Development of a Radial Basis Function Neural Network Model toPredictHigh b-value Diffusion MR Signals of the Prostate

In prostate cancer detection and diagnosis, high b-value diffusion MR signals of the prostate tissues are of great concern. MR scanners and computation methods facedifficultiesin obtaining these signals. This study aims development of a neural network model to predict the MR signal amplitudes at high b-values from the amplitudes at low b-values. Synthetic diffusion MR signals are generated using a kurtosis model for noise-free and noisy conditionsconsidering nine b-values: the low b-valuesare 0, 50, 250, 500, 750s/mm2and the high b-values are 1000, 1250, 1500, 2000s/mm2. Four radial basis functionsneural networks (RBF-NN) connected in parallelare designed to accept the signal amplitudesat low b-values and to provide signal amplitudesat thehigh b-values.RBF-NNs housing altered number of neurons with radial basis functions attributing different widthsin the hidden layersof the networks are analyzed. Learning and prediction performances of the NNsare assessed from training and testing datasets. For the noise-free condition, RBF-NNsreveal perfect predictions(r= 1.000) withvery good learnings(MSE= 0.76-0.02×10-6). For the noisy conditions, the RBF-NNs achievemoderate to strong predictions(r= 0.981-0.463)withgood learnings (MSE= 0.32-10.33×10-3). Prediction performance reduces as the level of noise and/or targetedhigh b-value increases.RBF-NNsfacilitate prediction of high b-value diffusion MR signals of the prostate byrequiringno diffusion signal decay function, optimization algorithm or initial/boundary values for the optimization algorithm. They may be quite functional in accuratevoxel-wise generation of high b-value MR images for early detection and diagnosis of prostate cancer. Further prospective studies are needed tojustify the potentialbenefitsin clinicalpractice.

PDF

___

[1]L. Tangand X.J. Zhou, “Diffusion MRI of cancer: From low to high b-values,” J Magn. Reson. Imaging, vol. 49, no. 1, pp. 23-40. Jan. 2019.

[2]A.R. Padhani, G. Liu, D.M. Koh,et al.“Diffusion-weighted magnetic resonance imaging as a cancer biomarker: consensus and recommendations,”Neoplasia, vol. 11, no. 2, pp. 102-125, Feb. 2009.

[3]E.H. de Figueiredo, A.F. Borgonoviand T.M. Doring, “Basic concepts of MR imaging, diffusion MR imaging,and diffusion tensor imaging,” Magn Reson Imaging Clin N Am., vol. 19, no. 1, pp. 1-22, Feb.2011.

[4]D.M. Koh, M. Blackledge, A.R. Padhani, et al.“Whole-body diffusion-weighted MRI: tips, tricks, and pitfalls,” AJR Am J Roentgenol., vol. 199, no. 2, pp. 252-262, Aug. 2012.

[5]M.M. Berl, L. Walker, P. Modi, et al.“Investigation of vibration-induced artifact in clinical diffusion-weighted imaging of pediatric subjects,”Hum. Brain Mapp., vol. 36, no. 12, pp. 4745-57, Dec 2015.

[6]M.H. Maurerand J.T. Heverhagen, “Diffusion weighted imaging of the prostate-principles, application, and advances,”Transl. Androl. Urol., vol. 6, no. 3, pp. 490-498, Jun. 2017.

[7]Y.R. Ueno,T. Tamada, S. Takahashi, et al.“Computed diffusion-weighted imaging in prostate cancer: basics, advantages, cautions, and future prospects,” Korean J Radiol., vol. 19, no. 5, pp. 832-837, Sep. 2018.

[8]M.D. Blackledge, M.O. Leach, D.J. Collins, et al.“Computed diffusion-weighted MR imaging may improve tumor detection,”Radiology, vol. 261, no. 2, pp. 573-581, Nov. 2011.

[9]L.N. Mazzoni,S.Lucarini, S. Chiti, et al.“Diffusion‐weighted signal models in healthy and cancerous peripheral prostate tissues: Comparison of outcomes obtained at different b‐values,”J. Magn. Reson. Imaging, vol. 39, no. 3, pp.512-518, Mar. 2014.

[10]Y. Mazaheri, A.M. Hötker, A. Shukla-Dave, O. Akin,H.Hricak, “Model selection for high b-value diffusion-weighted MRI of the prostate,” Magn.Reson.Imaging, vol. 46, pp. 21-27, Feb. 2018.

[11]J.H.Jensen, J.A.Helpern, A.Ramani, et al.“Diffusional kurtosis imaging: the quantification of non-Gaussian water diffusion by means of magnetic resonance imaging,”Magn.Reson.Med., vol. 53, no. 1, pp. 1432-1440, Jun. 2005.

[12]M. Bertleff, S. Domsch, S. Weingärtner, et al.“Diffusion parameter mapping with the combined intravoxel incoherent motion and kurtosis model using artificial neural networks at 3T,”NMR Biomed, 30:e383, Sep. 2017.

[13]G. Ertas, “Estimating the distributed diffusion coefficient of breast tissue in diffusion-weighted imaging using multilayer perceptrons,” Soft Comput(2018). https://doi.org/10.1007/s00500-018-3412-6X

[14]C. Tamura, H. Shinmoto, S. Soga, et al.“Diffusion kurtosis imaging study of prostate cancer: Preliminary findings,” J. Magn. Reson. Imaging,vol. 40, no. 3, pp. 723-729, Sep. 2014.

[15]S. Suo, X.Chen, W. Lianming, et al.“Non-Gaussian water diffusion kurtosis imaging of prostate cancer,” Magn. Reson. Imaging, vol. 32, no. 5, pp. 421-427, Jun. 2014.

[16]M.C. Roethke, T.A. Kuder, T.H. Kuru, et al. “Evaluation of diffusion kurtosis imaging versus standard diffusion imaging for detection and grading of peripheral zone prostate cancer,” Invest. Radiol., vol. 50, no. 8, pp. 483-489, Aug. 2015.

[17]T. Tamada, V.Prabhu, J. Li, et al.“Prostate cancer: diffusion-weighted MRimaging for detection and assessment of aggressiveness-comparison between conventional and kurtosis models,” Radiology, vol. 284, no. 1,pp. 100-108, Jul. 2017.

[18]Z. Feng Z, X. Min X, D.J. Margolis, et al.“Evaluation of different mathematical models and different b-value ranges of diffusion-weighted imaging in peripheral zone prostate cancer detectionusing b-value up to 4500 s/mm2,” PLoS One, 15;12(2):e0172127, Feb. 2017.

[19]F. Langkilde, T.Kobus, A. Fedorov, et al.“Evaluation of fitting models for prostate tissue characterization using extended-range b-factor diffusion-weighted imaging,” Magn.Reson.Med., vol.79, no. 4, pp. 2346-2358, Apr. 2018.

[20]Y. Xi, A. Liu, F. Olumba, et al. “Low-to-high b value DWI ratio approaches in multiparametric MRI of the prostate: feasibility, optimal combination of b values, and comparison with ADC maps for the visual presentation of prostate cancer,” Quant. Imaging Med. Surg., vol. 8, no. 6, pp. 557-567,Jul.2018.

[21]Cho, G. Y. Cho, L. Moy, J.L. Zhang, et al.“Comparison of fitting methods and b-value sampling strategies for intravoxel incoherent motion in breast cancer,” Magn. Reson. Med., vol. 74, no. 4, pp. 1077-1085, Oct. 2015.

[22]I. Jambor, H. Merisaari, P. Taimen, et al.“Evaluation of different mathematical models for diffusion-weighted imaging of normal prostate and prostate cancer using high b-values: a repeatability study,” Magn. Reson. Med., vol. 73, no. 5, pp. 1988-98, May 2015.

[23]S. Chen, C.N. Cowan, P.M. Grant, “Orthogonal least squares learning algorithm forradial basis function networks,” IEEE Trans. Neural Netw., vol. 2, no. 2, pp. 302-309, Mar. 1991. [24]M.D. Pérez-Godoy, A.J. Rivera, C.J. Carmona, et al.“Training algorithms for radial basis function networks to tackle learning processes with imbalanced data-sets,” Applied Soft Computing, vol. 25, pp. 26-39, Dec. 2014.

[25]P. Zhouand J. Austin, “Learning criteria for training neural network classifiers,”Neural Comput. & Applic.,vol. 7, no. 4. pp. 334-342, 1998.

[26]J.D. Evans,Straightforward statistics for the behavioral sciences. Belmont, CA: Thomson Brooks/Cole, 1996.

[27]H. Lee,S.I. Hwang, H.J. Lee, et al. “Diagnostic performance of diffusion-weighted imaging for prostate cancer: Peripheral zone versus transition zone,”PLoS One, 13(6):e0199636,Jun 2018.

[28]J.B. Gomm and D.L. Yu, “Selecting radial basis function network centers with recursive orthogonal least squares training,” IEEE Trans. Neural Netw., vol 11, no. 2, pp. 306-314, Mar. 2000