AutoCumulus: an Automated Mammographic Density Measure Created Using Artificial Intelligence
Al-qershi, O.; Nguyen, T. L.; Elliott, M. E.; Schmidt, D. F.; Makalic, E.; Li, S.; Fox, S. K.; Dowty, J.; Pena-Solorzano, C. A.; Kwok, C. F.; Chen, Y.; Wang, C.; Lippey, J.; Brotchie, P.; Carneiro, G.; McCarthy, D. J.; Jeong, Y.; Sung, J.; Frazer, H. M.; Hopper, J. L.
Show abstract
BackgroundMammographic (or breast) density is an established risk factor for breast cancer. There are a variety of approaches to measurement including quantitative, semi-automated and automated approaches. We present a new automated measure, AutoCumulus, learnt from applying deep learning to semi-automated measures. MethodsWe used mammograms of 9,057 population-screened women in the BRAIx study for which semi-automated measurements of mammographic density had been made by experienced readers using the CUMULUS software. The dataset was split into training, testing, and validation sets (80%, 10%, 10%, respectively). We applied a deep learning regression model (fine-tuned ConvNeXtSmall) to estimate percentage density and assessed performance by the correlation between estimated and measured percent density and a Bland-Altman plot. The automated measure was tested on an independent CSAW-CC dataset in which density had been measured using the LIBRA software, comparing measures for left and right breasts, sensitivity for high sensitivity, and areas under the receiver operating characteristic curve (AUCs). ResultsBased on the testing dataset, the correlation in percent density between the automated and human measures was 0.95, and the differences were only slightly larger for women with higher density. Based on the CSAW-CC dataset, AltoCumulus outperformed LIBRA in correlation between left and right breast (0.95 versus 0.79; P<0.001), specificity for 95% sensitivity (13% versus 10% (P<0.001)), and AUC (0.638 cf. 0.597; P<0.001). ConclusionWe have created an automated measure of mammographic density that is accurate and gives superior performance on repeatability within a woman, and for prediction of interval cancers, than another well-established automated measure.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Model uncertainty estimates for deep learning mammographic density prediction using ordinal and classification approaches 96%
- Mammographic density assessed using deep learning in women at high risk of developing breast cancer: the effect of weight change on density 95%
- Breast density prediction from low and standard dose mammograms using deep learning: effect of image resolution and model training approach on prediction quality 93%
Similar papers in this journal
- Automated and Manual Quantification of Tumour Cellularity in Digital Slides for Tumour Burden Assessment 95%
- Automated detection of the HER2 gene amplification status in Fluorescence in situ hybridization images for the diagnostics of cancer tissues 91%
- Reproducible And Clinically Translatable Deep Neural Networks For Cervical Screening 90%
Similar papers in this journal
- Improved accuracy of breast volume calculation from 3D surface imaging data using statistical shape models 94%
- Classification performance bias between training and test sets in a limited mammography dataset 92%
- BREAst screening Tailored for HEr (BREATHE) - A Study Protocol On Personalised Risk-based Breast Cancer Screening Programme 91%
Similar papers in this journal
- Assessing generalizability of an AI-based visual test for cervical cancer screening 91%
- Classification of Hyper-scale Multimodal Imaging Datasets 90%
- Development and Validation of a Deep Learning Model for Detecting Signs of Tuberculosis on Chest Radiographs among US-bound Immigrants and Refugees 89%
Similar papers in this journal
- SCU-Net: A deep learning method for segmentation and quantification of breast arterial calcifications on mammograms 94%
- Fully Automated Explainable Abdominal CT Contrast Media Phase Classification Using Organ Segmentation and Machine Learning 89%
- Selective ensemble methods for deep learning segmentation of major vessels in invasive coronary angiography 88%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.