Table 4.

A summary of articles that use deep learning approaches for image reconstruction in other modalities.

Reference	Brief overview
[56]	Wave flow is a deep learning-based tool. The technology was evaluated using data acquired from wire and cyst phantoms. Both GPU and CPU are supported by the tool.
[57]	For ultrasound image reconstruction, a generative adversarial network (GAN) framework was developed. The suggested framework produced higher-quality ultrasound reconstructions.
[58]	A method for faster B-mode ultrasound imaging. When compared to other current approaches, PSNR, CNR, and SSIM all increased significantly.
[59]	PET image reconstruction using an encoder-decoder system. The use of synthetic data rather than genuine patient data is a drawback.
[60]	To overcome the mismatch of noise levels, a framework for iterative PET reconstruction employing denoising CNN and a local linear fitting function has been developed. It beats traditional approaches in terms of total variation.
[30]	In electromagnetic tomography, a strategy for resolving imaging difficulties has been developed (EMT). Its practicality has been confirmed by preliminary results.
[61]	A diffuse optical tomography (DOT) projection data-based image reconstruction model. Validation of the model clinical situations is required.
[40]	In optical microscopy, the work offered an overview of DNNs. DNNs increase the quality of image reconstruction in optical microscopy, according to the findings.