The Concept Of Artificial Intelligence In Medical Imaging: Overview Of Its Origin

Have you heard of artificial intelligence in medical imaging? What is it about? This article will give you the history, concept and types of AI in medical imaging and its implication in the medical imaging diagnosis.

Artificial intelligence (AI) is a game-changing technology that use computerized algorithms to decipher complex data. Diagnostic imaging is one of the most potential clinical applications of AI, and increasing effort is being paid to establishing and fine-tuning its performance to help detect and quantify a wide range of clinical problems. Computer-aided diagnostics investigations have shown great accuracy, sensitivity, and specificity for the detection of minor radiographic abnormalities, potentially improving public health.

History of AI

A group of scientists from many fields (mathematics, psychology, engineering, economics, and political science) began debating the prospect of developing an artificial brain. During the summer of 1956, they gathered on the Dartmouth College campus for a workshop. Dartmouth Workshop, as it is known, is a society dedicated to artificial intelligence (AI). 1 The field then went through multiple cycles of peaks and dips. Marvin Minsky, an MIT cognitive scientist, and other Dartmouth Workshop attendees were tremendously hopeful about AI's future. They expected AI would be largely solved within a decade.

There was, however, no meaningful development. Following multiple critical studies and continued congressional pressure, government funding and interests began to dwindle. The first AI winter was from 1974 until 1990. AI resurrected in the 1980s as a result of British and Japanese competition. The winter of 1983–93 was a watershed moment for AI, as the market for essential computer power collapsed, forcing funding to be withdrawn once more. Following then, research began to ramp up again. IBM's Deep Blue, the first computer to defeat a chess champion, is a well-known example. Watson, IBM's question-answering system, won the quiz program Jeopardy in 2011, ushering in a new era of AI development. In addition, The amount of imaging data has expanded tremendously in recent years in medical imaging studies. The pressure on doctors to process the photos has risen as a result. They must be able to read images more quickly while maintaining the same level of precision. Fortunately, processing power has increased dramatically at the same period. These obstacles and opportunities have created the ideal environment for artificial intelligence to flourish in medical imaging research.

The application of artificial intelligence (AI) in diagnostic medical imaging is currently being researched. AI has demonstrated outstanding sensitivity and accuracy in the detection of imaging abnormalities, and it has the potential to improve tissue-based detection and characterization. However, when sensitivity improves, a significant disadvantage emerges: the detection of tiny changes of unclear significance. In radiology, researchers have effectively used AI to discover findings that are detectable or not by the naked eye. Radiology is transitioning from a subjective perceptual talent to a science that is more objective. AI has been effectively utilized to automatic tumor and organ segmentation, as well as tumor monitoring during treatment for adaptive treatment in Radiation Oncology. Lambin P, a Dutch researcher, initially suggested the notion of "Radiomics" in 2012, defining it as follows: a high-throughput method for extracting a large number of picture attributes from radiation photographs.

TYPES OF AI IN MEDICAL IMAGING

Radiomics (in the context of radiology) is a branch of medicine that tries to extract a large number of quantitative aspects from medical pictures utilizing data characterization techniques. The information is analyzed to help in decision-making. It has the ability to reveal illness traits that are difficult to detect with just human vision.

Radiomics can be used to characterize tumor aggressiveness, viability, response to chemotherapy

and/or radiation. Therefore, a radiomic approach can help to reveal unique information about

tumor biological behavior. It can be used for prognosis estimation in confirmed lung cancers or

to estimate the risk of distant metastasis. Radiomics has also been used to predict histology and

mutational profile of lung tumors.

CONVOLUTIONAL NEURAL NETWORK (CNN)

A Convolutional Neural Network (ConvNet/CNN) is a Deep Learning system that can take an input image, assign relevance (learnable weights and biases) to various aspects/objects in the image, and distinguish between them. When compared to other classification methods, the amount of pre-processing required by a ConvNet is significantly less. While basic approaches require hand engineering of filters, ConvNets can learn these filters/characteristics with enough training. The goal of this field is to enable machines to see and perceive the world in the same way that humans do, and to apply that knowledge to a variety of tasks such as image and video recognition, image analysis and classification, media recreation, recommendation systems, natural language processing, and so on. Advancements in Computer Vision using Deep Learning have been built and developed over time, mostly with one algorithm: the Convolutional Neural Network. The arrangement of the Visual Cortex influenced the architecture of a ConvNet, which is similar to that of the connectivity network of Neurons in the Human Brain. Individual neurons only respond to stimuli in the Receptive Field, which is a small area of the visual field. A group of similar fields can be stacked on top of one another to fill the whole visual field.

COMPUTER AIDED DIAGNOSIS (CAD)

The use of a computer generated output as a helping tool for a physician making a diagnosis is known as computer aided diagnosis (CAD). It's not the same as automated computer diagnosis, where the final diagnosis is solely based on a computer algorithm. Computer aided diagnosis methods have been utilized widely in radiology for many years as an early type of artificial intelligence . The most prevalent uses are for mammography breast cancer detection and chest CT pulmonary nodule detection. Traditional feature engineering based on domain expertise was used in these systems, while recent approaches use machine learning to uncover latent characteristics in imaging data. CADe (computer-assisted detection): labels specific parts of images that appear aberrant, reducing the possibility of missing disorders of interest.

A practitioner can use computer-aided diagnosis (CADx) to examine and classify pathology in medical images.

ARTIFICIAL NEURAL NETWORKS (ANN)

Artificial neural networks (ANNs) are a type of technology derived from brain and nervous system research. These networks are modeled after biological neural networks, although they only utilize a subset of biological neural system ideas. ANN models, for example, replicate brain and nervous system electrical activity. The processing elements (also known as a neurode or a perceptron) are linked to one other. Artificial neural networks are a versatile sort of model that can handle a wide range of inputs. Originally inspired by the connections between biological brain networks, modern artificial neural networks only have a passing resemblance to their biological counterparts at a high level. The analogy is still conceptually useful, as evidenced by some of the vocabulary employed. Numerous more neurons in the more superficial levels provide variable-weighted input to individual "neurons" in the network. The cumulative input of these more superficial neurons is needed for any one neuron to activate. They link to a large number of deeper neurons, all of which have different weightings.

IMPLICATIONS OF AI IN MEDICAL IMAGING

As obtained images are incorporated in patients' clinical histories for diagnosis, treatment planning, screening, follow-up, or prognosis, radiology generates a large amount of digital data. Furthermore, as the usage of computers and data has grown, artificial intelligence (AI) has been successfully applied to a variety of jobs, resulting in more accurate and up-to-date findings. The application of artificial intelligence (AI) in diagnostic medical imaging is currently being researched. AI has demonstrated outstanding sensitivity and accuracy in the detection of imaging abnormalities, and it has the potential to improve tissue-based detection and characterization. However, when sensitivity improves, a significant disadvantage appears, notably the detection of tiny changes of unclear significance. The future of AI will be magnificent in diagnosis, treatment planning and outcome.

References

1. https://www.thelancet.com/journals/landig/article/PIIS2589-7500(20)30160-6/fulltext#seccestitle10
2. https://radiopaedia.org/articles/neural-network-overview-1
3. https://towardsdatascience.com/a-comprehensive-guide-to-convolutional-neural-networks-the-eli5-way-3bd2b1164a53
4. https://www.mja.com.au/journal/2014/201/1/impact-genomics-future-medicine-and-healt

EMMANUEL AMPOFO

LEVEL 400 STUDENT AT UCC READING RADIOGRAPHY

TELEPHONE:0246578851

EMAIL:[email protected]