Interactive machine learning-based multi-label segmentation of solid tumors and organs

Authors

Dimitrios Bounias, National Technical University of Athens
Ashish Singh, University of Pennsylvania
Spyridon Bakas, University of Pennsylvania
Sarthak Pati, University of Pennsylvania
Saima Rathore, University of Pennsylvania
Hamed Akbari, University of Pennsylvania
Michel Bilello, University of Pennsylvania
Benjamin Greenberger, Thomas Jefferson UniversityFollow
Joseph Lombardo, Thomas Jefferson UniversityFollow
Rhea Chitalia, University of Pennsylvania
Nariman Jahani, University of Pennsylvania
Aimilia Gastounioti, University of Pennsylvania
Michelle Hershman, University of Pennsylvania
Leonid Roshkovan, University of Pennsylvania
Sharyn Katz, University of Pennsylvania
Bardia Yousefi, University of Pennsylvania
Carolyn Lou, University of Pennsylvania
Amber Simpson, Queen’s University
Richard Do, Memorial Sloan Kettering Cancer Center
Russell Shinohara, University of Pennsylvania
Despina Kontos, University of Pennsylvania
Konstantina Nikita, National Technical University of Athens
Christos Davatzikos, University of Pennsylvania

Document Type

Article

Publication Date

8-2-2021

Comments

This article is the author’s final published version in Applied Sciences (Switzerland), Volume 11, Issue 16, August 2021, Article number 7488.

The published version is available at https://doi.org/10.3390/app11167488. Copyright © Bounias et al.

Abstract

We seek the development and evaluation of a fast, accurate, and consistent method for general-purpose segmentation, based on interactive machine learning (IML). To validate our method, we identified retrospective cohorts of 20 brain, 50 breast, and 50 lung cancer patients, as well as 20 spleen scans, with corresponding ground truth annotations. Utilizing very brief user training annotations and the adaptive geodesic distance transform, an ensemble of SVMs is trained, providing a patient-specific model applied to the whole image. Two experts segmented each cohort twice with our method and twice manually. The IML method was faster than manual annotation by 53.1% on average. We found significant (p < 0.001) overlap difference for spleen (DiceIML/Dice Manual = 0.91/0.87), breast tumors (DiceIML/DiceManual = 0.84/0.82), and lung nodules (DiceIML/DiceManual = 0.78/0.83). For intra-rater consistency, a significant (p = 0.003) difference was found for spleen (DiceIML/DiceManual = 0.91/0.89). For inter-rater consistency, significant (p < 0.045) differences were found for spleen (DiceIML/DiceManual = 0.91/0.87), breast (DiceIML/DiceManual = 0.86/0.81), lung (DiceIML/DiceManual = 0.85/0.89), the non-enhancing (DiceIML/DiceManual = 0.79/0.67) and the enhancing (DiceIML/DiceManual = 0.79/0.84) brain tumor sub-regions, which, in aggregation, favored our method. Quantitative evaluation for speed, spatial overlap, and consistency, reveals the benefits of our proposed method when compared with manual annotation, for several clinically relevant problems. We publicly release our implementation through CaPTk (Cancer Imaging Phenomics Toolkit) and as an MITK plugin.

Recommended Citation

Bounias, Dimitrios; Singh, Ashish; Bakas, Spyridon; Pati, Sarthak; Rathore, Saima; Akbari, Hamed; Bilello, Michel; Greenberger, Benjamin; Lombardo, Joseph; Chitalia, Rhea; Jahani, Nariman; Gastounioti, Aimilia; Hershman, Michelle; Roshkovan, Leonid; Katz, Sharyn; Yousefi, Bardia; Lou, Carolyn; Simpson, Amber; Do, Richard; Shinohara, Russell; Kontos, Despina; Nikita, Konstantina; and Davatzikos, Christos, "Interactive machine learning-based multi-label segmentation of solid tumors and organs" (2021). Department of Radiation Oncology Faculty Papers. Paper 157.
https://jdc.jefferson.edu/radoncfp/157

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 License.

Language

English