NYU Researchers Using Google AI to Determine Lung Cancer Type

Researchers at NYU School of Medicine have recently developed an artificial intelligence (AI) system that is capable of analyzing lung tumors to identify specific types of cancers, and even identify genetic abnormalities. Published in Nature Medicine, their study showed that this machine learning program is capable of differentiate between squamous cell carcinoma and adenocarcinoma with 97% accuracy. These two types of lung cancer are particularly difficult for physicians to identify without extensive testing.

In the clinical study, the AI platform was found capable of detecting mutated forms of 6 different genes tied to lung cancer were present in cells. The accuracy range attained was 73-86% depending on which gene was analyzed. Researchers developed this impressive AI system by retraining Google’s Inception v3, a neural network originally conceived for object identification. The images analyzed in the study were attained from The Cancer Genome Atlas, a database managed by the National Cancer Institute (NCI) and the National Human Genome Research Institute (NHGRI) that contains images of cancers with known diagnoses.

Inception v3 was tested on 1,634 images from this database to compare its rulings with the known diagnoses to determine its efficacy. Once the system proved capable of diagnosing general cancerous cells with 99% accuracy, the AI was tested on the common adenocarcinoma and squamous cell carcinoma forms of lung cancer to yield the range listed above.

Among the researchers’ findings was that roughly half of the cancer images that the AI misinterpreted were also falsely diagnosed by professional pathologists, showing how difficult it is to differentiate between the two tumors. The AI bolstered its ability to compliment a physician’s diagnosis however, with 45 of the 54 images misinterpreted by physicians being correctly diagnosed by the system.

“Delaying the start of cancer treatment is never good,” says study author Dr. Aristotelis Tsirigos, professor of Pathology at NYU School of Medicine and NYU Langone Health’s Perlmutter Cancer Center. “Our study provides strong evidence that an AI approach will be able to instantly determine cancer subtype and mutational profile to get patients started on targeted therapies sooner.”

With over 200,000 people being diagnosed with lung cancer every year, this AI system could have a profound impact in the diagnostic side of oncology. The NYU research team has plans to continue this AI training process until the system is capable of identifying which specific genes contain mutations in a cancer with over 90% accuracy. Once this is achieved, the researchers will initiate the process of attaining government approval for clinical applications of their technology in cancer diagnostics.

VB > NYU applies open source Google AI to diagnose lung cancer > https://t.co/q118glBXjD #BigData #DataScience #BusinessIntelligence #BI #ArtificalIntelligence #AI #CognitiveComputing #CC #Blockchain #Technology #Analysis #Research

— Big Data Love ?❤️ (@BigDataLove) September 17, 2018

Sources: Science Daily, Venture Beat