🔢 mathematics

Complexity of Radon transforms

This paper investigates the emergence of complexity in Radon transforms, particularly when dealing with pure-real functions, by analyzing singular points via the Fourier slice theorem and proposing methods such as hybrid Wigner-like functions to naturally introduce this complexity, thereby enhancing reconstruction and optimization procedures for ill-posed inverse problems.

Original authors: I. V. Anikin

Published 2026-10-14

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: I. V. Anikin

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 trying to see the inside of a solid object without ever cutting it open. This is the fundamental challenge behind computed tomography, or CT scanning, a technology that has revolutionized medicine and materials science by allowing us to peer inside the human body or complex machinery. The mathematical engine that makes this possible is known as the Radon transform. In simple terms, this process takes a three-dimensional object and creates a series of two-dimensional "shadows" or projections from different angles. To reconstruct the original object from these shadows, scientists use an inverse process, essentially working backward from the shadows to the source. For decades, this reconstruction has been treated as a difficult but solvable puzzle, often requiring special mathematical tricks to smooth out errors and make the image clear. However, a new analysis suggests that the standard way we think about this mathematical reversal might be missing a crucial piece of the picture, one that involves a hidden layer of complexity.

In a recent study, a researcher at the Bogoliubov Laboratory of Theoretical Physics in Dubna, Russia, has re-examined the mathematical foundations of this reconstruction process. The work focuses on a specific relationship known as the Fourier slice theorem, which acts as a bridge connecting the original object to its shadow projections. The author demonstrates that when we look closely at the mathematical points where the data becomes singular or degenerate—essentially the edges or centers where the geometry of the projection changes behavior—standard methods fail to capture the full story. The study argues that if the original object is described by a simple, real-valued function, the complexity of the Radon transforms becomes a significant mathematical issue. Rather than being an inherent physical property, this complexity arises as a necessary mathematical requirement for regularization to handle singular points and ensure a valid reconstruction.

The researcher found that this complexity arises from a phenomenon called non-trivial holonomy. To understand this, imagine rotating a line around an object to take measurements. If the object has different properties in different regions, or if there are hidden "defects" within the material, the mathematical result of the measurement does not simply return to its starting value after a full rotation. Instead, it arrives at a slightly different state, creating a mismatch that standard linear math cannot resolve on its own. This mismatch forces the introduction of complex numbers into the equation to balance the scales. The study shows that ignoring this leads to an incomplete reconstruction, while embracing it allows for a more accurate and optimized recovery of the object's internal structure.

To make this complexity work in practice, the paper proposes a new method that blends spatial coordinates with momentum coordinates, creating what is called a hybrid function. This approach is similar to how a musician might analyze a sound not just by its pitch, but by how that pitch changes over time, but here it is applied to the geometry of the object itself. By treating the object's data as a combination of position and a related momentum-like variable, the researcher can generate the necessary complex values naturally during the calculation. This hybrid method allows the reconstruction algorithm to handle the "defects" or irregularities in the object that would otherwise cause the image to blur or distort. The study demonstrates that this technique is particularly effective when dealing with the discrete slices of data that modern CT scanners actually produce, turning a theoretical mathematical necessity into a practical tool for better imaging.

The implications of this work extend beyond just making sharper pictures. The author suggests that acknowledging this hidden complexity changes how we approach "ill-posed" problems, which are mathematical tasks where small errors in the input can lead to huge errors in the output. By including the complex contributions derived from the hybrid functions, the reconstruction process becomes more stable and robust. This means that the mathematical framework used to visualize the internal structure of objects can be improved without needing more expensive hardware, simply by refining the way the data is processed. The study does not claim to have solved every problem in medical imaging, but it provides a clear, mathematically rigorous path forward for handling the singularities and defects that have long complicated the reconstruction of complex objects.

Ultimately, this research highlights that the path from a shadow back to the object is not a straight line. It involves navigating through points where the geometry breaks down and requires a more sophisticated mathematical language to traverse. By introducing these hybrid functions and accepting the natural complexity of the Radon transform, the study offers a new way to ensure that what we see in a reconstructed image is a true reflection of the hidden world inside. It is a reminder that even in well-established fields like medical imaging, there are still deep mathematical layers waiting to be uncovered, layers that, once understood, can make the invisible visible with greater clarity than ever before.

Technical Summary: Complexity of Radon Transforms

Problem Statement
The paper addresses the reconstruction problem in computed tomography (CT) technologies, which relies on the application of direct and inverse Radon transforms. While the inverse Radon transform is essential for visualizing the internal structure of objects, it is an ill-posed problem requiring regularization. The author focuses on the "universal representation" of the inverse Radon transform, previously proposed in works [3, 4], which introduces a new contribution compared to standard methods based on Courant-Hilbert identities. This universal representation involves integration with a complex measure. A central problem identified is that if the outset (origin) function generating the Radon image is a pure-real function, the necessity and origin of the required complexity in the Radon transforms become unclear. The paper seeks to analyze the sources of this complexity, specifically investigating degenerated (singular) points in the Fourier slice theorem and proposing methods to naturally generate the necessary complex components on intermediate calculation stages.

Methodology
The analysis proceeds through several theoretical steps:

Fourier Slice Theorem Analysis: The author re-examines the Fourier slice theorem, which relates the Fourier image of an outset function to its direct Radon image via a one-dimensional Fourier transformation. They specifically analyze "degenerated" or singular points, particularly the axis singularity at $\lambda = 0$ (in polar coordinates) and the behavior at $\lambda = +\infty$ . The study demonstrates that regularization of these points leads to complex continuations of the Radon radial component (e.g., via $\epsilon$ -regularization resulting in terms like $1/(\eta - i\epsilon)$ ).
Linearity and Defect Analysis: The paper investigates the linearity of Radon transforms when the outset function is a combination of functions localized in different spatial regions (e.g., separated by characteristic indicators $\Theta_I$ and $\Theta_{III}$ ). By performing angular shifts (rotations) on the Radon functional, the author demonstrates the existence of "non-trivial holonomy." This phenomenon arises when the function contains "defects" or localized anomalies, leading to a discrepancy between the Radon transform of the combined function and the sum of its parts under specific rotational shifts.
Hybrid Function Construction: To naturally generate the required complexity, the author proposes introducing "hybrid (Wigner-like) functions." These functions depend on both spatial and momentum coordinates. The method involves taking the direct Fourier transform of the outset function with respect to one coordinate (e.g., $x_3$ ) to create a hybrid function $F(x_1, x_2; k)$ . In the context of discrete slice approximations (common in CT), this corresponds to a Fourier series expansion.
Complex Radon Transforms: The Radon transform is then applied to these hybrid functions. Since the hybrid functions are inherently complex (even if the original spatial function is real), their Radon images possess both real and imaginary parts.

Key Contributions and Results

Identification of Singularities as Sources of Complexity: The paper establishes that the regularization of singular points in the Fourier slice theorem (specifically the $\lambda=0$ and $\lambda=\infty$ limits) necessitates complex measures in the inversion formula. The imaginary part of the resulting distribution (related to $\delta_+(\eta)$ ) is shown to be crucial for the universal inversion, particularly when angular integration is restricted.
Non-Trivial Holonomy: The author demonstrates that Radon transforms exhibit non-trivial holonomy when applied to functions with spatial defects. This non-triviality generates complexity and allows for the extraction of "defect" effects from a common background, which is significant for reconstruction optimization.
Hybrid Function Framework: A primary contribution is the introduction of hybrid (Wigner-like) functions as a mechanism to ensure the Radon transform is complex-valued by construction. The paper shows that the inverse Radon operator acting on these functions splits into two contributions:
- $F_S$ : A term related to the real part of the Radon image (principal value integration).
- $F_A$ : A term related to the imaginary part of the Radon image (involving derivatives at $\eta=0$ ).
Reconstruction Scheme: The paper outlines a reconstruction scheme where the complex Radon image of the hybrid function is inverted to recover the hybrid function, which is then transformed back to the original spatial function $f(x_1, x_2, x_3)$ . This scheme aggregates 2D slices using a continuous momentum parameter $k$ , which the author notes is convenient for optimization procedures.

Significance and Claims
The paper claims that the discussed complexity is not merely a mathematical artifact but provides an "additional contribution" to the inverse Radon transforms that is essential for reconstruction and optimization procedures, particularly within the framework of "incorrect" (ill-posed) problems.

The author asserts that the introduction of hybrid functions ensures the corresponding complexity "naturally," avoiding ad-hoc additions.
They state that this approach offers a new perspective on the universal inverse Radon transform, distinct from standard methods relying on Courant-Hilbert identities.
The work suggests that these methods can be effectively used for practical reconstruction tasks, specifically by leveraging the complex nature of the Radon image to improve visualization and handle defects within the object being scanned.
The paper concludes that the complexity of Radon transforms, often questioned when dealing with real-valued outset functions, is a necessary feature derived from singularities and the structure of the inversion operator, and that the proposed hybrid function method provides a robust tool for managing this complexity in reconstruction algorithms.

More like this