A recent study demonstrates that a computer program, using data from about 500,000 tissue images and powered by artificial intelligence, can effectively diagnose cases of adenocarcinoma, the most prevalent form of lung cancer. Researchers at NYU Langone Health’s Perlmutter Cancer Center and the University of Glasgow are responsible for developing and testing the program. They claim that due to its reliance on a large dataset and advanced technology, the program is capable of accurately diagnosing adenocarcinoma.The software analyzed tumor structure data from 452 adenocarcinoma patients, part of the National Cancer Institute’s Cancer Genome Atlas, which includes over 11,000 patients. It provides an unbiased, detailed second opinion for both patients and oncologists regarding the presence of cancer and the prognosis for its return. The program is independent and has identified the most statistically significant structural features that impact disease severity and tumor recurrence on its own.A new research study published in the journal Nature Communications on June 11 introduced a program, known as histomorphological phenotype learning (HPL), which can accurately differentiate between similar types of lung cancer, such as adenocarcinoma and squamous cell cancers, 99% of the time. It was also found to be 72% accurate in predicting the likelihood and timing of cancer recurrence after treatment, surpassing the 64% accuracy of predictions made by pathologists who directly examined the same patients’ tumor images. The researchers claim that their new HPL program has the potential to significantly improve cancer diagnosis and prognosis.”The potential to offer cancer specialists and their patients a quick and unbiased diagnostic tool for lung adenocarcinoma that, once further testing is complete, can also be used to help validate and even guide their treatment decisions,” said study lead investigator Nicolas Coudray, PhD, a bioinformatics programmer at NYU Grossman School of Medicine and Perlmutter Cancer Center.
“Patients, physicians, and researchers know they can rely on this predictive modeling because it is self-taught, provides explainable decisions, and is based only on the knowledge drawn specifically from each patient’s tissue, The computer program can quickly analyze lung tissue samples to make accurate predictions about cancer recurrence, including information about dying cells, immune cells, and tumor cell density,” Coudray stated. “This is better than current standards for predicting prognosis in lung adenocarcinoma,” stated study co-senior investigator Aristotelis Tsirigos, PhD. Tsirigos is a professor at NYU Grossman School of Medicine and Perlmutter Cancer Center, where he is also the co-director of precision medicine.Tsirigos, a professor and director of its Applied Bioinformatics Laboratories, stated that with the help of tools and other advancements in lung cancer biology, pathologists will shift from examining tissue scans through microscopes to doing so on their computer screens. They will then use AI programs to analyze the images and generate their own detailed breakdown of the tissue’s content, known as a “landscape.” This breakdown might indicate the presence of 5% necrosis and 10% tumor infiltration, and what that implies for survival rates. Such readings may have statistical significance.There is an 80% likelihood of being cancer-free for two years or more, according to data from all patients in the program.
The HPL program was developed by first analyzing lung adenocarcinoma tissue slides from the Cancer Genome Atlas. Adenocarcinoma was selected as the test model due to its characteristic features, such as tumor cells grouping in acinar, or saclike patterns, and spreading predictably along the surface lining of lung cells.
From the analysis of the digitally scanned visual images of the slides, which were divided into 432,231 small quadrants. In a new study, researchers identified 46 key characteristics, which they call histomorphological phenotype clusters, in normal and diseased tissue. Some of these characteristics were found to be statistically linked to either early cancer recurrence or long-term survival. These findings were then validated through additional testing on tissue images from 276 individuals who were treated for adenocarcinoma at NYU Langone between 2006 and 2021.
The researchers aim to utilize the HPL algorithm to assign a score between 0 and 1 to each patient, indicating their statistical likelihood of survival and tumor recurrence over a five-year period. This program is designed to be self-The researchers emphasize that the accuracy of HPL will improve as more data is accumulated. In order to establish trust with the public, the researchers have made their programming code available online and intend to release the HPL tool for free once additional testing is completed.
Factors associated with the recurrence of tumors were high levels of dead cancer cells and lymphocytes, as well as dense clustering of tumor cells in the lung linings. On the other hand, factors linked to higher chances of survival were high percentages of preserved lung sac tissue and low levels of inflammation.mmatory cells.
Tsirigos and his team are planning to develop HPL-like programs for other types of cancer, such as breast, ovarian, and colorectal. These programs will be based on the distinctive morphological features and additional molecular data of each type of cancer. Additionally, they are also working on improving the accuracy of the current adenocarcinoma HPL program by incorporating other data from hospital electronic health records, including information about other illnesses, diseases, income, and home ZIP code.
The study was supported by funding from the National Institutes of Health grant P30CA016087 and United Kingdom Research Council grants Ep/R018634/1.This study was supported by the US National Institutes of Health/National Cancer Institute grants R01 CA160674, R01 CA196643, R01 CA205150, P30 CA016087, T32 CA009161, and T32 CA009161-41, as well as U.S. Department of Defense grants W81XWH-13-1-0263 and W81XWH-13-1-0215, and BB/V016067/1, and European Union Horizon 2020 grant no. 101016851.
In addition to Tsirigos and Coudray, other researchers from NYU Langone involved in this study include Anna Yeaton, Bojing Liu, Hortense Le, Luis Chiriboga, Afreen Karimkhan, Navneet Natula, Christopher Park, Harvey Pass, and Andre Moreira. Co-lead investigator Adalberto Claudio Quiros, co-investigators Xinyu Yang and John Le Quesne, and senior investigator Ke Yuan are all at the University of Glasgow, UK. Co-investigator David Moore is at the University College London, UK.
