From 79d037601c0ac215816fa983a3e7d6566bac01ae Mon Sep 17 00:00:00 2001 From: Santeri Rytky Date: Wed, 14 Jun 2023 10:25:27 +0300 Subject: [PATCH] Update README.md --- README.md | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) diff --git a/README.md b/README.md index 8b98f53..3c4ac9e 100644 --- a/README.md +++ b/README.md @@ -1,11 +1,12 @@ # BoneEnhance Methods for improving image quality on computed tomography -imaged bone -(c) Santeri Rytky, University of Oulu, 2020-2021 +(c) Santeri Rytky, University of Oulu, 2021-2023 ![Analysis pipeline](https://github.com/MIPT-Oulu/BoneEnhance/blob/master/images/Flowchart.png) ## Background +Clinical cone-beam computed tomography (CBCT) devices are limited to imaging tissues of submillimeter scale. This repository is used to create super-resolution models trained on high-resolution micro-computed tomography (µCT) images. For a detailed description of the method, refer to the publication by Rytky SJO et al. ## Prerequisites - [Anaconda installation](https://docs.anaconda.com/anaconda/install/) @@ -19,13 +20,11 @@ conda env create -f environment.yml ### Model training -- Create a training dataset: input images in folder `images` and target masks in `masks`. -For 2D data, just add the images to the corresponding folders, making sure that the image and mask names match. -For 3D data, create a subfolder for each scan (sample name for the folder), and include the slices in the subfolder. +- Create a training dataset: Use the script `create_training_data.py` to simulate image pairs from high-resolution µCT scans. Set the data and save paths as well as resolution and save parameters at the beginning of the script. - Set the path name for training data in [session.py](../master/rabbitccs/training/session.py) (`init_experiment()` function) -- Create a configuration file to the `experiments/run` folder. Four example experiments are included. +- Create a configuration file to the `experiments/run` folder. Example experiments are included in the folder. All experiments are conducted subsequently during training. ``` @@ -36,9 +35,9 @@ python scripts/train.py ### Inference For 2D prediction, use `inference_tiles_2d.py`. For 3D data, use `inference_tiles_3d.py`. -Running `inference_tiles_large_3d.py` allows to run inference on larger samples, and merge on CPU. +Running `inference_tiles_large_3d.py` allows to run inference on larger samples, and merge on CPU. Using `inference_tiles_large_pseudo3d.py` allows merging 2D predictions on orthogonal planes. Update the `snap` variable, image path and save directory. ## License -This software is distributed under the MIT License. \ No newline at end of file +This software is distributed under the MIT License.