Artificial Intelligence-based Tumor Segmentation in Mouse Models of Lung Adenocarcinoma.
Journal
Journal of pathology informatics
ISSN: 2229-5089
Titre abrégé: J Pathol Inform
Pays: United States
ID NLM: 101528849
Informations de publication
Date de publication:
2022
2022
Historique:
received:
09
08
2021
accepted:
14
12
2021
entrez:
4
3
2022
pubmed:
5
3
2022
medline:
5
3
2022
Statut:
epublish
Résumé
Mouse models are highly effective for studying the pathophysiology of lung adenocarcinoma and evaluating new treatment strategies. Treatment efficacy is primarily determined by the total tumor burden measured on excised tumor specimens. The measurement process is time-consuming and prone to human errors. To address this issue, we developed a novel deep learning model to segment lung tumor foci on digitally scanned hematoxylin and eosin (H&E) histology slides. Digital slides of 239 mice from 9 experimental cohorts were split into training ( The best model on patch-based validation was DeepLabV3+ using a Resnet-50 backbone, which achieved Dice 0.890 and 0.873 on validation and testing cohort, respectively. This result corresponded to 91.3 Sensitivity and 51.0 PPV in the validation cohort and 93.7 Sensitivity and 51.4 PPV in the testing cohort. False positives could be reduced 10-fold with thresholding artificial intelligence (AI) predicted output by area, without negative impact on Dice Coefficient. Evaluation at various stain normalization strategies did not demonstrate improvement from the baseline model. A robust AI-based algorithm for detecting and segmenting lung tumor foci in the pre-clinical mouse models was developed. The output of this algorithm is compatible with open-source software that researchers commonly use.
Sections du résumé
BACKGROUND
BACKGROUND
Mouse models are highly effective for studying the pathophysiology of lung adenocarcinoma and evaluating new treatment strategies. Treatment efficacy is primarily determined by the total tumor burden measured on excised tumor specimens. The measurement process is time-consuming and prone to human errors. To address this issue, we developed a novel deep learning model to segment lung tumor foci on digitally scanned hematoxylin and eosin (H&E) histology slides.
METHODS
METHODS
Digital slides of 239 mice from 9 experimental cohorts were split into training (
RESULTS
RESULTS
The best model on patch-based validation was DeepLabV3+ using a Resnet-50 backbone, which achieved Dice 0.890 and 0.873 on validation and testing cohort, respectively. This result corresponded to 91.3 Sensitivity and 51.0 PPV in the validation cohort and 93.7 Sensitivity and 51.4 PPV in the testing cohort. False positives could be reduced 10-fold with thresholding artificial intelligence (AI) predicted output by area, without negative impact on Dice Coefficient. Evaluation at various stain normalization strategies did not demonstrate improvement from the baseline model.
CONCLUSIONS
CONCLUSIONS
A robust AI-based algorithm for detecting and segmenting lung tumor foci in the pre-clinical mouse models was developed. The output of this algorithm is compatible with open-source software that researchers commonly use.
Identifiants
pubmed: 35242446
doi: 10.1016/j.jpi.2022.100007
pii: S2153-3539(22)00007-4
pmc: PMC8860735
doi:
Types de publication
Journal Article
Langues
eng
Pagination
100007Informations de copyright
© 2022 Published by Elsevier Inc. on behalf of Association for Pathology Informatics.
Références
J Pathol Inform. 2018 Nov 14;9:38
pubmed: 30607305
JAMA Netw Open. 2020 Jun 1;3(6):e205842
pubmed: 32492161
J Pathol Inform. 2013 Sep 27;4:27
pubmed: 24244884
CA Cancer J Clin. 2021 May;71(3):209-249
pubmed: 33538338
J Med Imaging (Bellingham). 2021 Mar;8(2):027501
pubmed: 33681410
Comput Intell Neurosci. 2021 Apr 9;2021:5580914
pubmed: 33897774
Genes Dev. 2001 Dec 15;15(24):3243-8
pubmed: 11751630
EBioMedicine. 2021 May;67:103388
pubmed: 34000621
Mod Pathol. 2021 Dec;34(12):2098-2108
pubmed: 34168282
Sci Rep. 2020 Apr 8;10(1):6047
pubmed: 32269234
Surg Oncol Clin N Am. 2016 Jul;25(3):447-68
pubmed: 27261908
PLoS One. 2018 Aug 23;13(8):e0202708
pubmed: 30138413
Nat Med. 2018 Oct;24(10):1559-1567
pubmed: 30224757
Toxicol Pathol. 2021 Jun;49(4):815-842
pubmed: 33618634
Oncogene. 2006 Feb 23;25(8):1277-80
pubmed: 16247444
Comput Methods Programs Biomed. 2021 Mar;200:105837
pubmed: 33221056
Nat Protoc. 2009;4(7):1064-72
pubmed: 19561589
BMC Med Imaging. 2015 Aug 12;15:29
pubmed: 26263899
Cancer Manag Res. 2021 Jun 10;13:4605-4617
pubmed: 34140807
BMC Med. 2021 Mar 29;19(1):80
pubmed: 33775248
Sci Rep. 2017 Dec 4;7(1):16878
pubmed: 29203879
IEEE Trans Med Imaging. 2016 Aug;35(8):1962-71
pubmed: 27164577
Front Pharmacol. 2020 Oct 02;11:572372
pubmed: 33132910
Comput Methods Programs Biomed. 2021 Jun;204:106047
pubmed: 33789213
Cell. 2019 Feb 21;176(5):998-1013.e16
pubmed: 30712876
IEEE J Biomed Health Inform. 2021 Feb;25(2):429-440
pubmed: 33216724
Open Biol. 2021 Jan;11(1):200247
pubmed: 33435818
PLoS One. 2017 Jul 6;12(7):e0180540
pubmed: 28683129
Sci Rep. 2020 Jan 30;10(1):1504
pubmed: 32001752
Med Image Anal. 2017 Jul;39:194-205
pubmed: 28521242