Deep Tomographic Reconstruction

Deep Tomographic Reconstruction is an area where deep learning methods are used for tomographic reconstruction of medical and industrial images. It is a new frontier of the imaging field by utilizing artificial intelligence and machine learning, especially deep artificial neural networks or deep learning, to overcome challenges such as measurement noise, data sparsity, image artifacts, and computational inefficiency. This approach has been applied across various imaging modalities, including CT, MRI, PET, SPECT, ultrasound, and optical imaging. Rapid progress in this field marks a significant shift from traditional reconstruction methods to data-driven approaches since 2016.

Traditional tomographic reconstruction relies on analytic methods such as filtered back-projection, or iterative methods which incrementally compute inverse transformations from measurement data (e.g., Radon or Fourier transform data). However, these approaches are unsatisfactory in challenging scenarios, such as low-dose CT, fast MRI, metal artifacts, patient motion, and so on. In 2016, deep tomographic reconstruction emerged as a new paradigm.

In CT, deep learning models have been particularly effective in reducing radiation exposure while maintaining image quality. Deep networks can also reconstruct decent images from sparsely sampled data without sacrificing diagnostic performance. Deep learning-based generative AI models can reduce CT metal artifacts.

MRI reconstruction has benefited from deep learning by speeding up acquisition speed, which is also referred to as fast MRI. Also, MRI motion artifacts are reduced via deep learning. Deep learning has enabled significant improvements in low-field MRI by enhancing image quality despite lower signal-to-noise ratio (SNR), making these systems clinically viable.

For PET imaging, deep learning models provide substantial improvements in low-dose imaging and motion artifact correction. Also, deep learning helps SPECT for generation of attenuation background. A notable technique for PET denoising involves integrating MR data through multimodal networks, which leverage anatomical information from MRI to enhance PET image quality.

Deep learning enhances ultrasound imaging by reducing speckle noise and motion blur. For ultrasound beamforming, deep neural networks, allows superior image quality with limited data at high speed. There are deep learning-based iPad ultrasound probes on the market. Such a device combines ultrasound imaging with AI-powered features for point-of-care applications.

Diffuse optical tomography, optical coherent tomography and microscopy are improved by deep neural networks beyond traditional methods. Furthermore, deep learning has also enhanced photoacoustic imaging, addressing challenges like high noise, low contrast, and limited resolution.