Artificial Intelligence Aiding Medical Diagnosis Focusing on Diabetic Retinopathy

Introduction

Artificial intelligence (AI) is a collection of technologies that work in concert to enable the machines to detect, interpret, work, and learn at the levels of intelligence of human brain (Accenture, 2021). The artificial intelligence categorized as Weak AI, also termed as Artificial Narrow Intelligence (ANI), or simply as Narrow AI, is a form of artificial intelligence that has been trained and fixated on performing narrow tasks. The majority of the AI that surrounds us now is driven by weak AI. IBM Watson, autonomous vehicles and Apple’s Siri are all examples of this type of AI. Strong AI, also known as Artificial Super Intellect, is a speculative kind of AI in which a machine has an intelligence equivalent to humans, something that surpasses the intelligence and capacity of the human brain. Strong AI is purely a theoretical concept as yet and there are no real instances in use presently (Education, 2021).

Complex computing's capacity to do pattern recognition by building complex associations on the basis of incoming data, followed by its comparison with performance standards is a significant step forward towards dealing with such vast amount of data. This technology has the potentiality to ameliorate healthcare pricing, efficiency, and availability. It is essentially a digital system's capacity to demonstrate cognitive abilities. AI is much more than a massive collection of database. Just as people need to learn to execute things, AI systems need to be first exposed to a database that allows them to initially “learn” elementary targets related to a certain discovery or condition. Following the early learning stages, the machine or the system is taught to “develop,” i.e., to modify its original learning to be much more efficient and precise. This learning is augmented by the application of sophisticated differential algorithms to help the system grasp complex correlations between diverse variables via an information dissemination called “neural network.” (Doi, 2007; Hamet, 2017; Jiang, 2017).

Machine learning (ML) is basically, a subset of Artificial intelligence (AI) and deep learning (DL) is further, a subset of machine learning (Darcy et al., 2016). Deep learning has been widely used in predictive analysis. It works in a similar way as neurons work in our brain. Its structure and networking is also rooted on the neural networks in our brain. It is composed of the artificial neural network which when introduced with a vast amount of learning data understands it and makes a memory of it and when the test data similar to it is inputted, it recognizes and produces the output. It works similar to our brain that learns when introduced to new information, by first trying to compare with known information and recognizing it with the stored memory. Structurally, it is made up of many layers of artificial neurons, thus giving it the name ‘deep learning’. The basic structure comprises of deep neural network, constituted with an input layer of artificial neurons, few hidden layers and finally, the output layer, all arranged hierarchically. All these layers consist of the structural unit called ‘node’, analogous to a biological neuron. Function conducted by neurons of each layer is termed as activation function which is then forwarded to the connected neuron. The input is received and worked upon only when it is above threshold just like how our biological neurons work (Jigsaw Academy, 2021)

Artificial intelligence has a widespread utilization in the health and medical sciences. AI has been in use in medicine since the early 1950s, when the physicians sought to enhance the accuracy of their diagnostics by employing computer-aided algorithms. (Secinaro et al., 2021).These technologies are used to discover novel medications for health-care management and patient-care therapies (Burton, 2019; Howard, 2019; Murff et al., 2011; Yang, 2019).