In a recent study published in Nature Communications, a team of researchers used artificial intelligence (AI) to classify histopathological images and differentiate between endometrial cancer subtypes. The tool identified a subtype of endometrial cancer known as NSMP or No Specific Molecular Profile, which is characterized by aggressive disease and low survival rates.
Study: AI-based histopathology image analysis reveals a distinct subset of endometrial cancers. Image Credit: megaflopp/Shutterstock.com
Background
Endometrial cancer is divided into four subtypes, each requiring different treatments and having different outcomes.
Currently, classifying these subtypes is based on unreliable clinical and pathological methods, leading to inconsistent and inaccurate assessments. This results in either too much or too little treatment, causing recurrence and sometimes death.
The Cancer Genome Atlas project has shown that using advanced genetic techniques can better classify endometrial cancer into four subtypes based on specific genetic mutations.
Moreover, AI tools with deep learning models are increasingly being used in medicine to analyze large amounts of data. These tools help identify potential biomarkers and improve cancer diagnosis.
About the study
In this study, researchers created an AI tool using deep-learning to analyze histopathological images and distinguish between two subtypes of endometrial cancer: NSMP and p53 abnormal (p53abn).
Previously, they had developed a molecular classification system that categorized endometrial cancer into four subtypes for clinical use:
- POLE mutant subtype: Features pathogenic mutations in the POLE gene, which is involved in DNA proofreading and repair.
- Mismatch repair deficient (MMRd) subtype: Identified by the absence of key mismatch repair proteins through immunohistochemistry tests.
- p53 abnormal subtype: Detected by abnormalities in the p53 tumor suppressor protein via immunohistochemistry.
- NSMP subtype: Diagnosed by excluding the features of the other three subtypes.
In this study, the AI tool was used to analyze histopathological images to differentiate between NSMP and p53abn subtypes. Researchers hypothesized that some NSMP tumors resemble p53abn tumors histologically. By applying deep-learning models to stained tissue slides, they aimed to identify this subset.
The study included tissue samples from 368 endometrial cancer patients in a discovery cohort, with validation from two independent cohorts of 614 and 290 patients. Researchers also performed shallow whole-genome sequencing to analyze copy number and gene expression profiles of both subtypes and p53abn-like NSMP samples from the validation cohort.
Results
The study found that AI analysis of histopathological images successfully identified a subset of NSMP endometrial cancer patients with significantly lower survival rates and more aggressive tumors.
This aggressive subset accounted for nearly 20% of NSMP tumors and 10% of all endometrial cancers.
The results indicated that traditional methods like clinicopathological features, immunohistochemistry tests, next-generation sequencing, and gene expression profiles could not differentiate between p53abn subtypes and these p53abn-like NSMP cases.
The deep learning model also detected tumors with TP53 mutations that appeared normal in p53 immunostaining, which would have been false negatives with traditional immunohistochemistry.
The AI tool could identify aggressive p53abn-like cancers within the NSMP subtype, even when pathological and molecular features failed to predict poor survival outcomes.
Shallow whole-genome sequencing revealed that this NSMP subset had more altered and unstable genomes, similar to the p53abn subtype but with less instability.
The findings provided evidence of histopathological differences in this subset, despite the lack of distinctions through traditional pathological or immunohistochemical methods.
Conclusions
Overall, the findings indicated that the AI-based image classifier was able to distinguish between subsets of endometrial cancer patients and detect a subset with significantly inferior survival outcomes.
The researchers believe that this AI-based tool can easily be incorporated into the clinical diagnostic process to scan histopathological images routinely.
Furthermore, with additional refinement, this AI-based tool could potentially replace the more time-consuming and expensive method of molecular marker-based diagnosis.
Air pollution linked to head and neck cancer risk