- author = "Marco Virgolin and Ziyuan Wang and Tanja Alderliesten and Peter A. N. Bosman",
- title = "Machine learning for the prediction of pseudorealistic pediatric abdominal phantoms for radiation dose reconstruction",
- journal = "Journal of Medical Imaging",
- year = "2020",
- volume = "7",
- number = "4",
- pages = "046501",
- month = "30 " # jul,
- note = "Winner Silver HUMIES",
- keywords = "genetic algorithms, genetic programming, machine learning, pediatric cancer, radiation treatment, dose reconstruction, phantom, Spleen, Liver, Lawrencium, Databases, Machine learning, Cancer, Image segmentation, Computed tomography, Radiography",
- publisher = "SPIE",
- ISSN = "2329-4302",
-
URL = "
http://www.human-competitive.org/sites/default/files/humies_entry_virgolin.txt",
-
URL = "
http://www.human-competitive.org/sites/default/files/virgolinpaperapreprint_0.pdf",
-
DOI = "
doi:10.1117/1.JMI.7.4.046501",
-
video_url = "http://www.human-competitive.org/sites/default/files/humies2021_virgolin_1.mp4",
-
video_url = "https://youtu.be/jowH-xIvthU",
- size = "25 pages",
-
abstract = "Purpose: Current phantoms used for the dose
reconstruction of long-term childhood cancer survivors
lack individualization. We design a method to predict
highly individualised abdominal three-dimensional (3-D)
phantoms automatically.
Approach: We train machine learning (ML) models to map (2-D) patient features to 3-D organ-at-risk (OAR) metrics upon a database of 60 pediatric abdominal computed tomographies with liver and spleen segmentations. Next, we use the models in an automatic pipeline that outputs a personalized phantom given the patient's features, by assembling 3-D imaging from the database. A step to improve phantom realism (i.e., avoid OAR overlap) is included. We compare five ML algorithms, in terms of predicting OAR left-right (LR), anterior-posterior (AP), inferior-superior (IS) positions, and surface Dice-Sorensen coefficient (sDSC). Furthermore, two existing human-designed phantom construction criteria and two additional control methods are investigated for comparison.
Results: Different ML algorithms result in similar test mean absolute errors: approx 8mm for liver LR, IS, and spleen AP, IS; approx 5mm for liver AP and spleen LR; approx 80 percent for abdomen sDSC; and approx 60 percent to 65 percent for liver and spleen sDSC. One ML algorithm (GP-GOMEA) significantly performs the best for 6/9 metrics. The control methods and the human-designed criteria in particular perform generally worse, sometimes substantially (+5-mm error for spleen IS, -10 percent sDSC for liver). The automatic step to improve realism generally results in limited metric accuracy loss, but fails in one case (out of 60).
Conclusion: Our ML-based pipeline leads to phantoms that are significantly and substantially more individualized than currently used human-designed criteria.",
-
notes = "Modern genetic programming reconstructs radiation dose
for cancer patients better than humans. Most children
treated for cancer suffer adverse effects from their
treatment decades later.
Today: 3D libraries of humans organ layout used by manual expert, replace by genetic programming personalised 3D surrogate (4:42) GP replace whole manual pipeline (4:59) See also \cite{Virgolin:2020:PMB} Human-interpretable. Cutting-edge GP (8:27) https://github.com/marcovirgolin/GP-GOMEA \cite{Virgolin:2017:GECCO} Model size v. accuracy R^2 CWI, Amsterdam UMC, TUDelft, Institute of Oncology Ljubljana, Princess maxima center pediatric oncology, LUMC, UMC Utrecht, Princess Margret Cancer Centre UHN, the university of Manchester.
2021 HUMIES prize giving video https://www.youtube.com/watch?v=jrT0sfq6WjM 43:10 -- 47:08 Childhood radiation treatment 2D images, years later 3D radiation dose outside target human tissue",
