Title : Application of new techniques to transillumination imaging of animal bodies using NIR light
Abstract:
X-ray, MRI, and ultrasound imaging are well-known non-invasive methods for visualizing the internal structures of the human body. We are exploring the potential of using near-infrared (NIR) light as a new modality. NIR light penetrates living tissue more effectively than light in other wavelength ranges and holds promise for medical applications due to its features, such as enabling repeated continuous video imaging at the bedside. However, its practical application has been hindered by low spatial resolution caused by strong light scattering in biological tissues.
To address this issue, we have developed various methods to suppress scattering effects. One innovative hardware method involves applying the time-reversal principle of phase-conjugated light to transillumination imaging of animal bodies. We found that image quality significantly improves by incorporating intensity distribution information in addition to the previously used two-dimensional phase information. Additionally, the spatial resolution of transillumination images was greatly enhanced by using a chirp waveform with a time-sweep modulation frequency in the ultrasonic modulation of the phase-conjugated light, allowing for better focus within biological tissues.
On the software side, we have applied deep learning techniques. A computer learns the phenomenon of light scattering in biological tissue by analyzing pairs of images—one degraded by light scattering and one pre-degradation. When a new degraded image is input into the trained neural network, a non-degraded image is produced as output. This deep learning application requires a large number of training image pairs. We have successfully derived a point spread function (PSF) that models light scattering in biological tissue as an analytical solution by applying the diffusion approximation to the energy transport equation. Using the convolution of this PSF with non-degraded images, we can artificially generate a vast number of training image pairs. Furthermore, since the PSF is a function of the light absorber's depth, the neural network trained through deep learning can also determine the absorber's depth during the image degradation elimination process. This capability enables clear 3D imaging of internal structures from a single blurred 2D transillumination image.
The application of these new hardware and software techniques can significantly advance NIR transillumination imaging towards practical biomedical use.
Audience Take Away Notes:
- The audience will discover the potential of another noninvasive technique for biomedical imaging.
- They can see new techniques to solve the problem of severe blurring caused by light scattering.
- They can understand the basic principle of time-reversal propagation using phase-conjugate light.
- They will see practical applications of the proposed techniques in animal and human bodies.
- They will learn how deep learning techniques are used in optical transillumination imaging.
