add journal ref

Author: ncoudray

DeepPATH_code/00_preprocessing/0d_SortTiles.py

@@ -251,13 +251,13 @@ def sort_subfolders(metadata, load_dic, **kwargs):
     parser.add_argument("--SourceFolder", help="path to tiled images", dest='SourceFolder')
     parser.add_argument("--JsonFile", help="path to metadata json file", dest='JsonFile')
     parser.add_argument("--Magnification", help="magnification to use", type=float, dest='Magnification')
-    parser.add_argument("--MagDiffAllowed", help="difference allwed on Magnification", type=float, dest='MagDiffAllowed')
+    parser.add_argument("--MagDiffAllowed", help="difference allowed on Magnification", type=float, dest='MagDiffAllowed')
     parser.add_argument("--SortingOption", help="see option at the epilog", type=int, dest='SortingOption')
     parser.add_argument("--PercentValid", help="percentage of images for validation (between 0 and 100)", type=float, dest='PercentValid')
     parser.add_argument("--PercentTest", help="percentage of images for testing (between 0 and 100)", type=float, dest='PercentTest')
     parser.add_argument("--PatientID", help="Patient ID is supposed to be the first PatientID characters (integer expected) of the folder in which the pyramidal jpgs are. Slides from same patient will be in same train/test/valid set. This option is ignored if set to 0 or -1 ", type=int, dest='PatientID')
     parser.add_argument("--TMB", help="path to json file with mutational loads; or to BRAF mutations", dest='TMB')
-    parser.add_argument("--nSplit", help="interger n: Split into train/test in n different ways", dest='nSplit')
+    parser.add_argument("--nSplit", help="integer n: Split into train/test in n different ways", dest='nSplit')
     parser.add_argument("--outFilenameStats", help="Check if the tile exists in an out_filename_Stats.txt file and copy it only if it True, or is the expLabel option had the highest probability", dest='outFilenameStats')
     parser.add_argument("--expLabel", help="Index of the expected label within the outFilenameStats file (if only True/False is needed, leave this option empty).", dest='expLabel')
     parser.add_argument("--threshold", help="threshold above which the probability the class should be to be considered as true (if not specified, it would be considered as true if it has the max probability).", dest='threshold')

DeepPATH_code/03_postprocessing/0h_ROC_MultiOutput_BootStrap.py

@@ -600,7 +600,7 @@ def main():
 
 
 	# save data
-	print("******* FP / TP for average probabilitys")
+	print("******* FP / TP for average probability")
 	print(fpr)
 	print(tpr)
 	for i in range(n_classes):
@@ -747,7 +747,7 @@ def main():
 
 
 	# save data
-	print("******* FP / TP for average probabilitys")
+	print("******* FP / TP for average probability")
 	print(fpr)
 	print(tpr)
 	for i in range(n_classes):

DeepPATH_code/README.md

@@ -1,8 +1,18 @@
+If you use this code, please cite:
+
+Nicolas Coudray, Paolo Santiago Ocampo, Theodore Sakellaropoulos, Navneet Narula, Matija Snuderl, David Fenyö, Andre L. Moreira, Narges Razavian, Aristotelis Tsirigos. "Classification and mutation prediction from non–small cell lung cancer histopathology images using deep learning". Nature Medicine, 2018; DOI: 10.1038/s41591-018-0177-5
+
+
+This procedure is based on the inception v3 architecture from google which should be cited as well if you use this code. See [Inception v3](https://github.com/tensorflow/models/blob/f87a58cd96d45de73c9a8330a06b2ab56749a7fa/research/inception/README.md) for information about it.
+
+Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens, Zbigniew Wojna. "Rethinking the Inception Architecture for Computer Vision"
+http://arxiv.org/abs/1512.00567
+
 
 Preliminary comments:
 * For the path, it is advised to always put the full path name and not the relative paths.
-* For all the steps below, always submit the jobs via a qsub script (if on Pheonix) and always check the output and error log files are fine. 
-* This procesude is based on the inception v3 architecture from google. See [Inception v3](https://github.com/tensorflow/models/blob/f87a58cd96d45de73c9a8330a06b2ab56749a7fa/research/inception/README.md) for information about it. 
+* For all the steps below, always submit the jobs via a qsub script (if on Phoenix) and always check the output and error log files are fine. 
+ 
 
 The overall process is:
 * tile the svs images and convert into jpg
@@ -17,7 +27,7 @@ There are several ways to run the pre-processing steps depending on the desired
 * if you aim for a multi-output classification (mutations for example): run the svs tiling step once using all the svs images inside 1 output folder, then run the sorting program using option 10 (only 1 output folder generated with all the jpg. They will be sorted into train/valid/test but not assigned any label yet), then run the TFRecord conversion programs for multi-output classification (will assign the label to each tile)
 * if a pathologist has selected/labelled regions with Aperio (contours of the ROIs saved in xml format), then you run the svs tiling step on each class of xml file, run the sorting program using optin 10 on each of them (the output folder will have the same name as the one where each class has been tiled), and then run the TFRecord conversion programs
 
-Advised folder organization (directorties that may need to be created in plain, those generated by the programs in bold).
+Advised folder organization (directories that may need to be created in plain, those generated by the programs in bold).
 * For classes obtained from json files:
 
 | directories                                     	| Comments           |
@@ -27,7 +37,7 @@ Advised folder organization (directorties that may need to be created in plain,
 | `images/Raw/*svs`                         		| original svs images 		|
 | `images/Raw/*json `                       		| json file from TCGA database 	|
 | `images/<##>pxTiles_<##>Bkg`				| output folder for tiles. Replace ## tile size and background threshold used to run the tiling process` | 
-| **`images/<##>pxTiles_<##>Bkg/<slide name>_files/20.0/<X index>_<Y index>.jpeg`**	| Each svs image will have a folder. Inside, there will be as many sub-folders as mangnification available and the tiles jpeg images inside | 
+| **`images/<##>pxTiles_<##>Bkg/<slide name>_files/20.0/<X index>_<Y index>.jpeg`**	| Each svs image will have a folder. Inside, there will be as many sub-folders as magnification available and the tiles jpeg images inside | 
 | **`images/<##>pxTiles_<##>Bkg/<slide name>_files/10.0/<X index>_<Y index>.jpeg`**	|	|
 | `images/01_Cancer_Tumor/				| output folder for sorted tiled	|
 | **`images/01_Cancer_Tumor/Solid_Normal_Tissue/<t/v/t set>_<slide name>_<X index>_<Y index>.jpg`**	| folders with class names will be generated. Symbolic links to the tiled jpeg will be created and renamed.` 	|