Machine learning in oncology: the future is (almost) here

As we delve deeper into the complexity of cancer, data from traditional clinical trials are insufficient to help solve all our clinical questions. Using real-world data could help answer these questions and assist us to progress cancer care, and machine learning has the potential to help us derive supplementary information from large datasets that are already in existence or that can be created from new data collected from patients—i.e. ‘big data’.This technology is already a reality in many fields of medicine, such as radiomics to read radiographic images and the classification of skin lesions in dermatology, although it is mostly in research settings. In oncology, we are still learning how to integrate evidence from this new technology with traditional evidence from clinical trials, and how this might impact clinical practice.At ESMO Congress 2019, some studies report the potential of machine learning to speed up the diagnosis of cancers, such as thyroid cancer (Abstract 1871P), and enhance the diagnostic accuracy of other cancers, such as gastrointestinal stromal tumours (Abstract 1707P) and liposarcomas (Abstract 1708P). Also, we see early evidence of its prognostic value in melanoma (Abstract 1314PD), breast (Abstracts 261P, 176O, 242P, 1917P, 1984PD), bladder, lung and prostate cancers (Abstract 1984PD), and as a tool to enhance our understanding of immune phenotypes in ovarian cancer (Abstract 1875PD).As these studies suggest, machine learning is likely to make a real difference to many aspects of oncology; however, there are many methodological, ethical and legal issues that must first be overcome. These encompass the development of adequate methods to analyse these large amounts of data, integration of machine learning data with traditional research data such as clinical trial data, how to solve issues of contradictory results between these two types of data, and their joint impact on clinical practice guidelines. From an ethical–legal point of view, the provision of informed consent by patients, and conversely, the right of patients to withhold consent and the preservation of patient confidentiality are crucial points.Recognising the growing importance of big data in oncology, ESMO set up a Big Data Task Force—which I chair—to ensure its members are at the forefront of technological advances. Fears around big data and machine learning likely come from a lack of knowledge about the technology rather than negative evidence or experience. Thus, it is extremely important to eliminate these fears by taking measures to tackle any challenges to the integration of machine learning in oncology.Today’s Educational Session ‘The promise of machine learning in medical oncology’ (16.30 – 18.00, Leon Auditorium [Hall 3]).ESMO Congress 2019 abstracts:

  • 1314PD - Cell phenotypes associated with response and toxicity defined by high resolution flow cytometry in melanoma patients receiving checkpoint inhibition
  • 1707P - Radiomics of gastrointestinal stromal tumours, risk classification based on computed tomography images: A pilot study
  • 1708P - Differentiating well-differentiated liposarcomas from lipomas using a radiomics approach
  • 176O - Machine learning-assisted prognostication based on genomic expression in the tumour microenvironment of estrogen receptor positive and HER2 negative breast cancer
  • 1871P - Classification of thyroid nodule using DNA methylation profiling on tissue and circulating tumor DNA
  • 1875PD - Targeting molecular mediators of T cell exclusion for effective immunotherapy in ovarian cancer
  • 1917P - Luminal B breast cancer prognosis prediction by comprehensive analysis of Homeobox genes
  • 1984PD - Functional cell profiling (FCP) of ∼100,000 CTCs from multiple cancer types identifies morphologically distinguishable CTC subtypes within and between cancer types
  • 242P - Clinical validation of CanAssist breast in a Spanish cohort
  • 261P - OncotypeDX® predictive nomogram for recurrence score output: A machine learning system based on quantitative immunochemistry analysis - ADAPTED01

This site uses cookies. Some of these cookies are essential, while others help us improve your experience by providing insights into how the site is being used.

For more detailed information on the cookies we use, please check our Privacy Policy.

Customise settings
  • Necessary cookies enable core functionality. The website cannot function properly without these cookies, and you can only disable them by changing your browser preferences.