There’s an exciting buzz going around the world of Cardiology right now, and what’s making that buzz is the new technologies in artificial intelligence and machine learning. With the demand of medicine to increase its operational efficiency and develop a more personalised patient care, AI could not be developing at a better time. But what exactly are its benefits?
AI able to use the data collected from health records to determine a patient’s likelihood of re-admittance, possible diagnoses, and their vulnerabilities to a broad range of other cardiovascular diseases (Johnson et al 2018) taking into account the patient’s medical history. With health records typically having a large amount of information set out in ‘messy’ order, problems can often arise when using traditional statistical models to create predictions. The reason for this is that it’s very difficult for the researcher to decide which variables from health records are relevant enough to add to the model and which ones to leave out. Especially when there are more variables than potential outcomes, it becomes algebraically impossible to make a prediction model.
Trending AI Articles:
1. Tutorial: Stereo 3D reconstruction with openCV using an iPhone camera
2. Machine Learning using Logistic Regression in Python with Code
3. Bursting the Jargon bubbles — Deep Learning
4. How Can We Improve the Quality of Our Data?
What’s more, traditional models of prediction, such as logistical regression, do not count the relationship between the variables (Johnson et al 2018)-for example, how one drug could potentially react with another given the particular condition of the patient. Without including the relationship between such variables, there is likely to be a validity problem that will leave behind inaccuracies.
But AI with its deep learning algorithm and huge data sets utilises those variable relationships, putting them together with multiple sources of evidence of cardiovascular diseases and producing a more personalised diagnosis and providing an exact prognosis (Johnson et al 2018). This would help determine the best medical direction for the patient, such as choice of medication, future referrals and therapy. What’s more, the deep learning algorithm it uses will be constantly improving the AI’s representation of the data as it collects more information and is used more, without the need of human interaction.
Dr Anthony Chang, AIMed founder and chairman, chief intelligence and innovation officer of Children’s Hospital or Orange County (CHOC), said last year on the Pediheart podcast that not too long in the future AI will be able to assist the cardiologist with overall clinical practice, as it will be able to understand natural language and put what has been said straight on the record, instead of relying on a second person to write it up later into a language the machine understands. This creates a machine-human division of labour, cutting costs for the hospital.
The technology is known as natural language processing; its function is primarily to “target at turning texts [such as clinical notes] into machine-readable structured data, which can then be analysed by ML techniques” (Jiang et al 2017). This would remove the burden on physicians to make sure all records are correctly written in the system and up to date, preventing disruptions of hospital workflow.
Such AI has existed for quite a few years already. But the challenge facing medical world right now is transitioning every hospital onto a system that can ‘train’ the AI through inputting clean data of “screening, diagnosis, treatment assignment and so on, so that they can learn similar groups of subjects, associations between subject features and outcomes of interest” (Jiang et al 2017).