Support vector machine for breast cancer classification using diffusion‐weighted MRI histogram features: Preliminary study

Journal of Magnetic Resonance Imaging - Tập 47 Số 5 - Trang 1205-1216 - 2018
Igor Vidić1, Liv Egnell2,1, Neil P. Jerome2,3, Jose R. Teruel4,5, Torill Eidhammer Sjøbakk3, Agnes Østlie2, Hans E. Fjøsne6,7, Tone F. Bathen3, Pål Erik Goa2,1
1Department of Physics, NTNU, Norwegian University of Science and Technology, Trondheim, Norway
2Clinic of Radiology and Nuclear Medicine St. Olavs University Hospital Trondheim Norway
3Department of Circulation and Medical Imaging, NTNU–Norwegian University of Science and Technology, Trondheim, Norway
4Department of Radiation Oncology, NYU Langone Medical Center, New York, New York, USA
5Department of Radiology, University of California San Diego, La Jolla, California, USA
6Department of Cancer Research and Molecular Medicine, NTNU, Norwegian University of Science and Technology, Trondheim, Norway
7Department of Surgery, St. Olavs University Hospital, Trondheim, Norway

Tóm tắt

BackgroundDiffusion‐weighted MRI (DWI) is currently one of the fastest developing MRI‐based techniques in oncology. Histogram properties from model fitting of DWI are useful features for differentiation of lesions, and classification can potentially be improved by machine learning.PurposeTo evaluate classification of malignant and benign tumors and breast cancer subtypes using support vector machine (SVM).Study TypeProspective.SUBJECTSFifty‐one patients with benign (n = 23) and malignant (n = 28) breast tumors (26 ER+, whereof six were HER2+).Field Strength/SequencePatients were imaged with DW‐MRI (3T) using twice refocused spin‐echo echo‐planar imaging with echo time / repetition time (TR/TE) = 9000/86 msec, 90 × 90 matrix size, 2 × 2 mm in‐plane resolution, 2.5 mm slice thickness, and 13 b‐values.AssessmentApparent diffusion coefficient (ADC), relative enhanced diffusivity (RED), and the intravoxel incoherent motion (IVIM) parameters diffusivity (D), pseudo‐diffusivity (D*), and perfusion fraction (f) were calculated. The histogram properties (median, mean, standard deviation, skewness, kurtosis) were used as features in SVM (10‐fold cross‐validation) for differentiation of lesions and subtyping.Statistical TestsAccuracies of the SVM classifications were calculated to find the combination of features with highest prediction accuracy. Mann–Whitney tests were performed for univariate comparisons.ResultsFor benign versus malignant tumors, univariate analysis found 11 histogram properties to be significant differentiators. Using SVM, the highest accuracy (0.96) was achieved from a single feature (mean of RED), or from three feature combinations of IVIM or ADC. Combining features from all models gave perfect classification. No single feature predicted HER2 status of ER + tumors (univariate or SVM), although high accuracy (0.90) was achieved with SVM combining several features. Importantly, these features had to include higher‐order statistics (kurtosis and skewness), indicating the importance to account for heterogeneity.Data ConclusionOur findings suggest that SVM, using features from a combination of diffusion models, improves prediction accuracy for differentiation of benign versus malignant breast tumors, and may further assist in subtyping of breast cancer.Level of Evidence: 3Technical Efficacy: Stage 3J. Magn. Reson. Imaging 2018;47:1205–1216.

Từ khóa


Tài liệu tham khảo

Bihan D, 1985, Imagerie de diffusion in‐vivo par résonance magnétique nucléaire, Comptes‐Rendus Acad Sci, 93, 27

10.1002/mrm.25484

10.1002/jmri.10140

10.1002/jmri.22078

10.1002/jmri.20643

10.2214/AJR.06.1403

10.1148/radiology.168.2.3393671

10.1002/mrm.22740

10.1016/j.ejrad.2013.08.006

10.1148/radiol.2016151630

10.1002/jmri.24140

10.1007/s00330-015-3666-7

10.1007/s00330-016-4318-2

10.1038/nrclinonc.2015.73

10.1007/s00330-015-4087-3

10.1634/theoncologist.2011-S1-20

10.1172/JCI60534

10.1371/journal.pone.0087387

10.1002/jmri.25119

10.1007/BF00994018

10.1016/j.dsp.2006.10.008

10.5858/134.7.e48

10.1136/jcp.2011.089847

10.1002/mrm.25489

10.1002/mrm.10308

10.1016/j.neuroimage.2009.11.044

10.1016/S0262-8856(03)00137-9

RaoRB FungG RosalesR.On the dangers of cross‐validation. An experimental evaluation. In Proceedings of the 2008 SIAM International Conference on Data Mining.2008. Society for Industrial and Applied Mathematics.

Pedregosa F, 2011, Scikit‐learn: Machine learning in Python, J Mach Learn Res, 12, 2825

10.1002/jmri.24462

10.1002/jmri.25043

10.1002/cam4.943

10.1002/jmri.22400

10.1002/jmri.25514

10.1259/bjr/79381464

10.2463/mrms.2012-0095

10.1016/j.mri.2011.03.004

10.1002/mrm.26598

10.1002/jmri.24799

Thông tin
Thông tin xuất bản

Journal of Magnetic Resonance Imaging

Tập 47 Số 5

1205-1216

Thông tin tác giả