Over the past weekend, Budapest, Hungary played host to the EUSOBI Annual Scientific Meeting 2019. Two particular sessions stood out in their focus on artificial intelligence (AI): the keynote lecture, which took place on Friday, October 4, entitled “A step into the near future, what to expect from AI in radiology”, by Dr. Bram van Ginneken, and a lecture in the “Novel development” session by Dr. Katja Pinker-Domenig, entitled “Advanced post-processing and AI”.
We reached out to the EUSOBI Scientific Committee Chair, Dr. Ritse Mann, for a brief statement on the EUSOBI Annual Scientific Meeting and why the topic of AI is one of importance in breast imaging. “EUSOBI is the foremost scientific conference on breast imaging in the world,” stated Mann. “Therefore, it is obvious that we have a significant amount of interest on the value of AI for breast imaging.” Mann further explained that due to the large number of imaging studies in this field, and the relative importance of imaging in cancer detection and therapy selection, the potential for AI implementation in breast imaging is very high. And, given the rapid expansion of research in AI for breast imaging and the high number of start-up companies on breast AI, this potential is well recognized. “Still, as a scientific community, we need to remain vigilant on the real value AI provides for our patients, and to which extent it is mainly hype,” said Mann.
“At EUSOBI, the keynote lecture by Dr. Bram van Ginneken provided a wide scope on the possibilities and limitations of AI in medical imaging and painted the framework required for validation and clinical implementation of AI techniques in medical practice,” explained Mann. Furthermore, seen in several other presentations, AI techniques now feature in the background, already partly integrated into the clinical problems addressed by the different speakers. AI is becoming more commonplace, with 2 of 4 best-rated abstracts addressing possibilities for the implementation of various AI systems in clinical practice.
“While the obvious end-goal would be full integration of AI into clinical tools, we are currently not nearly at that level,” stated Mann. “Hence, we were of the opinion that a dedicated lecture on the current status and potential of AI applications for breast imaging would also still be of substantial interest.” The second lecture mentioned earlier, delivered by Katja Pinker-Domenig on Saturday morning, October 5, provided a well-balanced, real-world view of the current possibilities and impossibilities of AI in breast imaging.
In “Advanced post-processing and AI”, Dr. Katja Pinker-Domenig discussed the application of AI in biomedical imaging and how the use of AI for image analysis is slowly shifting the field of radiology from a skill of perception to a more objective science. “AI is rapidly moving from an experimental phase to an implementation phase in medicine, with radiology being a prime candidate for early adoption of AI,” said Pinker-Domenig in a statement to the ESR’s AI blog. She continued, “The implementation of AI in radiology is expected to significantly improve the quality, value, and depth of radiology’s contribution to patient care and population health and will revolutionize radiologists’ workflows.”
Because breast cancer screening is one of the best known and most researched areas when it comes to use-cases for AI in radiology, mammography was one of the earliest areas to incorporate AI tools and techniques. “In breast imaging, the application of AI tools is an unprecedented opportunity to better derive clinical value from imaging data and may reshape the way we care for our patients with respect to breast cancer detection, prediction, and prognosis,” stated Pinker-Domenig. “However,” she continued, “we are currently in the infancy of AI implementation in breast imaging and continued study, rigorous standardization, and further studies with large datasets and with subgroup analysis by patient group and/or tumor type, and subsequent independent testing are necessary to allow meaningful clinical implementation.” The evolution from computer-aided detection (CAD) systems in breast cancer screenings to the increased use of AI tools and how AI can continue to improve breast cancer detection, prediction, and prognosis is beneficial for both radiologists and patients alike.
