Health

AI helps rule out cancer in dense breasts

An example of a deep Shapley Additive Description (SHAP) overlay image. The Maximum Intensity Projection (MIP) image is on the left and the MIP image with SHAP overlay is on the right. Positive SHAP values ​​(red) indicate areas where lesions are likely to be present, and negative SHAP values ​​(blue) indicate areas with low probability. (A) Contrast-enhanced breast MRI scan sagittal MIP image of invasive ductal carcinoma in situ in a 57-year-old woman of Breast Imaging Reporting and Data System (BI-RADS) Category 4. The deep learning (DL) model has a 90% probability of lesion presence. A positive SHAP value (red) has been shown to match the location of the lesion (arrow). (B) Sagittal MIP images of contrast-enhanced breast MRI scans of lesion-free breasts in a 53-year-old woman with a BI-RADS1 score. In the DL model, the probability of lesions being present was 11%. Negative SHAP values ​​(blue) are diffusely distributed in the breast area. (C) Lateral MIP images of contrast-enhanced breast MRI scans of ductal carcinoma in situ in a 65-year-old woman with a BI-RADS4 score. In the DL model, the probability of lesions being present was 32%. This is the lowest probability of all breasts with malignancy in our study. A positive SHAP value (red) has been shown to match the location of the lesion (arrow).Credit: Radiological Society of North America

An automated system that uses artificial intelligence (AI) quickly and accurately sifts breast MRIs of densely breasted women, eliminating cancer-free women and allowing radiologists to focus on more complex cases. To Radiology..

Mammography breast cancer By providing early detection when cancer is most treatable. However, women with very dense breasts are less sensitive than women with fatty breasts. In addition, women with very dense breasts are 3 to 6 times more likely to develop the disease. chest Cancer than women who have almost completely fatty breasts and are twice as risky as the average woman.

supplement Sifting In women with very dense breasts, cancer detection. A large Dutch-based study of the High Density Tissue and Early Breast Neoplasmic Screening (DENSE) trial supported the use of supplemental screening with MRI.

“The DENSE trial has shown that additional MRI screening is beneficial for women with very dense breasts,” said Erik Verburg, principal investigator at the Institute of Imaging Sciences at the Utrecht University Medical Center in the Netherlands. , M.Sc. Says. “On the other hand, the DENSE trial confirmed that the majority of women screened had no suspicious findings on MRI.”

Most MRIs show normal anatomical and physiological variations that may not require a radiological review, so how to triage these normal MRIs to reduce the workload of the radiologist. Is required.

In the first study of this kind, Verburg et al. Set out to determine the feasibility of automated triage methods based on: Deep learning, A sophisticated type of AI. They used breast MRI data from the DENSE trial to develop and train a deep learning model to distinguish between lesioned and unaffected breasts. The model was trained with data from 7 hospitals and tested with data from 8th hospital.

Very over 4,500 MRI datasets Rich breasts Was included. Of the 9,162 breasts, 838 had at least one lesion, 77 of which were malignant and 8,324 had no lesions.

NS Deep learning model 90.7% of lesioned MRIs were considered abnormal and were triaged for a radiological review. Without missing cancer, we rejected about 40% of lesion-free MRIs.

“I showed that it can be used safely. Artificial intelligence Reject breast screening MRI without missing a malignant disease, “Babbergh said. 40 percent is a good start. However, it still needs to be improved by 60%. “

AI-based triage systems have the potential to significantly reduce the workload of radiologists, Barbergh said.Nearly 82,000 in the Netherlands alone woman You may be eligible for biennial MRI breast screening based on breast cancer.

“This approach can be used first to help radiologists reduce overall reading time,” says Verburg. “As a result, more time may be available to focus on the most complex breast MRI scans.”

Researchers plan to validate the model on other datasets and deploy it in subsequent screening rounds of the DENSE trial.


For women with dense breasts, MRI is cost effective in detecting breast cancer


For more information:
Stefanie GA Veenhuizen et al, Supplemental Breast MRI for Women with Very Dense Breasts: Results of the Second Screening Round of the DENSE Trial, Radiology (2021). DOI: 10.1148 / radiol.2021203633

Quote: AI is for high density breasts (October 5, 2021) obtained from https://medicalxpress.com/news/2021-10-ai-cancer-dense-breasts.html on October 5, 2021. Helps to rule out cancer

This document is subject to copyright. No part may be reproduced without written permission, except for fair transactions for personal investigation or research purposes. The content is provided for informational purposes only.



AI helps rule out cancer in dense breasts Source link AI helps rule out cancer in dense breasts

Related Articles

Back to top button