🔬 applied physics

Absorption and Phase-Contrast Microtomography Using Direct X-ray Detection With COTS CMOS Sensors

This paper presents a low-cost, high-resolution X-ray microtomography system that utilizes commercial off-the-shelf CMOS sensors as direct detectors to achieve both absorption and phase-contrast imaging of soft tissues, offering a simpler and more accessible alternative to synchrotron or nanofocus systems.

Original authors: Damian L. Corzi, Jose Lipovetzky, Fabricio Alcalde Bessia, German Mato, Andres Cicuttin, Maria L. Crespo, Martin Perez, Mariano Gomez Berisso

Published 2026-05-29

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Damian L. Corzi, Jose Lipovetzky, Fabricio Alcalde Bessia, German Mato, Andres Cicuttin, Maria L. Crespo, Martin Perez, Mariano Gomez Berisso

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you want to take a 3D "CT scan" of something tiny, like a microchip or a bug, to see what's inside without cutting it open. Usually, doing this requires massive, expensive machines the size of a building (like synchrotrons) or incredibly complex lenses.

This paper describes a clever, low-cost alternative: a homemade X-ray scanner built using off-the-shelf computer camera sensors (the same kind found in webcams or smartphones) and a small, affordable X-ray tube.

Here is how they did it, explained through simple analogies:

1. The "Naked Eye" Sensor

Normally, X-ray cameras use a special glowing screen (a scintillator) to turn invisible X-rays into visible light, which a camera then takes a picture of. It's like looking at a shadow through a foggy window.

The authors removed the "foggy window." They took a standard computer camera sensor, stripped off its glass cover, and let the X-rays hit the sensor directly. Because these sensors have tiny pixels (smaller than a human hair), they can see incredibly fine details without needing any fancy lenses or light-converting screens. It's like switching from looking at a shadow through a blurry curtain to seeing the shadow cast directly on a high-definition TV screen.

2. The "Shadow and the Ripple" (Two Ways to See)

The system can take pictures in two different modes, depending on how far the object is from the camera:

Mode A: The Shadow (Absorption Contrast)
- How it works: The object is placed very close to the sensor.
- The Analogy: Think of holding your hand close to a wall under a flashlight. You see a dark shadow where your hand blocks the light.
- What it sees: This is great for dense things like metal or plastic. The authors used this to look inside a computer chip, clearly seeing the tiny gold wires connecting the parts. It's like seeing the "bones" of the object.
Mode B: The Ripple (Phase Contrast)
- How it works: The object is moved far away from the sensor.
- The Analogy: Imagine dropping a pebble in a pond. The water ripples outward. If you look at the ripples from far away, you can see the shape of the pebble even if the water is clear.
- What it sees: This is the magic trick. Soft things (like a wasp's antenna or insect tissue) don't block X-rays well, so they usually look invisible in a standard shadow picture. But when placed far away, the X-rays create a "ripple" pattern around the edges of the soft tissue. This makes the invisible visible, revealing the delicate boundaries of a wasp's head that would otherwise look like a blank spot.

3. The "Aging Sensor" Problem and Fix

There was a catch. X-rays are harsh. If you leave a camera sensor in the X-ray beam for a long time (which is needed to take hundreds of pictures for a 3D scan), the sensor gets "tired" and starts producing noise (static), like an old TV with bad reception.

The Problem: The longer the scan goes, the "dirtier" the image gets.
The Fix: The authors wrote a smart computer program that acts like a dynamic eraser. Instead of just taking one "clean" picture at the start to compare against, the program watches how the sensor degrades over time. It creates a moving correction map that updates with every single picture, subtracting the "tiredness" noise in real-time. This allowed them to take long scans without the image getting ruined.

4. The Results

Using this simple setup, they achieved two things:

Microchip Inspection: They saw gold wires inside a computer chip that were only 18 micrometers wide (thinner than a human hair).
Insect Anatomy: They created a clear 3D model of a wasp's head, seeing soft parts that standard X-rays would have missed completely.

The Bottom Line

This paper proves you don't need a billion-dollar machine to do high-resolution 3D X-ray imaging. By using a cheap, standard computer camera sensor, a small X-ray tube, and some clever math to fix sensor noise, they built a system that rivals much more expensive setups. It's a "DIY" approach to high-tech science that makes seeing the invisible accessible to regular labs.

Technical Summary: Absorption and Phase-Contrast Microtomography Using Direct X-ray Detection With COTS CMOS Sensors

Problem Statement
High-resolution X-ray microtomography typically relies on expensive, complex infrastructure, such as synchrotron radiation facilities or specialized nanofocus sources coupled with scintillators and complex optical systems. While these setups achieve sub-micrometer resolution, they are often inaccessible for standard laboratory use due to cost and complexity. Furthermore, conventional absorption-contrast tomography struggles to visualize low-density materials (low-Z) and soft tissues, which provide insufficient attenuation contrast. The paper addresses the need for a simpler, lower-cost laboratory-scale system capable of high-resolution imaging that can also leverage phase-contrast mechanisms to visualize soft boundaries without the need for specialized optics or synchrotron access.

Methodology
The authors developed a high-resolution microtomography system utilizing Commercial Off-The-Shelf (COTS) CMOS image sensors as direct X-ray detectors. The experimental setup consists of the following key components:

Source: A microfocus tungsten-anode X-ray source (Micro X-ray Inc. Microbox) with a 5 µm focal spot, operating in the 10–90 kV range.
Detector: A Micron MT9M001 CIS with a 1280 × 1024 pixel array and a 5.2 µm pixel pitch. The sensor's glass cover was removed to eliminate attenuation and scattering effects, allowing direct X-ray detection.
Geometry: A cone-beam geometry where the object-to-sensor distance is adjustable. This allows the system to switch between geometric magnification for absorption contrast and propagation-based phase contrast by increasing the sample-to-detector distance.
Shielding and Cooling: The setup is housed in a lead-lined cabinet divided by a collimator to minimize secondary photons. A Peltier-based cooling system manages sensor temperature, and the entire assembly is isolated on a foam mat to reduce vibrations.
Data Acquisition and Correction: To mitigate shot noise, multiple 1-second frames are averaged. A critical methodological contribution is a dynamic flat-field correction algorithm. Because prolonged X-ray exposure induces Total Ionizing Dose (TID) effects, causing a linear increase in dark current and flat-field drift, the authors acquire reference images at the start and end of the scan. They use these to interpolate correction parameters ( $K_D$ and $K_F$ ) for every projection, effectively compensating for sensor degradation during long acquisitions.
Reconstruction: Tomographic reconstruction is performed using the Filtered Backprojection (FDK) algorithm via the Reconstruction Toolkit (RTK), preceded by sinogram analysis to correct for mechanical misalignment.

Key Contributions

Direct Detection with COTS Sensors: The work demonstrates that removing the glass cover of a standard CMOS sensor allows it to function as a high-resolution direct X-ray detector, achieving voxel sizes between 3.9 µm and 5.2 µm without scintillators or optical coupling.
Dynamic Flat-Field Correction: The authors propose and validate a linear interpolation method to correct for TID-induced sensor degradation (dark current increase and flat-field drift) in real-time during long tomographic scans. This overcomes a major limitation of using consumer-grade hardware for high-dose applications.
Dual-Mode Imaging: The system successfully operates in two distinct modes:
- Absorption Contrast: By placing the object close to the sensor, the system achieves high-resolution imaging of high-attenuation materials.
- Propagation-Based Phase Contrast: By increasing the object-to-detector distance, the system captures interference fringes, enabling the visualization of low-attenuation soft tissues that are invisible in conventional radiography.

Results
The system was validated using two distinct samples:

Integrated Circuit (Absorption Contrast): An AD8615 surface-mount operational amplifier was imaged. The system resolved internal metallic structures, including gold bond wires with a measured diameter of approximately 18 µm. This demonstrates the system's capability for failure analysis in microelectronics.
Insect Specimen (Phase Contrast): A wasp head was imaged. When placed at a distance of 19 cm (enabling phase contrast), the system clearly visualized soft tissue boundaries and antennae. In contrast, when the same specimen was placed at 25 mm (eliminating phase effects), the antennae and soft boundaries became indistinguishable due to low absorption contrast, and noise levels near the mounting needle increased, preventing successful segmentation.

Significance and Claims
The paper claims that COTS CMOS sensors represent a viable, accessible, and cost-effective alternative for laboratory-scale high-resolution microtomography. The authors emphasize that their solution avoids the complexity and high cost of synchrotron facilities or nanofocus systems while still providing voxel resolutions in the few-micrometer range over a several-millimeter field of view.

The work highlights that by leveraging the intrinsic resolution of the sensor and the coherence of a microfocus source, it is possible to perform propagation-based phase-contrast imaging without complex optics. The authors conclude that this approach makes high-resolution 3D imaging of both high-attenuation (e.g., microelectronics) and low-attenuation (e.g., biological soft tissue) samples accessible to standard research laboratories, provided that specific corrections for sensor aging are implemented.