Integration of machine learning and mechanistic models accurately predicts variation in cell density of glioblastoma using multiparametric MRI

Nathan Gaw, Andrea Hawkins-Daarud, Leland S. Hu, Hyunsoo Yoon, Lujia Wang, Yanzhe Xu, Pamela R. Jackson, Kyle W. Singleton, Leslie C. Baxter, Jennifer Eschbacher, Ashlyn Gonzales, Ashley Nespodzany, Kris Smith, Peter Nakaji, J. Ross Mitchell, Teresa Wu, Kristin R. Swanson, Jing Li

Research output: Contribution to journalArticle

Abstract

Glioblastoma (GBM) is a heterogeneous and lethal brain cancer. These tumors are followed using magnetic resonance imaging (MRI), which is unable to precisely identify tumor cell invasion, impairing effective surgery and radiation planning. We present a novel hybrid model, based on multiparametric intensities, which combines machine learning (ML) with a mechanistic model of tumor growth to provide spatially resolved tumor cell density predictions. The ML component is an imaging data-driven graph-based semi-supervised learning model and we use the Proliferation-Invasion (PI) mechanistic tumor growth model. We thus refer to the hybrid model as the ML-PI model. The hybrid model was trained using 82 image-localized biopsies from 18 primary GBM patients with pre-operative MRI using a leave-one-patient-out cross validation framework. A Relief algorithm was developed to quantify relative contributions from the data sources. The ML-PI model statistically significantly outperformed (p < 0.001) both individual models, ML and PI, achieving a mean absolute predicted error (MAPE) of 0.106 ± 0.125 versus 0.199 ± 0.186 (ML) and 0.227 ± 0.215 (PI), respectively. Associated Pearson correlation coefficients for ML-PI, ML, and PI were 0.838, 0.518, and 0.437, respectively. The Relief algorithm showed the PI model had the greatest contribution to the result, emphasizing the importance of the hybrid model in achieving the high accuracy.

Original languageEnglish (US)
Article number10063
JournalScientific reports
Volume9
Issue number1
DOIs
StatePublished - Dec 1 2019

Fingerprint

Glioblastoma
Cell Count
Magnetic Resonance Imaging
Neoplasms
Machine Learning
Information Storage and Retrieval
Growth
Brain Neoplasms
Radiation
Biopsy

ASJC Scopus subject areas

  • General

Cite this

Integration of machine learning and mechanistic models accurately predicts variation in cell density of glioblastoma using multiparametric MRI. / Gaw, Nathan; Hawkins-Daarud, Andrea; Hu, Leland S.; Yoon, Hyunsoo; Wang, Lujia; Xu, Yanzhe; Jackson, Pamela R.; Singleton, Kyle W.; Baxter, Leslie C.; Eschbacher, Jennifer; Gonzales, Ashlyn; Nespodzany, Ashley; Smith, Kris; Nakaji, Peter; Mitchell, J. Ross; Wu, Teresa; Swanson, Kristin R.; Li, Jing.

In: Scientific reports, Vol. 9, No. 1, 10063, 01.12.2019.

Research output: Contribution to journalArticle

Gaw, N, Hawkins-Daarud, A, Hu, LS, Yoon, H, Wang, L, Xu, Y, Jackson, PR, Singleton, KW, Baxter, LC, Eschbacher, J, Gonzales, A, Nespodzany, A, Smith, K, Nakaji, P, Mitchell, JR, Wu, T, Swanson, KR & Li, J 2019, 'Integration of machine learning and mechanistic models accurately predicts variation in cell density of glioblastoma using multiparametric MRI', Scientific reports, vol. 9, no. 1, 10063. https://doi.org/10.1038/s41598-019-46296-4
Gaw, Nathan ; Hawkins-Daarud, Andrea ; Hu, Leland S. ; Yoon, Hyunsoo ; Wang, Lujia ; Xu, Yanzhe ; Jackson, Pamela R. ; Singleton, Kyle W. ; Baxter, Leslie C. ; Eschbacher, Jennifer ; Gonzales, Ashlyn ; Nespodzany, Ashley ; Smith, Kris ; Nakaji, Peter ; Mitchell, J. Ross ; Wu, Teresa ; Swanson, Kristin R. ; Li, Jing. / Integration of machine learning and mechanistic models accurately predicts variation in cell density of glioblastoma using multiparametric MRI. In: Scientific reports. 2019 ; Vol. 9, No. 1.
@article{2a7aab0730f4418fb7788f6da3d88413,
title = "Integration of machine learning and mechanistic models accurately predicts variation in cell density of glioblastoma using multiparametric MRI",
abstract = "Glioblastoma (GBM) is a heterogeneous and lethal brain cancer. These tumors are followed using magnetic resonance imaging (MRI), which is unable to precisely identify tumor cell invasion, impairing effective surgery and radiation planning. We present a novel hybrid model, based on multiparametric intensities, which combines machine learning (ML) with a mechanistic model of tumor growth to provide spatially resolved tumor cell density predictions. The ML component is an imaging data-driven graph-based semi-supervised learning model and we use the Proliferation-Invasion (PI) mechanistic tumor growth model. We thus refer to the hybrid model as the ML-PI model. The hybrid model was trained using 82 image-localized biopsies from 18 primary GBM patients with pre-operative MRI using a leave-one-patient-out cross validation framework. A Relief algorithm was developed to quantify relative contributions from the data sources. The ML-PI model statistically significantly outperformed (p < 0.001) both individual models, ML and PI, achieving a mean absolute predicted error (MAPE) of 0.106 ± 0.125 versus 0.199 ± 0.186 (ML) and 0.227 ± 0.215 (PI), respectively. Associated Pearson correlation coefficients for ML-PI, ML, and PI were 0.838, 0.518, and 0.437, respectively. The Relief algorithm showed the PI model had the greatest contribution to the result, emphasizing the importance of the hybrid model in achieving the high accuracy.",
author = "Nathan Gaw and Andrea Hawkins-Daarud and Hu, {Leland S.} and Hyunsoo Yoon and Lujia Wang and Yanzhe Xu and Jackson, {Pamela R.} and Singleton, {Kyle W.} and Baxter, {Leslie C.} and Jennifer Eschbacher and Ashlyn Gonzales and Ashley Nespodzany and Kris Smith and Peter Nakaji and Mitchell, {J. Ross} and Teresa Wu and Swanson, {Kristin R.} and Jing Li",
year = "2019",
month = "12",
day = "1",
doi = "10.1038/s41598-019-46296-4",
language = "English (US)",
volume = "9",
journal = "Scientific Reports",
issn = "2045-2322",
publisher = "Nature Publishing Group",
number = "1",

}

TY - JOUR

T1 - Integration of machine learning and mechanistic models accurately predicts variation in cell density of glioblastoma using multiparametric MRI

AU - Gaw, Nathan

AU - Hawkins-Daarud, Andrea

AU - Hu, Leland S.

AU - Yoon, Hyunsoo

AU - Wang, Lujia

AU - Xu, Yanzhe

AU - Jackson, Pamela R.

AU - Singleton, Kyle W.

AU - Baxter, Leslie C.

AU - Eschbacher, Jennifer

AU - Gonzales, Ashlyn

AU - Nespodzany, Ashley

AU - Smith, Kris

AU - Nakaji, Peter

AU - Mitchell, J. Ross

AU - Wu, Teresa

AU - Swanson, Kristin R.

AU - Li, Jing

PY - 2019/12/1

Y1 - 2019/12/1

N2 - Glioblastoma (GBM) is a heterogeneous and lethal brain cancer. These tumors are followed using magnetic resonance imaging (MRI), which is unable to precisely identify tumor cell invasion, impairing effective surgery and radiation planning. We present a novel hybrid model, based on multiparametric intensities, which combines machine learning (ML) with a mechanistic model of tumor growth to provide spatially resolved tumor cell density predictions. The ML component is an imaging data-driven graph-based semi-supervised learning model and we use the Proliferation-Invasion (PI) mechanistic tumor growth model. We thus refer to the hybrid model as the ML-PI model. The hybrid model was trained using 82 image-localized biopsies from 18 primary GBM patients with pre-operative MRI using a leave-one-patient-out cross validation framework. A Relief algorithm was developed to quantify relative contributions from the data sources. The ML-PI model statistically significantly outperformed (p < 0.001) both individual models, ML and PI, achieving a mean absolute predicted error (MAPE) of 0.106 ± 0.125 versus 0.199 ± 0.186 (ML) and 0.227 ± 0.215 (PI), respectively. Associated Pearson correlation coefficients for ML-PI, ML, and PI were 0.838, 0.518, and 0.437, respectively. The Relief algorithm showed the PI model had the greatest contribution to the result, emphasizing the importance of the hybrid model in achieving the high accuracy.

AB - Glioblastoma (GBM) is a heterogeneous and lethal brain cancer. These tumors are followed using magnetic resonance imaging (MRI), which is unable to precisely identify tumor cell invasion, impairing effective surgery and radiation planning. We present a novel hybrid model, based on multiparametric intensities, which combines machine learning (ML) with a mechanistic model of tumor growth to provide spatially resolved tumor cell density predictions. The ML component is an imaging data-driven graph-based semi-supervised learning model and we use the Proliferation-Invasion (PI) mechanistic tumor growth model. We thus refer to the hybrid model as the ML-PI model. The hybrid model was trained using 82 image-localized biopsies from 18 primary GBM patients with pre-operative MRI using a leave-one-patient-out cross validation framework. A Relief algorithm was developed to quantify relative contributions from the data sources. The ML-PI model statistically significantly outperformed (p < 0.001) both individual models, ML and PI, achieving a mean absolute predicted error (MAPE) of 0.106 ± 0.125 versus 0.199 ± 0.186 (ML) and 0.227 ± 0.215 (PI), respectively. Associated Pearson correlation coefficients for ML-PI, ML, and PI were 0.838, 0.518, and 0.437, respectively. The Relief algorithm showed the PI model had the greatest contribution to the result, emphasizing the importance of the hybrid model in achieving the high accuracy.

UR - http://www.scopus.com/inward/record.url?scp=85068933002&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85068933002&partnerID=8YFLogxK

U2 - 10.1038/s41598-019-46296-4

DO - 10.1038/s41598-019-46296-4

M3 - Article

VL - 9

JO - Scientific Reports

JF - Scientific Reports

SN - 2045-2322

IS - 1

M1 - 10063

ER -