Technicalities in AI
According to experts, artificial intelligence algorithms have demonstrated remarkable progress in image-recognition tasks. Dr Prashant Warier, CEO, Qure.ai in his presentation, discussed the technicalities of AI in radiology. He began by explaining the advances in AI and its application in radiology. “All the advances in AI comes from a class of algorithm called deep neutral networks that can now perform tasks that previously required human expertise”, he mentioned.
In order to promote AI in radiology and to create an ecosystem that supports the transformation associated with this requires creation of strategies, training programmes that provides clear understanding of machine learning etc. He also mentioned that in the future, radiologists’ need to develop algorithms. This will lead to further engagement of radiologists rather than replacement.
Delving deep into understanding the kind of research happening in the field of AI and diagnostic imaging in India, Dr Warier said that research in AI should be focussed on small images. His presentation also highlighted the need for training in AI application. However, appropriate training programmes will require huge data. “Training deep learning models requires a lot of data”, he said. Therefore the imaging sector will need to work on an appropriate data collation system. Moreover, warning on the misuse of data he cautioned, “There should be a check of algorithm on own data before using it anywhere.”
Further, discussing on the technicalities he spoke sensitivity and specificity of AI application in radiology. “There is always a trade-off between sensitivity and specificity,” he maintained.
Moving ahead, he spoke on the visualisation aspects and informed, “Visualisation of algorithm output is critical.”
Watch full video here: