Dr. Lu is a Senior Investigator and NCBI’s Deputy Director for Literature Search at the National Library of Medicine (NLM). He leads research in Text Mining, Natural Language Processing (NLP), and Machine Learning and also directs the overall R&D efforts to improve search quality and usability in PubMed.
https://www.nlm.nih.gov/research/researchstaff/LuZhiyong.html
AD version:
Transcript:
Our research focuses on AI and machine learning for processing both biomedical text and image data.
Using machine learning and AI techniques, we address a set of different problems. So these are different examples. We find that it's increasingly difficult for individual researchers to keep up with the rapid growth of the biomedical literature. We're talking about over a million articles in PubMed every year and that means two or three papers every minute. We developed a new search algorithm called "Best match". Our priority is to put the most relevant papers at the top of their search results in the first page or in the second page because that's where most of the user action is and this algorithm is used in PubMed by millions of users on a daily basis. The second example is our response to the COVID-19 pandemic. At the beginning of the pandemic, it was difficult to locate all the papers on COVID-19 because the term "COVID-19" was not used by the research community until later in that year. As a result, there are many different expressions used in the literature to refer to the COVID-19 pandemic. So a simple keyword search would miss many relevant papers. In response, we put together a database called "LitCovid" using machine learning and in particular text classification algorithms that go beyond keyword matching that would find all the papers relevant to COVID-19 regardless what kind of expressions were used by the authors. My last example is our joint work with our clinical researchers at the National Eye Institute. We developed an AI tool called "DeepSeeNet" formation diagnosis and prognosis in retinal diseases such as age-related macular degeneration. Not only can we use retinal images for predicting eye diseases, we can use retinal images to gain insights for other systematic diseases in other parts of the body. In one of the recent works, we used retinal images to predict heart attack with very high accuracy and right now we are pursuing a dissimilar project where we use retinal images to predict brain diseases such as dementia and cognitive function decline. In the long run, what we really aim to do is to teach computers to read and understand scientific papers like scientists, to interpret x-rays or retinal images like radiologists and ophthalmologists, for disease diagnosis at a speed and added accuracy that's above and beyond human ability.