M. Haltmeier,O. Scherzer,G. Zangerl
Abstract:Fourier reconstruction algorithms significantly outperform conventional back-projection algorithms in terms of computation time. In photoacoustic imaging, these methods require interpolation in the Fourier space domain, which creates artifacts in reconstructed images. We propose a novel reconstruction algorithm that applies the one-dimensional nonuniform fast Fourier transform to photoacoustic imaging. It is shown theoretically and numerically that our algorithm avoids artifacts while preserving the computational effectiveness of Fourier reconstruction.
What problem does this paper attempt to address?
This paper attempts to solve the problem of image artifacts in Photoacoustic Imaging (PAI) based on the Fourier reconstruction algorithm. Specifically:
1. **Limitations of traditional methods**: Although the traditional back - projection algorithm has a long calculation time, it is widely used in practical applications. However, the reconstruction method based on the Fourier domain, although computationally efficient, requires interpolation in the Fourier space, which will lead to artifacts in the reconstructed image.
2. **The proposed new method**: In order to solve the above problems, the author proposes a new reconstruction algorithm, which utilizes the One - Dimensional Nonuniform Fast Fourier Transform (NUFFT). This new method not only avoids the generation of artifacts but also maintains the high efficiency of the Fourier reconstruction method.
3. **Theoretical and numerical verification**: Through theoretical analysis and numerical experiments, the author proves that the proposed algorithm can effectively reduce the artifacts in the reconstructed image without significantly increasing the computational cost, thereby improving the image quality.
### Formula summary
- **Inverse Fourier transform formula**:
\[
(Ff)(K_x, K_y)=\frac{2K_y}{\sqrt{|K_x|^2 + K_y^2}}\left(FQf\right)\left(K_x,\text{sign}(K_y)\sqrt{|K_x|^2 + K_y^2}\right)
\]
where \(K=(K_x, K_y)\in\mathbb{R}^{d - 1}\times\mathbb{R}\), and \(F\) represents the \(d\)-dimensional Fourier transform.
- **Nonuniform Fast Fourier Transform**:
\[
T[g](\omega_k):=\sum_{n = 0}^{N-1}e^{-i\omega_k n\frac{2\pi}{N}}g_n,\quad k =-\frac{N}{2},\ldots,\frac{N}{2}-1
\]
Directly calculating the above discrete Fourier transform requires \(O(N^2)\) operations, while using the classical Fast Fourier Transform (FFT) can reduce it to \(O(N\log N)\) operations. However, the classical FFT is only applicable to the case of equally spaced nodes, while the non - uniform FFT can handle arbitrary nodes.
### Main contributions
- **Innovation points**: The main innovation of this paper is to apply the non - uniform FFT to the reconstruction process of photoacoustic imaging. By using multiple one - dimensional non - uniform FFTs to replace a single high - dimensional non - uniform FFT, the cost of numerical calculation is reduced.
- **Application scenarios**: This algorithm is especially suitable for photoacoustic imaging under the planar recording geometric structure. It can effectively improve the quality of the reconstructed image while maintaining high - efficiency computational performance.
### Conclusion
This paper successfully solves the problem of artifacts in the Fourier reconstruction method in photoacoustic imaging, proposes an efficient reconstruction algorithm based on non - uniform FFT, and verifies its superiority through theory and experiments. This result is of great significance for improving the application effect of photoacoustic imaging technology.