Although its official inception was in April 2018, The Center for Artificial Intelligence in Medicine & Imaging (AIMI) at Stanford University was officially born at the computer science department’s 50th anniversary celebration in 2015. Back then, Dr. Curtis Langlotz, Professor of Radiology and Biomedical Informatics and the current AIMI Director heard a talk given by renowned AI researcher and faculty member Dr. Fei Fei Li.

Dr. Langlotz knew Dr. Li was the creator of ImageNet, a large visual image database designed for training object recognition software but he found himself captivated by her latest work on automated captioning. “That really blew me away,” he says. “I thought, hey, that’s what I do for a living. I look for images. I create text describing what’s on those images. Applying AI to medical imaging has always been really exciting,” he recently told the Radiological Society of North America (RSNA).

Duly inspired, Dr. Langlotz set up a lab and enlisted Dr. David Larson, Professor of Pediatric Radiology, who happened to be in the office next door, to lead the first deep learning project on bone age. Gradually, Dr. Langlotz noticed pretty much anyone dealing with clinical images would be using some form of machine learning technique.

He then began receiving feedback on wanting to see more institutional commitment from his lab. With the support from Deans and Chairs of medicine and pathology as well as larger investment, Dr. Langlotz’s lab became what is known as AIMI today. Now, the Center boasts over 120 faculty members, with millions of dollars of sponsored research, and is working closely with the School of Medicine to develop infrastructure for researchers to have easy access to large image databases for their studies.

AIMI also has a seed grant program, offering small grants to labs who are preparing data or developing systems building to larger initiatives. It also has an industrial affiliate program, to collaborate with external vendors or to have spin-off startups, to help translate some of the research work to the market place. The Center hosts regular panel discussions, seminars, a work-in-progress journal club and many other efforts to support the word around improving patient care.

But as Dr. Langlotz explains, one of the really fun things AIMI does is hosting interdisciplinary bootcamps which students attend in exchange for course credits. “That has been the real engine of some of the work that we have done,”says Dr. Langlotz. Faculty, clinical faculty or anyone with great datasets will come to the first day of class and pitch their problems for students to break into groups to solve them.

Dr. Langlotz is equally proud of CheXNeXt, AIMI’s in-house deep learning algorithm developed by Associate Director Dr. Matthew Lungren and his team, as well as their efforts in the public release of databases, which is driving the field forward. CheXNeXt can concurrently detect 14 diseases in chest radiographs. Indeed, their work on chest radiograph interpretation has been cited over 400 times since its publication just two years ago.

Undoubtedly, Dr. Langlotz’s leadership has proved pivotal to AIMI’s success. Back in 2017, he already appeared to have the foresight that AI would play an important role in medicine, particularly in radiology, when he tweeted ‘Radiologists who use AI will replace radiologists who don’t.’

“That Tweet took on a life of its own,” he laughs. “It was just one of those things that came to me one evening and I didn’t think twice about it and suddenly, it was circulating widely.”

But Dr. Langlotz believes we are not even getting close to seeing AI replace humans. “If you look at the AI industry today, a lot of the applications are quite narrow. They are usually single things to detect a nodule in the chest or a haemorrhage in the brain. So it’s more like AI is going to transform the way medicine is being practiced. AI is going to help in the interpretation of larger datasets; extracting additional information from images quickly – something radiologists don’t normally do – finding new ways to underline abnormalities. So it’s all going to be very interesting.”