Computers having exceeded human abilities in today’s time is no longer something unusual. Prof Hiroshi Fujita further highlighted the accuracy of computers in image recognition, especially for the use of detecting breast cancer with mammograms, in his webinar titled “Fundamentals of AI-deep-learning Applications to Medical Image Diagnosis”. The webinar was organised by Centre for Internet of Things and Big Data (CIoTBD) on 4 February 2021 via Microsoft Teams.

The webinar, which aimed to enlighten participants on the applications and benefits of AI deep learning, focused on Artificial Intelligence Computer-aided Detection/Diagnosis (AI-CAD) used in image diagnosis by the medical practitioners, including for the recent Covid-19 pandemic.

“CAD first started in 1960, and the first CAD system was developed at the University of Chicago in 1994 for the detection of breast lesions on mammograms. That period was known as the first AI boom, which computers use search to produce inferences. As the computer’s ability improved, the second AI boom saw computers using knowledge to produce inferences. However, the traditional CAD had its setback, including high development cost, inadequate performance, bad workflow and usage was troublesome, and it supports only specific lesions. The conventional CAD also used a rule-based approach, which was time-consuming and it required the creation of computer algorithm for each lesion in the application,” explained Prof Fujita.

He added, “Today we are in the third AI boom, which computers are known to produce reasoning by learning. This brings us to the concept of deep learning, which is a method of machine learning that mimics the neural network of the human brain. It is able to learn and improve itself. With deep learning, CAD algorithms are also improving.”

Some successful uses of AI-CAD were also shared at the talk, for instance, the MIT having developed AI tools that can predict the onset of breast cancer, and the Google AI accurately assesses CT lung screening scans.

The speaker also enlightened participants on the new development of CAD, which is known as triage-type CAD. Participants learnt that it will be a stroke platform to identity large vessel occlusion (LVO) in CT scans of emergency patients. He also explained that the triage-type CAD functions to connect to a CT scanner, to warn a stroke specialist, also known as a neuropathologist, that a suspected stroke has been confirmed and it will send the image directly to the specialist’s smartphone. He also spoke about the ethical issues of AI-CAD applications, and the possibility of AI replacing clinicians.

At the end of the webinar, Prof Fujita concluded that CAD will become much more powerful along with recent and new AI technology, such as deep learning. The webinar ended with an interactive Q&A session.