A group of researchers are predicting big improvements in how computers diagnose and treat disease. A new article published in Nature Machine Intelligence reviews the potential of a new hybrid machine learning technique, ensemble deep learning, to improve the performance (e.g. accuracy, reliability) of computer models used in medical research. By using ensemble deep learning, medical researchers can integrate knowledge from multiple “experts” to recognise and find solutions to treat disease.
Machine learning, when computers learn and find novel patterns from big data, already has many applications in medical research: diagnosis and prognosis of disease, drug discovery and development, and formulating precision medicine treatments. However, the authors of the new article have identified the shortcomings of the two current major machine learning techniques and have reviewed the ways that medical researchers can combine these two styles of machine learning into a more accurate and functional tool. Ensemble deep learning, a hybrid of deep learning and ensemble learning, means precision medicine and drug development research can proceed more quickly and accurately, leading to treatments and cures faster than currently possible.
Dr Pengyi Yang is Group Leader of Computational Systems Biology at Children’s Medical Research Institute in Sydney. As lead author of the new article, he explained why machine learning in medical research is just scratching the surface of what’s possible.
“Just like ‘many heads are better than one’, ensemble deep learning that combines multiple ‘computer brains’ with complementary knowledge has achieved high levels of performance unattainable by traditional methods. This review is timely given the explosion of data seen in the biological and biomedical field.”
The two machine learning components in this hybrid technique, deep learning and ensemble learning, have independently made a substantial impact on medical research.
Deep learning uses a model inspired by neural networks. This allows computers to spot patterns in data (such as images) and learn how to classify new information (such as a previously unseen image with a possible tumour) based on what the computer has learned previously.
Ensemble learning uses multiple models simultaneously to check model predictions against each other and improve the performance of the overall model.
Combining these two techniques, ensemble deep learning greatly improves the accuracy, reliability, and efficiency of computer models. It can also handle datasets with smaller sample sizes and ‘noisier’ data.
Dr Yang and his fellow authors hope to inspire researchers to find completely novel ways of applying machine learning in medical research.
“Perhaps in the future we’ll see ensemble deep learning implemented in hospital computers for diagnosing patients based on their genomic information and medical records (such as CT scans) with expert level of accuracy. These technology advances will reduce the current burden and waiting time in clinics and compliment the expertise of doctors.”
IMAGE SOURCE: Creative Commons
Sign up for Scientific Inquirer’s Steady State Newsletter for the week’s top stories, exclusive interviews, and weekly giveaways. Plenty of value added but without the tax. http://bit.ly/2VEF06u