*Important notice: This news reports on an unedited version of an accepted paper and is awaiting final editing. Therefore, the paper should not be regarded as conclusive or treated as established information.
Numerical simulations of a novel ultra-compact biosensor have distinguished glioblastoma from peritumoral brain tissue, achieving an optical sensitivity of approximately 933 nm/RIU and a footprint of just 76 μm2. The design combines a two-dimensional photonic crystal structure with neural network algorithms to accurately classify glioblastoma tissue.
Study: A compact 2D photonic crystal biosensor for optical detection of glioblastoma brain tissues enhanced by neural networks. Image Credit: Natali _ Mis/Shutterstock.com
These findings, published in the journal Scientific Reports, suggest that the system could be used to support label-free tissue analysis during brain surgery, potentially addressing the need for rapid and non-invasive diagnostic methods.
The Challenges of Glioblastoma Detection
Glioblastoma is an aggressive and lethal primary brain tumor that can spread into surrounding healthy brain tissue. This diffuse infiltration can make it difficult to identify the tumor boundaries during
surgery and remove the tumor while preserving healthy tissue.
Conventional methods for diagnosis and intraoperative tumor mapping, including magnetic resonance imaging (MRI), computed tomography (CT), and standard pathological biopsy, can be invasive, time-consuming, and resource-intensive. There is, therefore, a need for faster methods to identify tumor boundaries during surgery.
Optical sensors based on photonic crystals offer one possible approach by detecting small changes in refractive index at the cellular level without requiring chemical labels or destroying the tissue.
Design of the Photonic Crystal Sensor
To address the limitations of these diagnostic tools, researchers designed a 2D photonic crystal biosensor, consisting of a hexagonal lattice of silicon rods suspended in air. The array contains 17 × 16 rods, with a lattice constant of 600 nm and a standard rod radius of 120 nm.
The study introduced defects into the lattice to create separate input and output waveguides and a central ring resonator. The ring resonator serves as the sensing region and contains 11 dielectric rods with an optimized radius of 144 nm.
During operation, a single-mode laser produces light at a wavelength of 1.55 μm. An optical fiber couples the light into the input waveguide. When a biological tissue sample is placed on the sensing region, changes in its refractive index alter the resonance wavelength.
The sensor was evaluated using the finite-difference time-domain (FDTD) method in RSoft software. The simulations solved Maxwell's equations to model light-matter interactions and calculate optical properties such as transmission spectra. The study also combined the photonic crystal sensor with an artificial neural network (ANN).
The feedforward multilayer perceptron used a rectified linear unit (ReLU) activation function. It was trained using spectral parameters, including ambient temperature, resonance wavelength, and peak transmission intensity, to autonomously classify samples as tumorous or pre-tumoral tissue.
Download a copy of this page for later by clicking here!
Performance Metrics: Sensitivity and Thermal Stability
Numerical simulations demonstrated that the photonic biosensor could distinguish between the refractive indices of pre-tumoral tissue (1.341) and tumor tissue (1.344) at 25 °C. The sensor achieved a maximum Q-factor of 4777 and a figure of merit (FOM) of 2333 RIU-1, thereby indicating its exceptional performance.
Its maximum optical sensitivity was approximately 933 nm/RIU, with a detection limit of 4.2 × 10-5. The sensor's performance was also assessed across temperatures ranging from 25 °C to 40 °C. While resonance peaks shifted due to temperature-induced refractive index variations, the optical responses for peritumoral and cancerous tissues remained distinguishable throughout the modeled range.
Using eight baseline simulated profiles simulated with synthetic training data, the ANN achieved 100% classification accuracy under group-aware cross-validation. These results indicate that the combination of photonic crystal sensing and ANN-based classification can distinguish tissue types without needing manual selection.
Implications for Intraoperative Brain Tumor Mapping
This integrated photonic and artificial intelligence (AI) platform has the potential for intraoperative brain tumor mapping and surgical assistance. With a compact footprint of 76 μm2, the device could be incorporated into lab-on-a-chip (LOC) systems or compact diagnostic probes.
During glioblastoma surgery, such probes could monitor changes in brain tissue at tumor margins in real time, providing critical information to help differentiate tumor tissue from peritumoral tissue. Because the sensing method relies solely on refractive index changes, it does not require fluorescent labels or chemical staining, facilitating faster, label-free tissue analysis during surgical procedures.
A Foundation for Future Innovations
This study successfully combined a 2D photonic crystal ring resonator with a machine learning model for the optical detection of brain tissue. The simulated sensor achieved a sensitivity of 933 nm/RIU, a Q-factor of 4777, and a compact footprint of 76 μm2.
The results demonstrated that temperature changes from 25 °C to 40 °C influenced the resonance response but did not hinder the ability to distinguish between modeled tissue types.
Overall, this research lays a solid foundation for the further development of compact photonic biosensors for the detection of brain tumors. Future work should focus on fabricating the proposed sensor and testing it with actual tissue samples in ex vivo and clinical conditions. Additional validation will be essential to determine its accuracy, temperature stability, and suitability for real-time tumor-margin assessment during surgery.
Journal Reference
Mohamadpour, G., et al. (2026). A compact 2D photonic crystal biosensor for optical detection of glioblastoma brain tissues enhanced by neural networks. Scientific Reports. DOI: 10.1038/s41598-026-72361-w. https://www.nature.com/articles/s41598-026-72361-w.
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.