The new fiber optic sensing platform enables continuous real-time monitoring of multiple markers in the skull

On August 16, 2022, Yuqian Zhang, Dr Yubing Hu and Professor Ali K. Yetisen from Imperial College London, in collaboration with a team of researchers from Sichuan University Jiang Nan, published a study titled “Multiplexed optical fiber sensors for dynamic brain monitoring” on Matter.

The results report a multiplexed biomarker fiber optic sensor that integrates a programmable AI prediction platform. The sensor is capable of detecting levels of multiple biomarkers in cerebrospinal fluid in real time and continuously, and diagnosing and dynamically monitoring different stages of head injury. The corresponding authors of the paper are Hu Yubing and Jiang Nan; The first author is Zhang Yuqian. The other authors are Liu Qiao, Lou Kai, Wang Shuhan, Zhang Naihan, ali K. Yetisen.

Traumatic brain injury (TBI) is one of the leading causes of death and disability worldwide, is the number one cause of coma, and is the leading cause of brain injury in adolescents and children. As a result, the need for accurate disease diagnosis and treatment of TBI is increasing, and how to dynamically monitor various biomarkers in brain tissue has become an urgent problem to be solved. At present, the Lix probe is commonly used clinically to measure the oxygen content and intracranial pressure in the brain tissue of patients with TBI. However, the probe can only be used for the detection of a single biomarker, and monitoring of multiple biomarkers requires multiple probes to be inserted into brain tissue at the same time, which will further lead to brain tissue damage and increase the risk of inflammatory infection. The main means of monitoring multiple brain biomarkers is microdialysis. Despite its advances in multimodal brain function monitoring, microdialysis is a sampling-based approach that requires regular extraction of cerebrospinal fluid to an external test bench for subsequent in vitro analysis, a process that requires repeated manual operations by medical staff, with a minimum sampling interval of up to 30 minutes each time, making it difficult to provide continuous real-time monitoring.

In recent years, electrochemical (EC) sensors and optical sensors have been developed for clinical brain monitoring. However, the EC sensor’s metal electrodes and external circuitry can lead to problems such as high cost, foreign body reaction, and poor magnetic resonance compatibility, hindering its clinical application. Optical sensors on an optical fiber platform offer significant advantages of small size, immunity to electromagnetic interference, easy remote and deep brain sensing, and low signal drift in continuous monitoring. However, crosstalk of optical signals between optical sensors and cross-sensitivity of multiple parameters pose a challenge for simultaneous dynamic monitoring of multiple biomarkers.

With this in mind, the research team designed a multiplexed biomarker fiber optic sensor that integrates a programmable artificial intelligence (AI) prediction platform for simultaneous dynamic monitoring of pH, temperature, dissolved oxygen (DO), and glucose levels in artificial cerebrospinal fluid (aCSF). The fiber optic sensor exhibits high sensitivity, high selectivity and stability when monitoring four brain biomarkers simultaneously, and has a short delay time, which can reflect the dynamic changes of biomarkers and identify the transition of the TBI stage, with great clinical application potential.

Research content

Figure 1: Schematic of a multiplexed fiber optic sensor for brain biomarker detection.

To achieve this, the team connected four different sensing films at the tip of the Y-fiber for simultaneous detection of four different biomarkers (Figure 1A). The study enables reading and analysis of biomarker concentrations by connecting a light source at one end of the Y-fiber and a spectrometer at the other end for reflection spectroscopy (Figure 1B). In order to eliminate crosstalk between the four sensors and provide highly accurate and quantitative readings of biomarkers, a machine learning-based regression model was developed and optimized using the features of the reflective spectrum. The four optical sensors are based on a fluorescence and colorimetric approach that works to alter the photophysical properties of the film through the interaction between the sensing membrane and the biomarker (Figure 1C-i). For long-term stable monitoring, the sensor is encapsulated in a transparent silica film (Figure 1C-ii) that avoids indicator leakage and maintains high permeability to biomarkers. Four sensing films are cut into a 1/4 circle and placed 4 mm above the tip of the fiber, and a reflective barrier made of glass microfibers is added to the surface of the sensing membrane to enhance the reflected signal and block background noise.

The research team used multiplexed sensors to simultaneously detect the pH, brain oxygen content (DO), temperature and glucose of artificial cerebrospinal fluid (aCSF), and the measurements were as follows. As shown in the figure, the fiber optic sensor distinguishes pH, DO, temperature and glucose when performing multi-parameter simultaneous monitoring, and the sensor has good repeatability and can be used repeatedly. Also, the sensing film has visible color variations in different physiological environments, confirming that the sensor can effectively detect biomarkers in aCSF (Figure 2).

Figure 2: Multiplexed optical biosensor sensing results for pH, temperature, brain oxygen (DO), and glucose.

In the multi-parameter experiment, it is found that the four signals cannot avoid the phenomenon of signal interference and output overlap, which makes the sensitivity of pH, temperature and glucose sensing a certain reduction, and it is difficult to calculate the detection parameters using only the values of a single wavelength (Figure 3A). The research team proposed to use machine learning algorithms to build four models from the collected spectral information to predict the four parameters (Figure 3B). In order to better train the effect, the research team extracted 10 spectral features from the spectrum to train the model, including the intensity corresponding to the sensitive wavelength of each parameter, as well as features such as peaks and valleys. In order to correct the effect of temperature on oxygen signaling, temperature is also trained on oxygen models as one of the characteristics when the model is trained. After comparing the effects of multiple models, the team found that using the Bayesian RIGI regression algorithm best predicted the four parameters. Finally, the team verified the effect of the model using a five-fold cross-validation method, and the results showed that the R2 of the pH, brain oxygen content, temperature, and glucose prediction models reached 0.84, 0.93, 0.94, and 0.94, respectively (Figure 3C-F).

In order to further detect the effectiveness and clinical application effect of the multiplexed optical fiber optic sensor, the research team used sheep brain to conduct ex vivo tissue verification. Before the test, the sheep brain was cleaned and washed away with its own cerebrospinal fluid. The researchers then prepared the corresponding aCSF according to the pH, brain oxygen, temperature and glucose content in the cerebrospinal fluid of healthy people, pre-TBI, mid-TBI and late TBI patients in actual conditions, and soaked the sheep brain in cerebrospinal fluid for simulated monitoring of the disease. In continuous monitoring, the multiplexed sensor can realize the measurement of cerebrospinal fluid at different stages, successfully and accurately calculate the content of these four biomarkers, and the changes between different stages can also be detected within 3 minutes, which can accurately reflect the switching of different stages between diseases, further confirming that the system has important clinical application prospects (Figure 3G).

WeChat image_20220818172228.png

Figure 3: In vitro TBI brain monitoring model study based on machine learning model.

(Source: Science Network)

Related paper information:

Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button