Researchers have developed a model that is 71% more accurate in predicting the survival expectancy of lung cancer patients. The results show that machine learning has the potential to assist health professionals in their decisions about treatment of the disease. 

The Penn State Great Valley research team conducted a study in which they developed a deep learning model that performs significantly better than the traditional machine learning models the team had previously tested. Indeed, the old models showed an accuracy rate of approximately 61% in predicting the survival expectancy of lung cancer patients, which is much lower than the 71% accuracy rate achieved with the deep learning model. The research was published in the prestigious International Journal of Medical Informatics.

What is Deep Learning?

Deep learning is an innovative type of machine learning based on artificial neural networks, which are typically based on the functioning of the human brain’s own neural network.

Breaking new ground in patient care

The deep learning model allows the analysis of large amounts of data such as cancer type, tumour size, tumour growth rate and demographics. The team explained the benefits that this technology can bring to patient care, pointing out that the information it is able to process could help healthcare professionals make better decisions on topics such as drug selection, resource allocation, or determining the intensity of care for patients. 

In fact this is a very powerful and accurate system that aims to help doctors make these important decisions about how to care for their patients. However, this support tool cannot be used to replace doctors in their decisions about lung cancer treatment.

Nevertheless, the potential of the method is undeniable, as it can provide the solid analysis needed for cancer research. It is therefore particularly well suited to tackling the prognosis of lung cancer. Deep learning is a machine learning algorithm that establishes associations between the data itself and the labels we use to describe the sample data. By making these associations, it learns from the data.

Deep learning can effectively complement such data science tasks, especially when they involve the processing of large amounts of information, such as patient records. 

Deep learning can go further than humans or other traditional machine learning methods.  Before going on to explain that the former has “a simple structure of neural network layers. In each layer, you have a group of cells”, while “in deep learning, there are many layers of these cells that can be structured in a sophisticated way to better transform and extract features, giving you the ability to further improve the accuracy of any model”.

Ultimately, the researchers plan to further improve the model and test its ability to analyse other types of cancers and medical conditions.