Tremendous opportunities exist to apply ANN-based predictive analytics in diagnosis management for personalised treatment

Current scene at a hospital: A patient walks into a hospital for a diabetes test – fasting blood sugar, post lunch sugar levels, glycated haemoglobin — and the results are captured.

But this data lies dormant with the patient and the hospital.

If the hospital had an Artificial Neural Network (ANN)-based predictive analytics engine, the results could be benchmarked, the progress based on others who have the same problem can be shown to the patient and the treatment can be discussed with a clear line of sight on how the problem would evolve.

This is just an example of the potential that ANN-based predictive analytics can do to improve the quality of care. In a way, ANNs are akin to the headlights of a vehicle that can help us manoeuvre when there isn’t enough light. The doctors too need smart tools.

Hospitals for some time now have been trying to accomplish an extremely difficult mission: to provide better healthcare, personalised assistance to patients and at reduced costs but the journey hasn’t been easy. The absence of any predictive analytical tools makes their functioning highly reactive and not proactive. But now things are changing.

Data-based Prediction
New mathematical models aided with powerful computer programming create an opportunity to predict, which till a few years ago was considered impossible. These models are called Artificial Neural Networks or ANNs.

Simply put, it means training computers to make decisions. ANNs are adaptive models inspired by how neurons are wired and triggered in our brains, enabling them to process information. The neural networks learn from data, just like our brain learns from every experience. This ‘training’ of data can be achieved by many algorithms, which involve intuitive mathematical models. The availability of high computing machines (HPCs) makes training on volumes of data faster. Developments in information technology and molecular biology have exponentially increased the volumes of data.

In our day to day life, we come across many recommendation engines based on ANNs that tell us what to buy, where to buy, what to eat, how quickly we can get a cab and so on. These are smart applications that have disrupted the way we used to do things and have compelled us to act differently in our lives.

It is now possible to build predictive analytics tools in healthcare too. Spectacular developments in information technology and molecular biology have exponentially increased the volumes of data. Be it clinical data, patient historical data, functional genomics or digital imaging data, hospitals own large data sets per subject. When such large data sets are combined, the strength of relationships between them is unknown and the patterns extremely complex. ANNs are intuitive and powerful methods that are best suited to work with volumes of complex data to extract hidden patterns and make predictions.

ANNs have been used in different clinical settings to predict the effectiveness of instrumental evaluation (echocardiography, brain single photon emission computed tomography, lung scans, prostate biopsy) in increasing diagnostic effectiveness and laboratory medicine.
The most promising application of ANNs relates to the prediction of possible clinical outcomes with specific therapy. ANNs have been applied for cervical cytology, x-Ray mammography and early detection of acute myocardial infarction.

Potential of ANN
ANNs thus have the potential to become a very useful tool to assist the doctors in arriving at a diagnosis and personalising treatment that every individual patient need. The degree of success depends on data quality, the methods used and the expertise of people involved.
Data is the most critical input for the ANNs’ success simply because without data, the algorithm will not be able to produce the right results. An effective application can be built with the help of a team of people who actually understand the business domain, the actual problem, can convert that into a mathematical model, have a firm grip on latest technologies and can think beyond root cause analysis.

We see a tremendous opportunity to apply predictive analytics in diagnosis management for diseases such as diabetes, coronary artery disease and cancer. We choose these specific areas because a lot of research has been done and there is a greater understanding of the problem within the clinician community and even more awareness among those affected. In addition, the line of treatment using multiple devices makes it possible to capture data, which is the key input for training a neural network.

With millions of records available from different sources, the use of ANN in predictive medicine will be all the more important and becomes a very useful tool to assist clinicians in arriving at a decision and personalising treatment to every individual patient’s need. An additional area is also in efficiency improvements and cost reduction in the healthcare ecosystem comprising hospitals and insurers.

Efficiency & Cost Reduction
The key parameter for measuring efficiency for a hospital is: Average Length of Stay (occupancy rate and case mix among others) and its impact on cost/bed and revenue/bed. A Predictive Analytics Generating Engine (PAGE) can be designed to predict how one or more parameters influence the cost and quality of care. By doing so, it would be possible to anticipate the failure modes and have an action plan ready – moving from reactive to proactive.

ANNs will only get better with time. They bring a lot of computational speed combined with exceptionally quick learning abilities. These two features will help hospitals and doctors increase their patient outreach and improve the quality of care by making sure that no relevant information is overlooked while making the diagnosis.

With growing population, changing lifestyles and high incidences of diseases impacting quality of lives, it is time to look beyond the traditional business intelligence models with confidence as neural networks can be a strong force that can disrupt the way we look at healthcare, improve the quality of lives and bring personal touch in many aspects. As the saying goes —It’s not who has the best algorithm that wins; it’s who has the most data and knows what to do with it.

By: Dr Suhas Nimbalkar & Ramakrishna Prasad
(Dr Suhas Nimbalkar is a geneticist and is into ANN-based predictive analytics: This email address is being protected from spambots. You need JavaScript enabled to view it.; Ramakrishna Prasad is a data scientist & machine learning enthusiast: This email address is being protected from spambots. You need JavaScript enabled to view it.)

Source: https://telanganatoday.com/