DOI: https://doi.org/10.31071/kit2020.16.01

Inventory reference

ISSN 1812-7231 Klin.inform.telemed. Volume 15, Issue 16, 2020, Pages 28-34

Author(s)

Y. E. Liakh¹, V. G. Hurianov², V. O. Biloshenko³, M. V. Liakh¹, V. O. Melnichuk⁴

Institution(s)

¹National University Ostroh Akademy, Ostroh, Ukraine

²Bogomolets National Medical University Kyiv, Ukraine

³Donetsk institute for Physics and Engineering named after O.O. Galkin, Kyiv, Ukraine

⁴Lesia Ukrainka East European National University Lutsk, Ukraine

Article title

Neural network analysis of mammary gland thermograms using the estimate of fractal dimension in field temperature distribution

Abstract (resume)

Introduction. Thermography is one of the promising additional standard methods of mammary glands screening in a large group of population. This method is considered to be suitable for widespread use due to its non-invasiveness, lack of radiation exposure and thus safety for the health of patients, accessibility to patients and high detection effectiveness of pathological changes of the mammary gland.

Methods of thermograms evaluation and analysis. To identify the risk of mammary gland pathology we analyzed thermograms using 68 features, among which three indicators of general characteristics: age of the patient, minimal temperature of theMG field, size of the MG temperature field; 32 features of the relative area of temperature rise; and 33 features of thermograms characteristics according to Hurst exponent of high dimensional fractals. To analyze distribution of MG field temperature and to identify signs of thermograms associated with the risk of pathology, methods of constructing one-factor and multifactor regression models were used, as well as method of operating characteristic curves (ROC).

Quantitative analysis of the thermography results. On the basis of the selected factor signs, a linear model for predicting the risk of MG pathology was built — AUC = 0,85 (95% CI 0,82–0,87) and a nonlinear model (was used a multilayer perceptron — MLP, with one hidden layer with sigmoid activation functions) for predicting the risk of MG pathology AUC = 0,89 (95% CI 0,87–0,92). A non-linear neural network model on a reduced set of traits had better (p<0,05) prognostic characteristics (AUC) than a linear model on all 68 features or a linear model on significant factor features. The prognostic characteristics of the MLP model allow to use it in order to predict the risk of a pathological process.

Conclusions. To analyze mammary gland thermograms with assessment of the fractal dimension of the field temperature distribution in norm and in pathology was constructed a neural network MLP model for predicting the risk of MG pathology. Sensitivity of the proposed model is 90,2% (95% CI 86,7%–93,0%), specificity — 85,1% (95% CI 80,6%–88,9).

Keywords

Thermograms, Mammary gland, Fractal dimension, MG temperature, One-wayanalysis of variance, MLP model

References

1. Ng E. Y. A review of thermography as promising non-invasive detection modality for breast tumor. Int. J. Therm. Sci., 2009, vol. 48, no 5, pp. 849–859.
https://doi.org/10.1016/j.ijthermalsci.2008.06.015

2. Kovalchuk Y. S., Dunaevskyi V. Y., Venher E. F., Kotovskyi V. Y., Nazarchuk S. S. The possibility of remote-sensing infrared thermography in the diagnosis of breast diseases (benign changes). Ukrainskyi medychnyi chasopys [Ukrainian medical journal], 2013, vol. 95, no. 3, pp. 165–169. (In Russ.).

3. Prykhodchenko V. V., Prykhodchenko O. V., Beloshenko V. A., Doroshev V. D., Karnachev A. S. Improving the efficiency of the screening stage for selective screening of breast diseases. Medyko-sotsialni problemy sim'i [Medical and social problems of family], 2009, vol. 14, no. 4, pp. 20–25. (In Russ.).

4. Rozenfeld L. H., Kolotylov N. N.The remote infra-red thermography in the oncology. Onkolohyia [Oncology], 2001, vol. 3, no. 2–3, pp. 103–106. (In Russ.).

5. Markel A. L., Vainer B. H. Infra-red thermography in the diagnosis of breast cancer (review of foreign literature). Terapevtycheskyi arkhyv [Therapeutic archive], 2005, vol. 776, no.10, pp. 57–61.(In Russ.).

6. Prykhodchenko V. V., Prykhodchenko O. V. Diagnosis of breast diseases using a digital contact thermograph. Medyko-sotsialni problemy sim'i [Medical and social problems of family], 2005, vol.10, no. 3–4, pp. 61–64. (In Russ.).

7. Prykhodchenko V. V., Prykhodchenko O. V., Biloshenko V. O., Karnachov O. S., Doroshev V. D., Kalinina O. I. Potentialities of contact digital thermography in early preclinical diagnostics of breast tumor. Onkolohyia [Oncology], 2009, vol. 13, no. 2, pp. 125–129. (In Ukr.).

8. Beloshenko V. A., Doroshev V. D., Karnachev A. S., Prykhodchenko V. V. A system for early detection of cancer diseases by means of contact digital thermography. Nauka ta innovatsii [Science and innovation], 2007, vol. 3, no. 5, pp. 11–25. (In Russ.).
https://doi.org/10.15407/scin3.05.011

9. Prykhodchenko V. V., Dumanskyi Yu. V., Prykhodchenko O. V., Beloshenko V. A., Doroshev V. D., Karnachev A. S. Employment of the contact digital thermograph ТKTS-1 in the diagnosis of breast diseases. Donetsk, 2012, 189 p. (in Russ.).

10. Carbone A. Algorithm to estimate the Hurst exponent to fhighdimensional fractals. PhysRev E, 2007; no. 76:056703.
https://doi.org/10.1103/PhysRevE.76.056703

11. Dumansky Y. V., Lyakh Y. E., Gorshkov O. G., Gurianov V. G., Prihodchenko V. V. Fractal dimensionality analysis of normal and cancerous mammary gland thermograms. Chaos, Solitons and Fractals, 2012, vol. 45, pp. 1494–1500.
https://doi.org/10.1016/j.chaos.2012.07.006

12. Etehad Tavakol Mahnaz, Lucas Caro, Sadri Saeed, Ng Eyk. Analysis of breast thermography using fractal dimension to establish possible difference between malignant and benign patterns. J Healthcare Eng., 2010, vol. 1, no. 43.
https://doi.org/10.1260/2040-2295.1.1.27

13. Baish J. W., Jain R. K. Fractals and cancer, American Association for Cancer Research. Cancer Res., 2000, vol. 60, pp. 3683–3688.

14. Ramirez-Cobo P., Vidakovic B. A 2D wavelet-based multiscale approach with applications to the analysis of digital mammograms. Computational Statistics and Data Analysis, 2013, vol. 58, no. 2, pp. 71–81.
https://doi.org/10.1016/j.csda.2011.09.009

15. Hurianov V. H., Liakh Yu. Ye., Parii V. D., et. al. Posibnyk z biostatystyky. Analiz rezul'tativ medychnykh doslidzhen' u paketi EZR (R–statistics). [Handbook of Biostatistics. Analysis of the results of medical research in the EZR package (R–statistics)], Kyiv, Vistka, 2018, 208 p. (In Ukr.).

16. Kanda Y. Investigation of the freely available easy-to-use software "EZR" for medical statistics. Bone Marrow Transplant., 2013, no. 48, pp. 452–458.
https://doi.org/10.1038/bmt.2012.244

17. Liakh Yu. E., Hurianov V. H. Mathematic modeling for classification problems in biomedicine. Ukrainskyi zhurnal telemedytsyny ta medychnoi telematyky [Ukrainian Journal of telemedicine and medical telematics], 2012, vol. 10, no. 2, pp.69–76. (In Russ.).

18. Liakh Yu. E, Hurianov V. H., Yakobson E. A. Fractal dimencionality analysis of mammary gland thermograms. Medychna informatyka ta inzheneriia [Medical informatics and engineering], 2016, no. 1, pp. 88–90. (In Russ.).

19. Liakh Yu. Ye., Biloshenko V. O., Hurianov V. H., Prykhodchenko V. V. Automated system for evaluation of the mammary glands examination results for contact digital thermography. Medychna informatyka ta inzheneriia [Medical informatics and engineering], 2019, no. 2, pp. 25–37. (In Ukr.).
https://doi.org/10.11603/mie.1996-1960.2019.2.10315

20. Liakh Yu. Ye., Dumanskyi Yu. V., Hurianov V. H., Prykhodchenko V. V. Development of automatized expert system for the preliminary screening estimation of data of contact digital thermography. Unyversytetskaia klynyka [University clinic], 2011, vol. 7, no. 1, pp.109–112. (In Russ.).

Full-text version http://kit-journal.com.ua/en/viewer_en.html?doc/2020_16/001.pdf