modified: readme.md

readme.md

@@ -5,24 +5,39 @@
 sc-MTOP is an analysis framework based on deep learning and computational pathology. It consists of three steps: 1) Hover-Net based nuclear segmentation and classification; 2) Nuclear morphological and texture feature extraction; 3) Multi-level pairwise nuclear graph construction and spatial topological feature extraction. This framework aims to characterize the tumor ecosystem diversity at the single-cell level. We have a [demo](http://101.132.124.14/#/dashboard) website to show this work.
 
 This is the offical pytorch implementation of sc_MTOP. According to the above description, we use three functions to finish three steps: segment, feature and visual. Note that only segment step support batch processing.
+
 In the segmentation steps, it uses the [HoVer-Net](https://github.com/vqdang/hover_net) model. We doesn't provide the model parameter because it is large. We use the pretrain model based on PanNuke dataset, you can download it from the this [url](https://drive.google.com/file/d/1SbSArI3KOOWHxRlxnjchO7_MbWzB4lNR/view) which provided from the repositories of HoVer-Net.  It needs a folder path including WSI files of all samples as input. We use `.ndpi` file in our work, but we have not tried other formats of wsi files. In theory it supports all file formats allowed by HoVer-Net. The step gives a folder with json file which including all information of cell segmentation and classification.
+
 In the feature step, you have to provide the path of `.json` file, WSI file, and output folder. You can also provide an `.xml` annotation file to only compute features for cells within the annotation range. The annotation is the square shape which only have two pairs of coordinate, and it supports multiple annotation. Similarly, the `.xml` annotation is follow the format of software's `.xml` annotation [ImageScope](https://www.leicabiosystems.com/zh/digital-pathology/manage/aperio-imagescope/). The annotation file from other software may not be support. In the output folder there will be a secondary folder named after the input WSI file. In this folder, it will include three `.csv` files of cell features and one `.csv` file of graph edge information.
+
 In the visual step, we make a visualization of the graph and segmentation. You have to provide the path of feature, WSI file and `.xml` file. The path of feature is the output of feature step, and this step needs the cell ID and the edge information in it. `.xml` file is the annotation file same as the feature step. We only plot the range in the annotation. Note that if the annotation is too large then it will failed.
 
 ## Repository Structure
 `Hover`: the implementation of HoVer-Net, which clone from the offical [implementation](https://github.com/vqdang/hover_net)
+
 `main.py`: main function
+
 `F1_CellSegment.py`: segment step by calling `Hover`.
+
 `F3_FeatureExtract.py`: feature step by calling `WSIGraph.py`.
+
 `F4_Visualization.py`: visual step by calling `utils_xml.py`.
+
 `utils_xml.py`: define some tools to finish visualization.
+
 `WSIGraph.py`: define the process of feature extract.
 
 ## Usage
 Here are some example to use it.
+
 segment step
+
 `python main.py segment --input_dir='./wsi' --output_dir='./output'`
+
 feature step
+
 `python main.py feature --json_file='./output/json/sample.json' --wsi_path='./wsi/sample.ndpi' --output_path='./feature'`
+
 visual step
+
 `python main.py visual --feature_path='./feature/sample' --wsi_path='./wsi/sample.ndpi' --xml_path='./xml/sample.xml'`