Slima abstract XAI Deep learning for health using fuzzy logic
1. Title: Explainable neuro-fuzzy recurrent neural network to predict
colorectal cancer with different timeframe data
Abstract
Deep learning algorithms have proven to be an effective mechanism for making highly accurate predictions from
vast and complex data sources [1]. In particular, long short term memory (LSTM) deep neural networks are used
for analyzing data with different temporal granularity [2]. In the area of health, this feature is particularly useful
in analyzing trends of diseases in several time spans:decades,years,months, hours or seconds.Previous research
has shown that the identification of temporal patterns from electronic medical records (EMR) data can be highly
beneficial in the prediction of colorectal cancer (CRC) [3]. Nevertheless, and despite LSTM could predict with
high level of certainty the occurrence of colorectal cancer, it does not offer an explanation for how the output is
generated. Without understanding how deep learning algorithms arrive to a solution, there is no guarantee that
these neural networks will transition from controlled laboratory environments to production environments. An
explainable AI (XAI) or transparent AI allows a better understanding ofthe actions that are performed within the
“black box” (machine learning) to be trusted and easily understood by humans [4, 5]. This abstract introduces an
approach for adjusting the parameters of a fuzzy systemusing a LSTM neural network in order to achieve high
precision but without losing interpretability. The fuzzy systemwill be built using the standard IEEE 1855-2016
fuzzy markup language (FML)[6] and the free software tool GUAJE (Generating Understandable and Accurate
fuzzy models in a Java Environment) [7]. GUAJE was created for building fuzzy interpretable systems that are
fed by expert knowledge or machine learning techniques. There has been no previous efforts to integrate these
fuzzy interpretable systems with neural networks and apply this combination to CRC prediction. The final goal is
to propose a conceptualframework for predicting colorectal cancer that is both accurate and explainable and can
be used in several hospitalareas such as oncology,radiology and disease prevention to improve medical decision-
making. Developing the proposed systemwill allow effective communication between autonomous systems and
analysts. This work will improve the understanding of the system’s strengths and limitations, quantify the
uncertainty, and develop methods for neuro-fuzzy integration in the medical field.
References
[1] Deng, L., & Yu, D. Deep learning: methods and applications. Foundations and Trends® in Signal
Processing,7(3–4), 197-387. (2014).
[2] Hochreiter, S., and Schmidhuber, J. Long short-term memory. Neural computation 9(8):17351780. (1997).
[3] Dmitrii Bychkov et al. Deep learning based tissue analysis predicts outcome in colorectal cancer. Nature
magazine, Scientific reports (2018) 8:3395 | DOI:10.1038/s41598-018-21758-3
[4] Core, M. G., Lane, H. C., Van Lent, M., Gomboc, D., Solomon, S., & Rosenberg, M. Building explainable
artificial intelligence systems.In AAAI 2006 (pp. 1766-1773).
[5] Samek, W., Wiegand, T., & Müller, K. R. Explainable artificial intelligence: Understanding,visualizing and
interpreting deep learning models. arXiv preprint arXiv:1708.08296. (2017).
[6] J.M. Soto-Hidalgo, Jose M. Alonso,G. Acampora, and J. Alcala-Fdez, "JFML: A Java Library to Design
Fuzzy Logic Systems According to the IEEE Std 1855-2016", IEEE Access (ISSN:1556-603X),
DOI:10.1109/ACCESS.2018.2872777, https://dx.doi.org/10.1109/ACCESS.2018.2872777
[7] J.M. Alonso and L. Magdalena, “Generating understandable and accurate fuzzy rule-based systems in a Java
environ-ment,”Lecture Notes in Artificial Intelligence - Proc. 9th InternationalWorkshop on Fuzzy Logic and
Applications,LNAI6857, 212–219 (2011)