Detecting Cancer Early in Slides Using Artificial Intelligence (AI)
Using artificial intelligence (AI) to categorize endometrial biopsy whole slide images (WSI) allows the prioritization of malignant slides, which in turn reduces the time to diagnosis for patients with endometrial cancer, according to new research published in the journal PLoS ONE.
Cancer of the uterus is one of the top ten most common cancers in women worldwide and the fourth most common in the UK. Endometrial cancers are the most common type of cancer of the uterus. An endometrial biopsy is a key step in the diagnosis of endometrial cancer.
‘An Artificial intelligence model could be used to automatically sort endometrial biopsy slides, allowing faster cancer diagnosis.’
The tissue obtained from the biopsies is fixed in formalin, then undergoes a series of laboratory processing steps to generate a slide, including staining with Haematoxylin and Eosin (H&E) before being examined by a pathologist and a report generated.
In this new study, researchers used artificial intelligence (AI) to categorize endometrial biopsy whole slide images (WSI) from digital pathology as either "malignant", "other or benign" or "insufficient". An endometrial biopsy is a key step in the diagnosis of endometrial cancer, biopsies are viewed and diagnosed by pathologists.
Automated Sorting Can Diagnose Endometrial Cancer Faster
Pathology is increasingly digitized, with slides viewed as images on screens rather than through the lens of a microscope. The availability of these images is driving automation via the application of AI.A model that classifies slides in the manner proposed would allow prioritization of these slides for pathologist review and hence reduce the time to diagnosis for patients with cancer. Previous studies using AI on endometrial biopsies have examined slightly different tasks, for example using images alongside genomic data to differentiate between cancer subtypes.
They took 2909 slides with "malignant" and "other or benign" areas annotated by pathologists. A fully supervised convolutional neural network (CNN) model was trained to calculate the probability of a patch from the slide being "malignant" or "other or benign".
Heat maps of all the patches on each slide were then produced to show malignant areas. These heatmaps were used to train a slide classification model to give the final slide categorization as either "malignant", "other or benign" or "insufficient".
The final model was able to accurately classify 90% of all slides correctly and 97% of slides in the malignant class; this accuracy is good enough to allow prioritization of pathologists' workload.
The final model is in two stages. Firstly, the very large images are split into smaller patches and a deep learning model is trained to classify each patch as malignant or not. Next, a second stage model combines the small patches back together and predicts a classification for the whole slide, this compensates for noise in the predictions of the first stage.
Source: Eurekalert