Scientists Say Machine Learning Could Help Diagnose Mesothelioma Earlier

Researchers in Toronto say machine learning could help doctors diagnose mesothelioma earlier.

Molecular geneticists at the University of Toronto teamed up with cancer and heart doctors to conduct the new research.

The team collected data from hundreds of suspected mesothelioma cases and ran it through different statistical analysis programs. The goal was to see which one was best at determining who had pleural mesothelioma and who did not.

The results revealed some important information – both about the machine learning tools and about how doctors might diagnose mesothelioma earlier, even without them.

Why Doctors Need to Diagnose Mesothelioma Earlier

Mesothelioma is almost always caused by asbestos exposure. In the first few years of the disease, there may be no mesothelioma symptoms.

When patients do develop mesothelioma symptoms like chest pain and shortness of breath, it can mean that the disease is in an advanced stage.

Mesothelioma diagnosis usually involves a number of tests. Patients may have imaging scans, blood tests, and a biopsy. By the time mesothelioma is finally diagnosed, it may already be too late for treatment to help.

Right now, only a small percentage of mesothelioma patients live beyond 18 months. If doctors could find a way to diagnose mesothelioma earlier, it could improve the odds of surviving it.

How Machine Learning Could Help

Machine learning is the process by which artificial intelligence can “learn” to get better at analyzing data, by analyzing a lot of it. There are many different kinds of machine learning algorithms and they all work a little differently.

To see which one might help diagnose mesothelioma earlier, the Toronto researchers used data from the health records of 324 Turkish people. The people were all exposed to asbestos and had mesothelioma symptoms.

After testing five different types of machine learning tools, they concluded that one called “random forest” was the most effective. It was able to quickly and accurately identify the people who ended up having mesothelioma.

“Our results show that machine learning can predict diagnoses of patients having mesothelioma symptoms with high accuracy, sensitivity, and specificity in a few minutes,” concludes lead researcher Davide Chicco.

Two Symptoms May Also Boost Efforts to Diagnose Mesothelioma Earlier

Another important fact to come out of the study are the two features that point to mesothelioma more than all the others.

The Random Forest tool showed that lung side and platelet count could be used to diagnose mesothelioma earlier.

The researchers say even doctors who do not have access to artificial intelligence could be pretty confident in making a mesothelioma diagnosis based largely on these factors.

“The importance of pleural plaques in lung sides and blood platelets in mesothelioma diagnosis indicates that physicians should focus on these two features when reading records of patients with mesothelioma symptoms,” concludes. Dr. Chicco.

Source:

Chicco, D, et al, “Computational prediction of diagnosis and feature selection on mesothelioma patient health records”, 2019, PLoS One, https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0208737