the challenges of current HBsAg tests, a team of researchers from The First Hospital of Jilin University—led by Weihong Sun, Professor Jingjie Nan, Hongqin Xu, Lei Wang, Professor Junqi Niu, Professor Junhu Zhang, and Professor Bai Yang developed a new diagnostic platform that pairs plasmonic biosensing with deep learning. Their aim was to develop a tool that’s not only highly accurate but also versatile enough for quick point-of-care use. They call it the Deep Learning-Enhanced Plasmonic Biosensor (DLPB) platform. By combining the sensitivity of plasmonic biosensors with the pattern-recognition strengths of neural networks, they’ve made a significant step toward a more practical and effective way to detect HBsAg and possibly other biomarkers in the future. The authors started by creating a unique, layered “sandwich” structure, specifically designed to pick up even the faintest traces of HBsAg. This setup relies on a concept known as localized surface plasmon resonance (LSPR), where the interaction of light with tiny plasmonic particles generates strong, measurable signals. Whenever HBsAg binds to the biosensor, the optical signal shifts, directly reflecting the concentration of the antigen and setting the foundation for an accurate detection process.

But, as you can imagine, raw data like this can be complex. Just capturing those signals doesn’t automatically make the results clear and interpretable. To make sense of the data, the team turned to neural networks—specifically a multilayer perceptron and a convolutional neural network (CNN)—to analyze the subtleties in the optical signals. They trained these networks on various signal patterns produced by different levels of HBsAg, teaching the algorithms to identify even the smallest changes that indicated the presence of the antigen. This approach significantly boosted accuracy. The multilayer perceptron model, skilled at recognizing patterns, handled the variability in the signals smoothly, while the CNN focused on specific features associated with HBsAg. Together, the models achieved an impressive detection accuracy of up to 99.6%, proving how well the plasmonic biosensor and neural networks worked together. The authors didn’t stop there. They wanted to make sure the platform could detect incredibly low levels of HBsAg. They began testing with smaller and smaller amounts of the antigen, pushing the limits of the sensor’s sensitivity. The result? The platform could identify HBsAg at concentrations that traditional methods would likely miss entirely. They set a new advancement for sensitivity, detecting HBsAg at just a few picograms per milliliter. This level of sensitivity makes the platform ideal for early detection, especially in diseases like hepatitis B, where catching the infection early on can make a huge difference in treatment outcomes. With an eye on real-world application, the researchers also explored the system’s potential as a point-of-care diagnostic. They adapted the setup to be portable, recreating the kind of conditions often found in lower-resource environments. To test how well it could work without lab-grade equipment, they even swapped out a high-powered microscope for a smartphone camera. The results were impressive—the smartphone setup maintained nearly the same accuracy, showing that the DLPB system could reliably detect HBsAg without needing expensive lab tools. This adaptability means it could work well in diverse healthcare settings, including remote or underserved areas where full lab setups aren’t available.

At every stage, this study highlighted the exciting potential of combining plasmonic biosensors with neural networks for medical diagnostics. The deep learning models made it possible to interpret complex signal shifts with high precision, while the sensitive plasmonic structure ensured the signals were strong enough for accurate detection. Altogether, this work sets the stage for a new type of diagnostic tool that is not only highly sensitive and cost-effective but also versatile enough to transform early disease detection and real-time healthcare diagnostics across the board. In conclusion, the new study takes a really fresh approach to disease detection, combining the best of two worlds: the incredible sensitivity of plasmonic biosensors and the analytical muscle of neural networks. What the researchers have come up with is a tool that can pick up even the tiniest traces of HBsAg with remarkable speed and accuracy. Indeed, we think this innovation could make a huge difference, especially for healthcare settings that don’t have the resources or equipment for standard testing. Instead of relying on lengthy lab work, this tool offers a quick, reliable way to detect diseases like hepatitis B early on, which could be a game-changer for managing infectious diseases. And because the platform can work with something as simple as a smartphone, it’s easy to imagine it being used for mobile testing or even at-home screening. Just picture a healthcare worker visiting a remote community and running reliable tests on the spot—this tool could make early detection a reality for so many more people, leading to faster interventions and better health outcomes. Moreover, the new platform can be adapted for detecting many kinds of biomarkers for other diseases. This adaptability makes it a powerful tool not only for infectious disease but also for personalized medicine, where understanding specific markers can guide individualized care.