Deep-learning powered whispering gallery mode sensor based on multiplexed imaging at fixed frequency

The authors proposed a first example of an affordable self-learning whispering gallery mode sensor and analyzed its performance on the refractive index variations detection. The distinguishing feature of the introduced sensing method is simultaneous collection of the radiated illumination that has been modulated by the microcavities. The light from the cost- and energy-effective laser source operating at the fixed frequency has been coupled in a robust manner into the multiplexed sensor with hundreds of spherical microcavities and then has been in parallel collected. In order to interpret the complex non-linear modulation and multidimensional response of the instrument, the authors suggested a solution based on the deep-learning engine. The reported results demonstrate the possibility for construction of the self-learning sensing solutions with the affordable instrument configuration, reduced complexity and device size for the first time, and are expected to significantly contribute to the change of the sensing paradigm from model-based to the machine learning inspired approach. The proposed sensor supplemented by the essential set of training data that can be automatically collected may be utilized in a wide range of the practice-oriented sensing tasks where the a priori known data about the response model is redundant.

Fig. Artistic representation of the introduced whispering gallery mode sensor: the fixed frequency laser signal is collected in parallel from multiple microcavities and is processed with a deep-learning engine.

In the last years the numerous advantages such as high sensitivity, miniaturization potential and cost efficiency have designated the optically based sensing as one of the most demanded measurement solutions. Special attractiveness has been gained by the label-free measurement methods where the sensing agent interacts directly with the sensor. The phenomenon of the optical resonance in the dielectric circular microcavities referred to as the effect of the whispering gallery modes (WGMs) stands out here due to the extremely low detection limit down to single nanoparticle level. The optical microcavity-based sensing has thus a great potential to be implemented in the form of a non-destructive, label-free, real-time single-nanoparticle on-chip sensor with prospects for re-usability. In addition, the WGM sensing technology has been already painstakingly researched so that the respective robust devices are currently expected to emerge for the applications of the personalized medicine, environmental monitoring, chemical threat sensing, food and water quality control, etc. Nonetheless, despite of being on the forefront of the today’s sensing paradigm, the WGM sensing approach suffers from several limitations so that its transition from the laboratory to the real-world deployable systems has not succeeded yet.

For the widespread implementation of the WGM resonator sensors the further advancements have to be made in terms of cost, speed, portability, and multiplexity. Here the main restriction of the known configurations of the WGM instruments is referred to the single microresonator interrogation concept and is based on the collection of the transmitted (unmodified) or scattered (doped with fluorescence label) intensity with a photodiode to trace the resonance position in the acquired spectrum. Moreover, the common WGM instrument configurations require either a narrow linewidth frequency sweeping source or a high-resolution spectroscopic unit supplemented with the appropriate spectral data processing methods that taken together limits the sensor performance in terms of the achievable temporal and/or spectral resolution and restricts the utilization of the WGM instrument outside the laboratory.

In order to push the WGM sensing technology towards the real-life tasks solving, a robust platform for parallel light coupling into and signal collection from the multiple microcavities allocated on a single chip has been introduced. The relatively simple sensor fabrication, reusability, possibility for multicavity signal collection and intensity-based signal detection captured at a fixed laser frequency makes this solution especially attractive. Since the complex and multidimensional nature of the captured signal in this case complicates the interpretation of the external variations utilizing the analytical descriptions, the latter has been addressed by the deep-leaning engine which enabled to reach the refractive index prediction accuracy comparable to the other instrument configurations but in an affordable way. The article is entitled “Deep-learning powered whispering gallery mode sensor based on multiplexed imaging at fixed frequency” and published in Opto-Electronic Advances Issue 11 2020.

About The Group

The international group of researchers from the Chair of Applied Laser Technologies at the Ruhr University Bochum and from the Belarusian State University has a long-term cooperation on development and application of the WGM microcavities for physical and biochemical parameters sensing. Special emphasis in the research activities of the group has been made on sensors’ fabrication, establishment of the measuring methods, development of the optical processes for the laser-based microsystems manufacturing and on the development of the automated data processing algorithms. This group of researches has a many-years expertise in the development and advancement of different innovative laser-based methods for additive manufacturing, sensor technology, material characterization and processing in the fields of micro- and nanophotonics, biomedical technology, microsystem technology, material science and photovoltaics, machine learning algorithms and data mining approaches.


Saetchnikov A V, Tcherniavskaia E A, Saetchnikov V A, Ostendorf A. Deep-learning powered whispering gallery mode sensor based on multi-plexed imaging at fixed frequency. Opto-Electron Adv 3, 200048 (2020).

DOI: 10.29026/oea.2020.200048