A better AI-based tool for mesothelioma

Our plain-English summary of A better AI-based tool for mesothelioma, by D. Romero (Nature Reviews. Clinical Oncology, 2019).

The study

• Author: D. Romero
• Journal: Nature Reviews. Clinical Oncology
• Published: 2019;16(12):722
• Type of article: Comment
• Cited by: 1 article, as counted by NIH iCite
• Read it: PubMed · Publisher

In plain English

This one-page news item (a 'Research Highlight') did not report new research. It summarized a study by Courtiol and colleagues published in Nature Medicine. That study described MesoNet, a deep-learning tool that gave an overall survival score to whole-slide tissue images from patients with malignant pleural mesothelioma. The tool was developed in a training set (n = 2,300) and a testing set (n = 681), then checked in 56 images from The Cancer Genome Atlas. According to the highlight, MesoNet predicted survival better than the standard histology-based classification, with c index values of 0.643 versus 0.598 in the testing set. In the testing set, 60 patients whom MesoNet predicted to have shorter survival had a median survival similar to the 7.2 months seen in the 60 patients with sarcomatoid tumours, although only 34% of the MesoNet group had sarcomatoid tumours. Image areas linked to the lowest survival scores mainly showed stromal features. The highlight did not discuss asbestos exposure. The author wrote that the tool has the potential to improve diagnostic accuracy and that prospective validation is awaited.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: News summary (Research Highlight) of a published study that developed and tested a deep-learning (artificial intelligence) tool, MesoNet, for predicting survival from tissue slide images; not original research by this author
• Who was studied: Whole-slide images of tissue specimens from patients with malignant pleural mesothelioma: a training set (n = 2,300) and a testing set (n = 681), with further validation in 56 whole-slide images from The Cancer Genome Atlas. No location or dates are given.
• Main results:
  • MesoNet predicted overall survival better than histology-based classification in all three sets (c index): training 0.642 versus 0.596, testing 0.643 versus 0.598, TCGA 0.656 versus 0.590.
  • In the test set, 60 patients with sarcomatoid tumours had a median overall survival of 7.2 months. The 60 patients MesoNet predicted to have shorter survival had a similar median survival, but only 34% of them had sarcomatoid tumours.
  • In the test set, 80 patients with grade 1 epithelioid tumours had a median overall survival of 28.1 months. The 80 patients MesoNet predicted to have longer survival all had epithelioid tumours, but of mixed grades (46% grade 1).
  • Image tiles with high survival scores usually showed tubular structures and good blood supply. Tiles with low survival scores mainly showed stromal features.
  • In patients with epithelioid tumours and shorter survival, 39% of tiles showed features linked to poor survival (described as including inflammation, cellular diversity, vacuolization and stromal structures).
• The authors' conclusion: The highlight's author wrote that the tool has the potential to improve accuracy in diagnosing malignant pleural mesothelioma and other cancers.
• Limitations the authors state:
  • The tool has not yet been validated prospectively.

Topics

The National Library of Medicine indexes this study under subjects our site covers: Lung cancer, Mesothelioma.

Medical terms in this study, explained in our Lexicon: Mesothelioma.

Related studies

• Factoring-in agglomeration of carbon nanotubes and nanofibers for better prediction of their toxicity versus asbestos — A.R. Murray, E.R. Kisin, et al. (2012)
• A meta-analysis of asbestos and lung cancer: is better quality exposure assessment associated with steeper slopes of the exposure-response relationships? — V. Lenters, R. Vermeulen, et al. (2011)
• Women with malignant pleural mesothelioma have a threefold better survival rate than men — E. Taioli, A.S. Wolf, et al. (2014)
• Loss of heterozygosity of chromosome 3p21 is associated with mutant TP53 and better patient survival in non-small-cell lung cancer — C.J. Marsit, M. Hasegawa, et al. (2004)
• Low homozygous/high heterozygous deletion status by p16 FISH correlates with a better prognostic group than high homozygous deletion status in malignant pleural mesothelioma — M. Hamasaki, S. Matsumoto, et al. (2016)
• PD-L1 Testing for Immune Checkpoint Inhibitors in Mesothelioma: For Want of Anything Better? — S. Lantuejoul, N. Le Stang, et al. (2017)

Browse more research in Nature Reviews. Clinical Oncology, or return to medical research.