Accurate Patient-Specific Machine Learning Models of Glioblastoma Invasion Using Transfer Learning

L.S. Hu; H. Yoon; J.M. Eschbacher; L.C. Baxter; A.C. Dueck; A. Nespodzany; K.A. Smith; P. Nakaji; Y. Xu; L. Wang; J.P. Karis; A.J. Hawkins-Daarud; K.W. Singleton; P.R. Jackson; B.J. Anderies; B.R. Bendok; R.S. Zimmerman; C. Quarles; A.B. Porter-Umphrey; M.M. Mrugala; A. Sharma; J.M. Hoxworth; M.G. Sattur; N. Sanai; P.E. Koulemberis; C. Krishna; J.R. Mitchell; T. Wu; N.L. Tran; K.R. Swanson; J. Li

doi:10.3174/ajnr.A5981

Abstract

BACKGROUND AND PURPOSE: MR imaging–based modeling of tumor cell density can substantially improve targeted treatment of glioblastoma. Unfortunately, interpatient variability limits the predictive ability of many modeling approaches. We present a transfer learning method that generates individualized patient models, grounded in the wealth of population data, while also detecting and adjusting for interpatient variabilities based on each patient's own histologic data.

MATERIALS AND METHODS: We recruited patients with primary glioblastoma undergoing image-guided biopsies and preoperative imaging, including contrast-enhanced MR imaging, dynamic susceptibility contrast MR imaging, and diffusion tensor imaging. We calculated relative cerebral blood volume from DSC-MR imaging and mean diffusivity and fractional anisotropy from DTI. Following image coregistration, we assessed tumor cell density for each biopsy and identified corresponding localized MR imaging measurements. We then explored a range of univariate and multivariate predictive models of tumor cell density based on MR imaging measurements in a generalized one-model-fits-all approach. We then implemented both univariate and multivariate individualized transfer learning predictive models, which harness the available population-level data but allow individual variability in their predictions. Finally, we compared Pearson correlation coefficients and mean absolute error between the individualized transfer learning and generalized one-model-fits-all models.

RESULTS: Tumor cell density significantly correlated with relative CBV (r = 0.33, P < .001), and T1-weighted postcontrast (r = 0.36, P < .001) on univariate analysis after correcting for multiple comparisons. With single-variable modeling (using relative CBV), transfer learning increased predictive performance (r = 0.53, mean absolute error = 15.19%) compared with one-model-fits-all (r = 0.27, mean absolute error = 17.79%). With multivariate modeling, transfer learning further improved performance (r = 0.88, mean absolute error = 5.66%) compared with one-model-fits-all (r = 0.39, mean absolute error = 16.55%).

CONCLUSIONS: Transfer learning significantly improves predictive modeling performance for quantifying tumor cell density in glioblastoma.

ABBREVIATIONS:

FA: fractional anisotropy
GBM: glioblastoma
LOOCV: leave-one-out cross-validation
MD: mean diffusivity
OMFA: one-model-fits-all
rCBV: relative CBV
T1 + C: T1-weighted postcontrast
TCD: tumor cell density
TL: transfer learning
EPI+C: post-contrast T2*WI

Footnotes

Leland S. Hu, Hyunsoo Yoon. Kristin R. Swanson, and Jing Li contributed equally to this work.
Disclosures: Leland S. Hu—RELATED: Grant: National Institutes of Health, Comments: National Institutes of Health/National Institute of Neurological Disorders and Stroke: R21-NS082609, National Institutes of Health/National Cancer Institute: U01-CA220378, R01-CA221938, P50-CA108961*; UNRELATED: Grants/Grants Pending: National Institutes of Health, Comments: National Institutes of Health/National Institute of Neurological Disorders and Stroke: R21NS082609, National Institutes of Health/National Cancer Institute: U01-CA220378*; Patents (Planned, Pending or Issued): patent application title: Methods for Using Machine Learning and Mechanistic Models for Cell Density Mapping of Glioblastoma with Multiparametric MRI, Patent Application No. 62/684,096, application type: Provisional, Country: USA, filing date: June 12, 2018, Mayo Clinic Case No. 2017–498. Hyunsoo Yoon—RELATED: Grant: National Institutes of Health U01.* Leslie C. Baxter—RELATED: Grant: several grants from the National Cancer Institute (National Institutes of Health).* Amylou C. Dueck—RELATED: Grant: National Institutes of Health, Comments: R21-NS082609 and U01-CA220378.* Peter Nakaji—UNRELATED: Consultancy: Carl Zeiss Meditec, Comments: microscope company that does tumor imaging and fluorescence work; Payment for Lectures Including Service on Speakers Bureaus: Carl Zeiss Meditec, Comments: microscope company that does tumor imaging and fluorescence work, for which I sometimes lecture; Patents (Planned, Pending or Issued): GT Medical Technologies, Comments: creates brachytherapy solutions for recurrent brain tumors, not related to current work. I was a founder and inventor; Stock/Stock Options: GT Medical Technologies, Comments: creates brachytherapy solutions for recurrent brain tumors, not related to current work; *Money paid to the individual (P.N.). Other: Stryker, SpiWay, Thieme. Yanzhe Xu—RELATED: Grant: National Institutes of Health U01.* Lujia Wang—RELATED: Grant: National Institutes of Health U01.* Andrea J. Hawkins-Daarud—RELATED: Grant: National Institutes of Health.* Pamela R. Jackson—RELATED: Grant: National Institutes of Health.* Jing Li—RELATED: Grant: R21-NS082609, U01-CA220378.* Teresa Wu—RELATED: Grant: R21-NS082609, U01-CA220378.* Chad Quarles—RELATED: Grant: National Institutes of Health.* Kristin R. Swanson—RELATED: Grant: National Institutes of Health, James S. McDonnell Foundation, Ivy Foundation, Arabidopsis Biological Resource Center.* Mithun G. Sattur—UNRELATED: Stock/Stock Options: MRI interventions. *Money paid to the institution.
This work was supported by R21-NS082609, R01-CA221938, U01-CA220378, P50-CA108961, R01-CA158079 of the National Cancer Institute; the Mayo Clinic Foundation; the James S. McDonnell Foundation; the Ivy Foundation; and the Arizona Biomedical Research Commission.