2D signal classification system for decision support

The paper presents the results of testing the decision support system based on AI. It is used to detect geometrically described irregularities on 2Dsignals. There are two favorite cases for the system. First is the pollution detection on ocean surface. The second one is estimation of the brain damage volume in multiple sclerosis. 300 mri studies were classified for the test. All of them were depersonalized and tagged. Parameters under research were: sensitivity, specificity, average processing time of studies, detection probability, area under the ROC curve in two groups, percentage of false positive, false negative lesions, percentage of the volume of lesions in the pathology group. It was found that the average accuracy of the tested system is 0,89, the average sensitivity is 0,89, and the average specificity is 0,88. Classification system based innovative product has excellent results not only in accuracy, but also in speed and reliability, since testing takes only 60 seconds on average for automatic segmentation of multiple sclerosis lesions.

Keywords: classification of 2D signals, multiple sclerosis, MRI diagnostics, artificial intelligence, quantitative measurements.

References

  1. B. V. Strelkov, M. L. Belov, S. A. Tukhvatullina, V. A. Gorodnichev. Lazernyy metod obnaruzheniya neftyanykh zagryazneniy na vzvolnovannoy morskoy poverkhnosti, ispol'zuyushchiy uglovoe skanirovanie//Nauka i obrazovanie. MGTU im. N. E. Baumana. № 07. Iyul' 2012.
  2. M. A. Kudrina. Ispol'zovanie preobrazovaniya Khafa dlya obnaruzheniya pryamykh liniy i okruzhnostey na izobrazhenii//Izvestiya Samarskogo nauchnogo tsentra Rossiyskoy akademii nauk. T. 16. №4 (2). 2014.
  3. Neurohive. https://neurohive.io/ru/vidy-nejrosetej/u-net-image-segmentation.
  4. J. Le. How to do Semantic Segmentation using Deep learning. 2018. https://nanonets.com/blog/how-to-do-semantic-segmentation-using-deep-learning.
  5. M. D. Zeiler, R. Fergus. Visualizing and understanding convolutional networks. In European conference on computer vision//Springer, Cham. September, 2014. P. 818-833. https://arxiv.org/abs/1311.2901.
  6. I. Shirokova, I. Sidorova. Diagnoz: rasseyannyy skleroz//Remedium. 2015. № 1-2. https://cyberleninka.ru/article/n/diagnoz-rasseyannyy-skleroz.
  7. L. I. Sokolova, T. A. Kobys', N. V. Domres. Sovremennye kriterii Makdonal'da v diagnostike rasseyannogo skleroza//Mezhdunar. nevrol. zhurn. MNZh. 2012. № 7 (53). https://cyberleninka.ru/article/n/sovremennye-kriterii-makdonalda-v-diagnostike-rasseyannogo-skleroza.
  8. V. Martinelli, M. Rodegher, L. Moiola, G. Comi. Late onset multiple sclerosis: clinical characteristics, prognostic factors and differential diagnosis//Neurological Sciences, 25(4), 2004. P. 350-355.
  9. C. Walton, R. King, L. Rechtman et al. Rising prevalence of multiple sclerosis worldwide: Insights from the Atlas of MS, third edition//Mult Scler. 2020. Dec 26(14): 1816-1821. doi: 10.1177/1352458520970841.
  10. E. E. Altay, E. Fisher, S. E. Jones et al. Reliability of Classifying Multiple Sclerosis Disease Activity Using Magnetic Resonance Imaging in a Multiple Sclerosis Clinic//JAMA Neurol. 2013, 70 (3): 338-344. doi: 0.1001/2013.jamaneurol.211.

Authors