If you haven’t heard of Chat GPT or other programs of its ilk, where have you been hiding?
Chat GPT and other software like it are natural language processing (NLP) artificial intelligence chatbots.
They are trained on a massive amount of text data from the internet and can generate human-like text responses to a given prompt.
Clinical NLP is a specialisation of NLP that allows computers to understand the rich meaning behind a doctor’s written analysis of a patient.
While NLPs sound smart in theory, they are currently prone to being inaccurate due to the technology still being in its infancy.
Dentists need their work and records to be domain specific, which is where PhD candidate Hanna Pethani and her team’s system comes in.
The World Health Organisation’s (WHO) Global oral health status report found that out of every disease, condition or category, oral diseases are the most prevalent, whether in a high- or low-income country.
The report’s suggested approach is better information around the social determinants of health and how they interact.
In response to this, Hanna and her team are developing a domain specific ChatGPT-like system for oral health records.
Hanna says that in general, oral health doesn’t collect data very well.
"We keep saying that oral diseases are generally speaking a social disease that tend to impact the socially disadvantaged more than anyone else, but we've never really explored the how and why,” explains Hanna, who is undertaking a Doctor of Philosophy (Medicine and Health) in Biomedical Informatics and Digital Health.
Supervised by Professor Adam Dunn and Professor Heiko Spallek, Hanna, who is retraining in dentistry after previously studying accounting and finance and working in auditing, is hoping to develop a system that can pick up on social determinants of health from oral health professional's clinical notes.
In partnership with the CSIRO, the University of Melbourne and the University of Utah, this project hopes to unlock information, such as whether someone was experiencing homelessness or a substance abuse problem, to determine how social determinants interact.
“Being able to have more information to do an analysis on could help get more answers,” notes Hanna.
Right now, there is no systematic data collection for this other than Australian Bureau of Statistics (ABS) surveys, which are typically expensive and labour intensive.
Hanna's method automatically collects information without the patient or oral health practitioner having to do anything different.
Patients go to their dentist, the dentist writes notes on the appointment and the system picks up on the notes to build data.
“Even in the medical world there’s not much information on how social determinants interact,” says Hanna.
“By collecting this we are hoping to see how these things interact - the fact that you're employed but smoke 50 cigarettes a day, what does that mean for your oral health?
“The answers might be surprising, but no one's looked into it with a great deal of detail.”
Once the system is developed and its performance evaluated, the collected data can be examined to identify gaps in how dentists are recording information as well as themes among oral health problems.
“Maybe dentists don’t realise that their notes could be used for such a big purpose – if they know the possibility and are asking the right questions to record the most useful data, it unlocks many more research opportunities,” says Hanna.