Data from: Denoising, Deblurring, and optical Deconvolution for cryo-ET and light microscopy with a physics-informed deep neural network DeBCR

Datasets deposition for a physics-informed deep learning model DeBCR for microscopy image restorations.

Leveraging optics-based physical models as its foundation, DeBCR model demonstrates superior performance in denoising, optical deconvolution, and more general deblurring (super-resolution, SR) tasks across both light microscopy (LM) and cryo-electron microscopy (EM) modalities.

To evaluate DeBCR on various image restoration tasks, several previously published datasets were assembled, pre-processed and provided as the following:

LM: 2D denoising (files: LM_2D_CARE_*.npz) - low/high exposure confocal dataset of Schmidtea mediterranea (Denoising_Planaria) from the publication of CARE network applied to fluorescent microscopy data (Weigert, Schmidt, Boothe et al., Nature Methods, 2018).

LM: 3D denoising (files: LM_3D_CARE_*.npz) - low/high exposure confocal dataset of Tribolium castaneum (Denoising_Triboleum) from the publication of CARE network applied to fluorescent microscopy data (Weigert, Schmidt, Boothe et al., Nature Methods, 2018).

LM: wide-field data deconvolution (SR) (files: WF_SIM_S.aureus_*.npz) - widefield/SIM dataset of Staphylococcus aureus from the deposition (Pereira & Pinho, Zenodo, 2021).

LM: confocal data deconvolution (SR) (files: CF_STED_FActin_*.npz) - confocal/STED dataset of F-actin from the deposition (Bouchard, Gagné & Lavoie-Cardinal, Zenodo, 2023).

EM: low-frequency denoising (files: EM_low_Tomo110_*.npz) - cryoET dataset of Chlamydomonas reinhardtii cilia (Tomo110) from the cryo-CARE publication (Buchholz et al., IEEE (ISBI), 2019).

EM: high-frequency denoising (files: EM_high_RyR1_*.npz) - cryoET dataset of native vesicles-embedded membrane protein RyR1 (EMPIAR-10452) from the publication (Sanchez, Zhang, Chen et al., Nat Commun, 2020).

and each include train (*_train.npz), test (*_test.npz), and validation (*_val.npz) parts.