From 640c16675bbd8f1528d69dbb875b172f11099667 Mon Sep 17 00:00:00 2001
From: Renato Cuocolo <42030660+rcuocolo@users.noreply.github.com>
Date: Mon, 14 Sep 2020 09:58:24 +0200
Subject: [PATCH] Update README.md

Updated readme to revert to prior version (to correct a push error) and added details on the lesion class file.
---
 README.md | 27 ++++++++++++++++++++++++++-
 1 file changed, 26 insertions(+), 1 deletion(-)

diff --git a/README.md b/README.md
index 1989e69..68aaba9 100644
--- a/README.md
+++ b/README.md
@@ -9,7 +9,9 @@ We encourage the use of this data for radiomics and machine learning investigati
 
 ## Files
 
-The files are in compressed NIFTI (.nii.gz) format and collected in separate folders for T2 and ADC images. Each filename contains the following information (separated by underscores):
+### Lesion masks
+
+The files are in compressed NIFTI (.nii.gz) format and collected in separate folders for T2 and ADC images (n = 299 lesions). Each filename contains the following information (separated by underscores):
 
 - "PROSTATEx Patient ID" + "Finding ID" + "Sequence name" + "ROI"
 
@@ -19,6 +21,29 @@ As sometimes multiple axial ADC/T2 image sequences are available for a patient,
 
 The sequence number corresponds to the number at the beginning of the PROSTATEx dataset DICOM image folders for each patient.
 
+Finally, the "PROSTATEx_classes.csv" file contains the ground truths for all lesions included in the dataset, obtained from the orignial PROSTATEx and PROSTATEx2 data. This file is structured as follows (separated by underscores):
+- ID column = "PROSTATEx Patient ID" + "Finding ID"
+- Clinically Significant column (i.e. Gleason Score > 3+3, Gleason Grade Group/ISUP score > 1) = "True"/"False"
+- Gleason Grade Group/ISUP score column = "No Biopsy" or 1-5
+
+Patients who were classified as PI-RADS score = 2 did not undergo biopsy and their lesions are always assumed to be not clinically significant.
+
+Full exam DICOM files are retrievable through the Cancer Imaging Archive at: [https://wiki.cancerimagingarchive.net/display/Public/SPIE-AAPM-NCI+PROSTATEx+Challenges](https://wiki.cancerimagingarchive.net/display/Public/SPIE-AAPM-NCI+PROSTATEx+Challenges)
+
+### Prostate and zonal anatomy masks
+
+The files are in compressed NIFTI (.nii.gz) format, and all segmentations have been performed on axial T2 images (n = 204 exams). Each filename contains the following information (separated by underscores):
+
+- "PROSTATEx Patient ID"
+
+The masks included in the "mask_prostate" directory represent whole-gland masks (binarized). Those included in "mask_pz" and "mask_tz" are limited respectively to the peripheral and rest (transition zone, central zone, anterior stroma) of the gland, also binarized. The last directory "mas_pz_tz" contains both annotations, with peripheral zone voxels having a value of 2 and the remaining (transition zone, central zone, anterior stroma) a value of 1. The backgorund always has a value of 0.
+ 
+As sometimes multiple axial T2 image sequences are available for a patient, the repository also includes a list in CSV format containing the correct images for each mask. The list is structured as follows (separated by underscores):
+
+- "PROSTATEx Patient ID" + "Sequence name" + "Sequence number"
+
+The sequence number corresponds to the number at the beginning of the PROSTATEx dataset DICOM image folders for each patient.
+
 Full exam DICOM files are retrievable through the Cancer Imaging Archive at: [https://wiki.cancerimagingarchive.net/display/Public/SPIE-AAPM-NCI+PROSTATEx+Challenges](https://wiki.cancerimagingarchive.net/display/Public/SPIE-AAPM-NCI+PROSTATEx+Challenges)
 
 ## License