Seeing Through Tissue Fog and Fruit Multisensory Tech for Hidden Objects

Detecting objects hidden within scattering materials like tissue, fog, or fruit poses a challenge for industries such as healthcare, agriculture, and navigation. Light at visible or nearinfrared wavelengths scatters in these media, making it hard to see embedded objects. A new study by Balaji et al. (2025) presents a multisensory approach, inspired by nature, to address this issue, allowing clearer imaging through dense scattering media compared to conventional methods.

Unpacking the Scattering Media Issue

Scattering media, such as biological tissue, foggy air, or fruit, scatter light, disrupting its path and making it difficult to detect objects like tumors in the body or defects in produce. This scattering randomizes light direction over a distance, known as the transport mean free path, after which light spreads out like heat, obscuring objects within the medium. Conventional methods to tackle this, including Magnetic Resonance Imaging (MRI), ultrasound, or radar, have drawbacks. MRI systems are costly and bulky, ultrasound often lacks the contrast needed for tasks like food inspection, and radar struggles with fine details. Optical techniques, like Diffuse Optical Spectroscopy (DOS), model light as a diffusion process to map scattering or absorption properties, but they rely on slow pointbypoint scanning, which limits how much of the medium can be sampled at once.

NatureInspired Multisensory Integration

Balaji et al. (2025) draw from nature to address these sampling constraints, using a multisensory integration strategy similar to how animals combine different senses for better perception. Think of how honeybees use visual cues to approach flowers, then rely on scent to find honey. The study combines two optical methods, a widefield neuromorphic camera and a pointscanning Frequency Modulated Continuous Wave (FMCW) LiDAR, to probe scattering media more effectively.

The neuromorphic camera, inspired by biological vision, captures a broad view by detecting changes in speckle patterns, which are random granular patterns formed when coherent light, like a laser, scatters through a medium. By tuning the laser frequency, the camera observes how these patterns shift over time, revealing the presence of hidden objects through changes in light behavior. This widefield view provides spatial cues to guide the FMCW LiDAR, a precise laserbased system that measures the time light takes to travel through the medium, known as the timeofflight distribution. This combination allows the system to first locate an object broadly, then focus in for detailed imaging, improving sampling density by ten times over standard uniform scanning methods.

WideField Detection Using Neuromorphic Sensing

The neuromorphic camera is key to detecting hidden objects across a wide area. It responds only to changes in light intensity, generating spikes or events, similar to how neurons in the brain fire. When the laser frequency is tuned, speckle patterns fluctuate, and the camera captures these shifts to create a lowresolution image of the embedded object. Balaji et al. (2025) tested this by using a laser at 780 nanometers to illuminate a thick scattering medium, inserting paper cutouts of various shapes as test objects. The camera monitored the speckle patterns at the far end of the medium, accumulating the events to form widefield images of the hidden shapes after applying a denoising filter.

This method performs better than standard approaches that use pulsed light and time-resolved detectors, like SPAD arrays, which often face issues with low resolution and high data volumes. The neuromorphic camera handles rapid light changes with a dynamic range exceeding 120 decibels, meaning it can detect both faint and bright signals effectively, while producing less data for quicker processing.

HighResolution Imaging with FMCW LiDAR

The spatial cues from the neuromorphic camera guide the FMCW LiDAR to focus on specific areas for detailed imaging. FMCW LiDAR uses a tunable laser in a Mach Zehnder interferometer setup, where light splits into a reference beam and a probe beam that passes through the scattering medium. The scattered light emerges at different time delays and interferes with the reference beam, creating beat signals that reveal the timeofflight distribution. This distribution acts like a fingerprint of the medium, showing how light travels through it, which helps identify properties of hidden objects, such as their scattering or absorption characteristics.

Speckle noise, from the interference of randomly phased light waves, can affect these measurements. To manage this, Balaji et al. (2025) used a method inspired by averaging techniques in signal processing, dividing the interferogram, a record of the interference pattern, into multiple segments with randomized time offsets. Each segment is processed to produce independent speckle patterns, which are then averaged to reduce noise. This approach allowed them to estimate the transport properties of scattering phantoms, artificial samples mimicking tissue properties, by fitting the timeofflight distribution to an analytical model. Their findings showed that as scattering increases, the peak arrival time of light shifts, indicating a delay consistent with theoretical expectations.

Multisensory Imaging Through Dense Media

The multisensory approach integrates the widefield neuromorphic camera with the FMCW LiDAR to image through a very thick scattering medium, resembling breast tissue. The process starts with the neuromorphic camera creating a widefield activity map to locate embedded objects, in this case, paraffin wax shapes inserted into the medium. While this initial image lacks fine detail due to crosstalk between adjacent areas, it highlights regions with significant information.

To refine these cues, Balaji et al. (2025) applied a Poisson Surface Reconstruction (PSR) technique, which reconstructs the object’s surface by analyzing the spatial data as a point cloud, a set of 3D points representing the object. PSR uses a mathematical function to define the object’s surface, identifying areas with high curvature where transport properties vary most. This method involves solving an optimization problem to find a function whose changes closely match the spatial data, ultimately generating a mesh of triangles. Areas with denser triangles indicate regions of interest, where the FMCW LiDAR then collects detailed timeofflight measurements.

These measurements are processed to create a dense spatiotemporal dataset, which is integrated over specific time windows to produce detailed images of the embedded objects. This approach distinguishes between scattering and absorption properties, revealing finer spatial details. Compared to a standard camera using LED illumination in the same setup, the multisensory method recovered sharper features, demonstrating its ability to see through dense media with fewer measurements, only 437 compared to 6400 needed for uniform sampling.

Applications Across Different Fields

This multisensory approach offers potential in several areas. In healthcare, it could improve tumor detection by imaging through tissue more effectively than standard optical methods, serving as a less invasive alternative to MRI. In agriculture, it might help identify defects in fruits, ensuring quality control without relying on ultrasound. For autonomous navigation, robots or vehicles could use this technology to see through fog, enabling safer travel in lowvisibility conditions. The reduced number of measurements needed makes it more practical for applications, like functional brain imaging, where quick detection of neuronal activity is important.

Bridging the Gap Future Directions for Multisensory Imaging

Some challenges remain in synchronizing the neuromorphic camera and FMCW LiDAR, which requires careful laser tuning and data processing. The approach also needs further testing with dynamic scattering media, like living tissue, where movement adds complexity. The detection speed of neuromorphic sensors can be limited by their readout circuitry, which processes events on a firstcomefirstserved basis, potentially causing delays if many events occur at once. Future work could focus on simplifying the system, perhaps by developing more compact hardware and exploring faster laser tuning methods, like drive current modulation. Techniques such as spiking neural networks, which process temporal data like the brain, might improve sensitivity and speed. Collaborative efforts between researchers and industry can ensure practical deployment in areas, like medical diagnostics or food safety, while advancing scalability for broader use.

Conclusion

The multisensory approach by Balaji et al. (2025) marks a step in optical imaging through scattering media, combining a neuromorphic camera and FMCW LiDAR to detect hidden objects with greater detail and fewer measurements. Inspired by natures integration of multiple senses, it achieves a tenfold reduction in sampling needs compared to conventional methods. From detecting tumors in tissue to navigating foggy roads, its applications are practical. Addressing synchronization challenges and testing in dynamic conditions will help realize its potential. Researchers can refine the technology, developers can streamline its design, and investors can support its growth, ensuring it benefits society responsibly.

Citation

Balaji M, Ahsanullah D, Rangarajan P. A Multisensory Approach to Probing Scattering Media. arXiv Preprint 2504.13349 [physics.optics] 2025. https://doi.org/10.48550/arXiv.2504.13349

Related Articles

PHP Code Snippets Powered By : XYZScripts.com
★

Track This Company

Create a free account to add companies to your watchlist and get notified when patents, funding, or key developments change.