A.P. Kazantsev
Ph.D. (Eng.), Associate Professor, Pushchino State Institute of Natural Sciences;
Senior Research Scientist, Acting Head of the laboratory of Biomedical Research Institute of Biological Instrumentation of the Russian Academy of Sciences,
E-mail: telemed.ak@gmail.com
L.M. Subbotina
Bachelor of Mathematics, Engineer, Institute of Biological Instrumentation of the Russian Academy of Sciences,
Post-graduate Student, Pushchino State Institute of Natural Sciences
E-mail: lilysubby@gmail.com
A.A. Senin
Ph.D. (Eng.), Research Scientist, Pushchino State Institute of Natural Sciences;
Institute of Biological Instrumentation of Russian Academy of Sciences
E-mail: digger_1@mail.ru
N.S. Minaev
Bachelor of Economics, Engineer, Institute of Biological Instrumentation of the Russian Academy of Sciences,
Post-graduate Student, Pushchino State Institute of Natural Sciences
E-mail: st.denko@rambler.ru
Ju.N. Ponomareva
Dr.Sc. (Med.), Professor, Department of Obstetrics and Gynecology,
Moscow State University of Medicine and Dentistry
E-mail: juliyapon@mail.ru
E.M. Chatskis
Head of the Department of Ultrasound Diagnostics,
Road Clinical Hospital at the Chita-2 Station of Russian Railways
E-mail: len130922@yandex.ru
E.M. Proshin
Dr.Sci (Eng.), Professor, Department of Information-measuring and Biomedical Engineering,
Ryazan State Radio Engineering University,
Honorary Worker of Higher Professional Education of the Russian Federation,
Laureate of the Ryazan region on science and technology academician V.F. Utkin
E-mail: proshin39@mail.ru

Introduction. The aim was design of a neural network for the stereometric determination of the location of a wide-aperture ultrasound probe against the fetal heart, according to the information of the sonographic probe audio signal with simultaneous identification of the profile of intracardiac fetal blood flow for subsequent detection of its cardiac cycle phases.

Materials and methods. The ultrasound location of the intracardiac fetal blood flow initially assumes the determination of the distance from the probe input surface to the conditional center of a fetal heart, which was taken as the origin of the three-dimensional polar coordinate system of the imaginary sphere and heart, and actually the fetus. The purpose of the location is the blood flow profile, as the carrier of the desired hemodynamic information. Due to the dependence of the Doppler effect on the angle of insonation, the localized profile depends on the angle of the probe. An arbitrary sonographic signal of intracardiac blood flow was taken, recorded for such an investigation with the help of a fetal L6 Smart Doppler from a certain angle. With the help of the Gaussian adaptive chirplet transform, the original signal was reduced to the form of continuous bipolar blood flow profiles. The section of the bipolar profile was cut out and was brought to the form of a pattern of input to the earlier trained committee convolutional neural network. To train each elementary network, the simulator of the intracardiac blood flow was used, which, depending on the stereometric position of the sensor relative to the fetal heart, yields a synthesized total signal from the fetal heart blood vessels. To recognize 20 different profections of the volume blood flow profiles in the vicinity of the fetal heart that uniquely correspond to the separate sectors (60×60°) of the axis of the Doppler proble in the polar fetal cardiac coordinate system, a deep five-layer convolutional neural network with two fully connected layers of the classifier was developed. To detect small angles of 6x6 (facets), a facet method for constructing and operating a cooperative neural network in the form of a lattice of 10x10 convolutional neural networks of an identical architecture was developed. The training was conducted on the caffe platform with the python interface using the method of back propagation of the error.

Results. The use of a cooperative neural network for locating complex impulse blood flow inside the fetal heart is proposed using a special Doppler ultrasound probe (fetal doppler - FD) with a wide (~60°) directional pattern or angular aperture.

The format of the graphic pattern of intracardiac fetal blood flow perceived as an echo in the aperture of the ultrasonic Doppler fetal monitor with a reference frequency of 2 MHz is developed, and after transformation of the echo into an audio signal and final demodulation of a 200x200 rendered on a pixel matrix in the form of four processes occurring during one cardiac cycle and representing four graphic elements of the pattern: positive and negative profiles of blood flow pulses; positive and negative 10-percent interference levels in pauses between pulses.

Conclusion. Demonstrated the feasibility of automation of Doppler with the help of computational intelligence, which is of fundamental importance for the telemonitoring of the fetus with the second trimester of pregnancy.

1. Ageeva M.I. Doplerograficheskoe issledovanie vnutriserdechnoj gemodinamiki ploda pri fiziologicheskom ego razvitii vo II–III trimestrah beremennosti // Ul'trazvukovaya i funkcional'naya diagnostika. 2005. № 3. S. 11–20.
2. Medvedeva D.A., Kazancev A.P., Ponomaryova YU.N., CHackis E.M., Senin A.A., Minaev N.S., Abramov A.M. Imitacionnaya model' doplerometrii vnutriserdechnogo krovotoka ploda // Biomedicinskaya radioehlektronika». 2016. № 6. S. 38–44.
3. Medvedeva D.A., Kazancev A.P., Ponomaryova YU.N., CHackis E.M., Senin A.A., Minaev N.S., Subbotina L.M., Kryakov V.G. Podhod k telemedicinskoj tekhnologii monitorirovaniya vnutriserdechnoj gemodinamiki ploda // Biomedicinskaya radioehlektronika. 2017. № 7. S. 85–94.
4. Kazantsev A., Ponomareva J., Kazantsev P. Developmentand validation of an AI-enabled mHealth technology for in-home pregnancy management // Proc. of the 2014 International Conference on Information Science, Electronics and Electrical Engineering. Apr. 26–28 2014. SapporoCity, Hokkaido, Japan. Beijing: IEEE. 2014. P. 927–931.
5. Kazantsev A.P.; Senin A.A.; Ponomareva J.N.; Mochalova M.N. An mHealth approach to remote fetal monitoring // Proc. of the Int. Healthcare Innovation Conference (HIC). 810 Oct. 2014. Seattle, WA, USA. IEEE. 2014. P. 239–242.
6. Kazancev A.P., Senin A.A., Ponomaryova YU.N., Mochalova M.N., Proshin E.M. Podhod k sozdaniyu massovoj telemedicinskoj tekhnologii domashnego monitoringa ploda // Biomedicinskaya radioehlektronika. 2015. № 5. S. 37–45.
7. Kazantsev A., Ponomareva J., Kazantsev P., Digilov R., Ping Huang. Development of e-health network for in-home pregnancy surveillance based on artificial intelligence // Proc. 2012 IEEE-EMBS Int. Conf. BHI. Hong Kong. 2012. R. 82–84.
8. Senin A.A., Kazancev A.P. Cifrovaya zapis' i indikaciya audiosignala ul'trazvukovogo doplerovskogo zonda s peredachej dannyh na udalennyj server pri monitoringe ploda posredstvom mobil'nogo ustrojstva na platforme Android // Svidetel'stvo o gosudarstvennoj registracii programmy dlya EHVM
   № 2014612803.
9. Mann S. and Haykin S. The chirplet transform: Physical considerations // IEEE Trans. Signal Processing. 1995. № 43. R. 2745–2461.
10. Rybakova M.K., Mit'kov V.V. Differencial'naya diagnostika v ehkhokardiografii (+ DVD). M.: VIDAR. 2011. 232 s.
11. Rybakova M.K., Mit'kov V.V. EHkhokardiografiya v tablicah i skhemah. Nastol'nyj spravochnik. M.: VIDAR. 2011. 288 s.
12. Bunin A., Strizhakov A., Medvedev M., Ageeva M. “Diagnostic Importance of Dopplerometry in Cases of Fetal Development Retardation // Obstetrics & Gynecology. 1999. V. 12. R. 41.
13. Nikolenko S., Kadurin A., Arhangel'skaya E. Glubokoe obuchenie. SPb: Piter. 2018. 480 s.
14. Haykin S. Neural Networks: A Comprehensive Foundation. 2-nd Edition. Prentice-Hallyu 1999. 842 p.
15. Krizhevsky A., Sutskever I., Hinton G.E. ImageNet Classification with Deep Convolutional Neural Networks // Advances in Neural Information Processing Systems 25. Curran Associates, Inc. 2012. P. 1097–1105.
16. Goodfellow I., Bengio Y., Couruille A. Deep Learning. MIT Press. 2016. http://www.deeplearningbook.org. (data obrashcheniya: 01.05.2018).
17. Goodfellow I. NIPS 2016 Tutorial: Generative Adversarial Networks // arXiv. 2017. http://arxiv.org/abs/1701.00160. (data obrashcheniya: 01.05.2018).
18. Russakovsky O. et al. ImageNet Large Scale Visual Recognition Challenge // International Journal of Computer Vision (IJCV). 2015. V. 115. № 3.
    P. 211–252.
19. Li Deng, Dong Yu. Deep Learning: Methods and Applications // Foundations and Trends in Signal Processing. 2014. V. 7. № 3–4. P. 197–387.