Update README.md

README.md

@@ -9,8 +9,8 @@ task_categories:
 ---
 
 ## Pan-Multiplex (Pan-M) dataset
-This dataset was to train the Nimbus model for the publication "Improving cell phenotyping by factoring in spatial marker expression patterns with Nimbus".
-The dataset contains multiplexed images from different modalities, tissues and protein marker panels. It was constructed by semi-automatic pipeline, where the cell types assigned by the authors of the original studies that published the data, where mapped back to their expected marker activity. 
+This dataset was constructed to train the Nimbus model for the publication "Improving cell phenotyping by factoring in spatial marker expression patterns with Nimbus".
+The dataset contains multiplexed images from different modalities, tissues and protein marker panels. It was constructed by a semi-automatic pipeline, where the cell types assigned by the authors of the original studies that published the data, where mapped back to their expected marker activity. In addition, for 3 FoVs of each dataset, 4 expert annotators proofread ~1.1M annotations which served as the gold-standard for assesing the algorithm. 
 More details to the construction of the dataset can be found in the paper. The dataset consists of five subsets named `codex_colon`,`mibi_breast`,`mibi_decidua`,`vectra_colon`,`vectra_pancreas`, each in an individual folder.
 After unzipping, the data should be stored in the following folder structure to use the code provided for [training](https://github.com/angelolab/Nimbus) and [inference](https://github.com/angelolab/Nimbus-Inference). To construct the binary segmentation maps used for training, you can use the code in `segmentation_data_prep.py` and `simple_data_prep.py` in the [training repository](https://github.com/angelolab/Nimbus).