Join our daily and weekly newsletters for the latest updates and the exclusive content on AI coverage. Learn more
Patient data files can be convoluted and sometimes incomplete, which means that doctors do not always have all the information they easily need. This adds the fact that health professionals cannot follow the case study dam, research documents, trials and other cutting -edge developments from industry.
Based in New York Nyu Langone Health found a new approach to meet these challenges for the next generation of doctors.
The Academic Medical Center – which includes the Nyu Grossman School of Medicine and the Nyu Grossman Long Island School of Medicine, as well as six hospital for hospitalized patients and 375 ambulatory locations – has developed a broad language model (LLM) which serves as a companion to companion Research respected and medical advisor.
Each night, the model treats electronic health files (DSE), by matching them with relevant research, diagnostic advice and essential general information which he then provides in concise and personalized emails to residents The next morning. This is elementary part of the pioneering approach of Nyu Langone to the schooling of medicine – what she calls “precision medical training” which uses AI and data To provide highly personalized student trips.
“This concept of” clarification in everything “is necessary in health care,” told Venturebeat Marc Triola, a associate dean for educational IT and director of the Innovations in Medical Education in Nyu Langone Health. “It is clear that the evidence emerges that AI can overcome many cognitive biases, errors, waste and ineffectiveness in the health system, that it can improve diagnostic decision -making.”
How Nyu Langone uses Llama to improve patient care
Nyu Langone uses an open weight model built on the latest version of Llama-3.1-8B-Instruct and Open-Source Chromium vector database for Recovery generation (CLOTH). But it is not only a question of accessing documents – the model goes beyond the cloth, actively employing research and other tools to discover the latest research documents.
Each night, the model connects to the DSE database of the establishment and removes medical data for patients seen in Langone the day before. He then searches basic information on diagnostics and medical conditions. Using a Python API, the model also searches for related medical literature in AdvertisingWho has “millions and millions of papers,” said Triola. The LLM reviews opinions, deep diving articles and clinical trials, selecting some of the apparently the most relevant and “puts all this in a pretty email”.
Early the next morning, medical students and internal medicine, neurosurgery and residents in radiotherapy receive a personalized email with detailed summaries of patients. For example, if a patient with congestive heart failure had been carried out for an examination the day before, the email will provide refreshment on basic pathophysiology of heart conditions and information on the latest treatments. It also offers questions of self-study and organized AI medical literature. In addition, this can give advice on the steps that residents could take the next or the actions or details they may have ignored.
“We have obtained excellent comments from students, residents and the faculty on how it keeps them updated without friction, how they incorporate this in the way they make choices concerning the care plan of a patient “Said Triola.
A key success metric for him was when a system breakdown interrupted emails for a few days – and the teachers and the students complained not to receive the kicks of the morning on which they had come.
“Because we send these emails just before our doctors start rounds – which is among the craziest and busiest moments of the day – and for them to notice that they did not receive these e- Mail and lacked them as part of their thinking they were great, “he said.
Transform the industry with precision medical education
This sophisticated AI recovery system is fundamental to Nyu Langone’s precision education model, which Troli explained is based on digital data “without density, without friction”, AI and strong algorithms.
The establishment has collected large amounts of data over the past decade on students – their performance, the environments in which they take care of patients, the notes of DSE they write, the clinical decisions They do and the way they reason through the interactions and care of patients. In addition, Nyu Langone offers a large catalog of all the resources available for medical students, whether videos, self-study or examination questions or online learning modules.
The success of the project is also thanks to the rationalized architecture of the medical establishment: it has centralized IT, a unique data warehouse on the health side and a unique data warehouse for education, allowing Langone d ‘marry its various data resources.
The head of medical information Paul Testa noted that major AI / ML systems are not possible without excellent data, but “it is not the easiest thing to do if you are sitting on data not managed in silos in your system ”. The medical system can be important, but it works as “a patient, a file, a standard”.
Gen Ai allowing Nyu Langone to move away from education “a single size”
As Triola said, the main question of his team sought to approach is: “How do they connect the diagnosis, the context of the individual student and all these learning materials?”
“All of a sudden, we have this big key to unlock this: the generating AI,” he said.
This allowed the school to move away from a “one -sized” model which was the norm, which the students intended to become, for example, a neurosurgeon or a psychiatrist – very different disciplines which require unique approaches.
It is important that students receive a tailor-made education throughout their schooling, as well as “educational kicks” that adapt to their needs, he said. But you can’t just tell the teachers to “spend more time with each student” – it’s humanly impossible.
“Our students were hungry for this, because they recognize that this is a period of high speed change in medicine and generator,” said Triola. “It will absolutely change … what it means to be a doctor.”
Serve as a model for other medical institutions
Not that there were no challenges along the way. The technical teams have notably worked by the “immaturity” model.
As Triola noted: “It is fascinating to see how expanded their integrated knowledge is, and sometimes how much. It will work perfectly, predictable, 99 times in a row, then at the 100th time, it will make an interesting set of choice. »»
For example, at the start of development, LLM could not make the difference between an ulcer on the skin and an ulcer in the stomach, which is “not linked at all,”, explained Triola. His team has since focused on rapid refining and earthing, and the result was “remarkable”.
In fact, his team is so confident in the battery and the process they believe that it can be an excellent example to follow. “We promote open source and open weight because we wanted to arrive at the point where we could say:” Hey, other medical schools, many of which have many resources, you can do so cheap »» Triola explained.
Testa agreed: “Is it reproducible?” Is this something we want to spread? Absolutely, we want to disseminate it in health care. »»
Revalle the practices of “sacrosanct” in medicine
Naturally, there is a lot of worries through the industry regarding nuanced biases that could be cooked in AI systems. However, Triola stressed that it was not a huge concern in this case of use, because it is a relatively simple task for AI. “This is research, he chooses from a list, he summarizes,” he noted.
On the contrary, one of the largest concerns on the surface concerns non-qualifying or detachment. Here is a correlation: those of a certain vintage will remember to have learned cursive in primary school – but they have probably forgotten the competence because they have found rare occasions to use it in their adult life. Now he is almost obsolete, rarely taught in primary education today.
Triola pointed out that there are “sacrosophying” parts to be a doctor, and that some resist them to abandon them to AI or digital systems “in any way whatsoever, form or form”. For example, there is a perception that young doctors should actively seek and nose in the last literature whenever they are not in a clinical setting. But the quantity of medical knowledge available today and the “frantic rhythm” of clinical medicine requires a different way of doing things, said Triola.
When it comes to looking for and recovering information, he noted: “AI does better, and it is an uncomfortable truth that many people hesitate to believe.”
Instead, he posed: “Let’s say that it will give superpowers to doctors and understand the co -pilot relationship between man and AI, not the competitive relationship of who will do what.”
