Este artigo faz parte de uma série de artigos redigidos por colaboradores do Departamento de Ensino e Ação Social da ANEEB. Apoie o autor lendo o artigo no seu LinkedIn.
Before focusing on Deep Learning (DL) application to Biomedical Engineering, let’s become acquainted with some basic concepts.
Generally, DL is a process, like data mining, that employs deep neural network architectures, which are particular types of machine learning algorithms. These are deep artificial neural networks (ANNs) with at least one hidden layer, a set of hidden nodes, connected to input and output nodes from which connections are unknown. This unique characteristic, of understanding a system’s hidden complexity, enables DL and provides a powerful set of problem-solving tools. Notice that ANNs are like a replica of the nervous system, recently and often inspired by biology.
In addition, there are a few concepts related to or specific of DL: perceptron, which is a simple linear binary classifier that takes inputs and associated weights (representing relative input importance), and combines them to produce an output, which is then used for classification; thus, a multilayer perceptron (MLP) is the implementation of several fully adjacently-connected layers of perceptrons resulting in simple feedforward neural network (the simplest neural networks architectures in which connections are non-cyclical, unidirectional flow); contrarily, a recurrent neural network forms a cycle, bidirectional flow. In neural networks, an activation function produces the output decision boundaries by combining the network’s weighted inputs. During neural network training, the correctness of output is assessed, as it’s known the expected correct output of training data, with which output of training can be compared. Furthermore, a cost function measures the difference between real and training outputs, which should be minimized. Moreover, and normally associated with computer vision and image recognition, convolutional neural networks (CNNs) employ the mathematical concept of convolution to mimic the neural connectivity mesh of the biological visual cortex.
Currently, as you might know, DL has been applied to analysis and learning of massive amounts of unsupervised data, making it a valuable tool for Big Data Analytics (massive amounts of domain-specific information, unlabelled and un-categorized, which can contain useful information about problems).
Now, moving forward on its application to Biomedical Engineering, it is rapidly increasing and developing in a wide range of fields, such as medical image processing, object recognition in 3D scenes, deep robot learning, and data mining in bioinformatics. In most cases, DL algorithms are being used for modeling biological systems described by ever-growing datasets (increasingly complexity).
In neurosciences, these algorithms are implemented on neural decoding (to estimate subjects’ intentions based on brain activity), for example, to predict intended movements in order to move an exoskeleton with their thoughts; and to neural encoding (involves the study of signals from a neuron or a brain region to understand how they relate to external variables).
Some studies apply DL in image processing, for example, in breast cancer diagnosis and prognosis,  in which the authors applied deep learning methods to a dataset of Finite Difference Time Domain (FDTD) numerical simulations of tumor models embedded in homogeneous adipose tissue, specifically Deep and Convolutional Neural Networks, resulting in a higher performance classifier, when compared with conventional Machine Learning algorithms; in lung cancer diagnosis,  whose authors used three types of deep neural networks, designed for lung cancer calcification, which were applied to the CT image classification task with some modification for the benign and malignant lung nodules, resulting in high-performance classifiers.
Finally, and I guess you agree with me, much more can be found by and made with Deep Learning in Biomedical Engineering. Recent research applies deep learning classifiers to diagnose COVID-19 in X-Ray Images , thus there’s always space for an objective and effective classification in health departments.
Go deep into data and find out its hidden features!
Mayo, M., 2020. Deep Learning Key Terms, Explained – Kdnuggets. [online] KDnuggets. Available at: <https://www.kdnuggets.com/2016/10/deep-learning-key-terms-explained.html> [Accessed 12 May 2020].
Najafabadi, M.M., Villanustre, F., Khoshgoftaar, T.M. et al. Deep learning applications and challenges in big data analytics. Journal of Big Data 2, 1 (2015). Available at: <https://doi.org/10.1186/s40537-014-0007-7> [Accessed 13 May 2020].
Koprowski, R., 2020. Deep Learning In Biomedical Engineering. [online] Biomedcentral.com. Available at: <https://www.biomedcentral.com/collections/Deep> [Accessed 13 May 2020].
KORDING, K., BENJAMIN, A., FARHOODI, R. and GLASER, J., 2020. The Roles Of Machine Learning In Biomedical Science. [online] Ncbi.nlm.nih.gov. Available at: <https://www.ncbi.nlm.nih.gov/books/NBK481619/> [Accessed 13 May 2020].
1: Gerazov, B. and Conceição, R., 2020. Deep Learning For Tumour Classification In Homogeneous Breast Tissue In Medical Microwave Imaging – IEEE Conference Publication. [online] Ieeexplore.ieee.org. Available at: <https://ieeexplore.ieee.org/document/8011175> [Accessed 13 May 2020].
2: Gao, Junfeng. Song, QingZeng. Zhao, Lei. Luo, XingKe. Dou, XueChen. 2017. Using Deep Learning for Classification of Lung Nodules on Computed Tomography Images. Journal of Healthcare Engineering. Hindawi. Available at: <https://doi.org/10.1155/2017/8314740> [Accessed 13 May 2020].
3: Hemdan, E., Shouman, M. and Karar, M., 2020. COVIDX-Net: A Framework Of Deep Learning Classifiers To Diagnose COVID-19 In X-Ray Images. [online] arXiv.org. Available at: <https://arxiv.org/abs/2003.11055> [Accessed 13 May 2020].