Chairs: 

  • Prof. Dr. Igor Meglinski, Aston University
  • Dr. Alexander Doronin, Victoria University of Wellington
  • Dr. Anna Baldycheva, Exeter University

Nowadays, artificial intelligence (AI) and machine learning have gained prominence through the development of computational facilities and play an important role in private and scientific life. Microchip manufacturers implement AI algorithms into hardware layers, providing unprecedented abilities in processing data “on the fly”, providing a glimpse of the future of smart sensors. One of the first available products is HiSilicon Kirin 970, that uses AI for processing of images obtained from cameras. Based on computer vision and closely related image processing and machine vision, AI has made significant progress in image and video analyses, but has not been widely and successfully used for spectroscopic data, while integration of sensors, computational units, and AI-specific ISA (instruction set architecture) into a single device has yielded autonomous hyperspectral multidimensional smart sensors which satisfy the concept of Internet of Things (IoT).

Smart sensors should be able to classify incoming data, which is a one of the tasks of machine learning (ML). The fastest implementation of ML methods for data classification is referred to as supervised learning. One aspect of ML pattern recognition is aiming to find regularities and features (patterns) in the experimentally obtained images. Support vector machine (SVM) is the well-developed and fast implementation of supervised learning which is widely used in pattern recognition. SVM is based on regression searching for such a set of hyperplanes in multidimensional space which is the furthest from each of the classes (datapoint groups) displayed in the training dataset. Data in SVM are presented in a form of n-dimensional vectors. When such a hyperplane is found, it is then used to sort incoming data into classes.

Augmented reality (AR) is a hot topic not only in entertainment but also in optics, since it allows for simultaneous visualization of features at an object together with sensing process, e.g., projecting features at an object surface while it is being viewed through a microscope. In prospective smart sensors, AR will reveal the information that is obtained and classified by the sensor itself, meaning that each sensor will have its own integrated circuit sensor and processor unit with algorithms for data.

The application of distributed computing, such as cloud computing and CUDA, speeds up data processing and helps to distribute data analyses between cloud, CPU, and GPU to ensure the best performance. Such an approach moves with the times, since smart sensors within the concept of IoT distribute the data analyses exactly in this manner. OpenCL is proposed as a main framework due to high flexibility and scalability. It can be used for programming, such GPU as CUDA, and also in embedded systems. Data transfer is proposed to be compatible with modern 3GPP communication standards as well as prospective 6G.

The aim of this Special Issue is to highlight the latest exciting developments in promoting the use of AI/ML in biomedical devices and applications by attracting leading researchers to present the results of their latest efforts. Accepted contributions will include the implementation of AI/ML tools for spectroscopic optical data analysis, processing of clinical/pre-clinical/biomedical images/data obtained with modern photonics-based technologies, automatic standalone tissue screening, biopsy pattern recognition, smart clothes and photonics sensors in concept of IoT, AI algorithms, cloud-based computing, how CUDA accelerates data/image processing, and more.

Session topics:

  • AI algorithms in biomedical imaging and optical diagnostics
  • Explosion of IOT and IOMT data with connected health
  • AI and machine learning in medical data structuring
  • Robotics and Cobotics
  • Machine learning in biophotonics
  • Optical tomography and biomedical visualization with AI
  • Smart optical biopsy
  • Automatic standalone screening of cancer and tissue characterization
  • Hyperspectral multidimensional smart sensors satisfying the concept of Internet of Things (IoT)
  • Smart optical and photonics sensors in concept of IoT
  • Smart image and video analyses
  • Support vector machine in pattern/tissues recognition
  • AI and machine learning in multimodal imaging
  • Advantages in AI algorithms for data and image processing
  • Cloud computing and the acceleration of data/image processing by CUDA