Transfer learning for the generalization of artificial intelligence in breast cancer detection: a case-control study.

BACKGROUND: Some researchers have questioned whether artificial intelligence (AI) systems maintain their performance when used for women from populations not considered during the development of the system. PURPOSE: To evaluate the impact of transfer learning as a way of improving the generalization of AI systems in the detection of breast cancer. MATERIAL AND METHODS: This retrospective case-control Finnish study involved 191 women diagnosed with breast cancer and 191 matched healthy controls. We selected a state-of-the-art AI system for breast cancer detection trained using a large US dataset. The selected baseline system was evaluated in two experimental settings. First, we examined our private Finnish sample as an independent test set that had not been considered in the development of the system (unseen population). Second, the baseline system was retrained to attempt to improve its performance in the unseen population by means of transfer learning. To analyze performance, we used areas under the receiver operating characteristic curve (AUCs) with DeLong's test. RESULTS: Two versions of the baseline system were considered: ImageOnly and Heatmaps. The ImageOnly and Heatmaps versions yielded mean AUC values of 0.82±0.008 and 0.88±0.003 in the US dataset and 0.56 (95% CI=0.50-0.62) and 0.72 (95% CI=0.67-0.77) when evaluated in the unseen population, respectively. The retrained systems achieved AUC values of 0.61 (95% CI=0.55-0.66) and 0.69 (95% CI=0.64-0.75), respectively. There was no statistical difference between the baseline system and the retrained system. CONCLUSION: Transfer learning with a small study sample did not yield a significant improvement in the generalization of the system.

Saliency of breast lesions in breast cancer detection using artificial intelligence.

The analysis of mammograms using artificial intelligence (AI) has shown great potential for assisting breast cancer screening. We use saliency maps to study the role of breast lesions in the decision-making process of AI systems for breast cancer detection in screening mammograms. We retrospectively collected mammograms from 191 women with screen-detected breast cancer and 191 healthy controls matched by age and mammographic system. Two radiologists manually segmented the breast lesions in the mammograms from CC and MLO views. We estimated the detection performance of four deep learning-based AI systems using the area under the ROC curve (AUC) with a 95% confidence interval (CI). We used automatic thresholding on saliency maps from the AI systems to identify the areas of interest on the mammograms. Finally, we measured the overlap between these areas of interest and the segmented breast lesions using Dice's similarity coefficient (DSC). The detection performance of the AI systems ranged from low to moderate (AUCs from 0.525 to 0.694). The overlap between the areas of interest and the breast lesions was low for all the studied methods (median DSC from 4.2% to 38.0%). The AI system with the highest cancer detection performance (AUC = 0.694, CI 0.662-0.726) showed the lowest overlap (DSC = 4.2%) with breast lesions. The areas of interest found by saliency analysis of the AI systems showed poor overlap with breast lesions. These results suggest that AI systems with the highest performance do not solely rely on localized breast lesions for their decision-making in cancer detection; rather, they incorporate information from large image regions. This work contributes to the understanding of the role of breast lesions in cancer detection using AI.

A New Benchmark and Method for the Evaluation of Chest Wall Detection in Digital Mammography.

CAD systems have shown good potential for improving breast cancer diagnosis and anomaly detection in mammograms. A basic enabling step for the utilization of CAD systems in mammographic analysis is the correct identification of the breast region. Therefore, several methods to segment the pectoral muscle in the medio-lateral oblique (MLO) mammographic view have been proposed in the literature. However, currently it is difficult to perform and objective comparison between different chest wall (CW) detection methods since they are often evaluated with different evaluation procedures, datasets and the implementations of the methods are not publicly available. For this reason, we propose a methodology to evaluate and compare the performance of CW detection methods using a publicly available dataset (INbreast). We also propose a new intensity-based method for automatic CW detection. We then utilize the proposed evaluation methodology to compare the performance of our CW detection algorithm with a state-of-the-art CW detection method. The performance was measured in terms of the Dice's coefficient similarity, the area error and mean contour distance. The proposed method achieves yielded the best results in all the performance measures.

A Comparison of Regions of Interest in Parenchymal Analysis for Breast Cancer Risk Assessment.

Computerized parenchymal analysis has shown potential to be utilized as an imaging biomarker to estimate the risk of breast cancer. Parenchymal analysis of digital mammograms is based on the extraction of computerized measures to build machine learning-based models for the prediction of breast cancer risk. However, the choice of the region of interest (ROI) for feature extraction within the breast remains an open problem. In this work we perform a comparison between five different methods suggested in the literature for automated ROI selection, including the whole breast (WB), the maximum squared (MS), the retro-areolar region (RA), the lattice-based (LB), and the polar-based (PB) selection methods. For the experiments, we built a retrospective dataset of 896 screening mammograms from 224 women (112 cases and 112 healthy controls). The performance of each ROI selection method was measured in terms of the area under the curve (AUC) values. The AUC values varied between 0.55 and 0.79 depending on the method and experimental settings. The best performance on an independent test set was achieved by the MS method (AUC of 0.59, 95% CI: 0.55-0.64). This method is fully-automated and does not require adjusting hyper-parameters. Based on our results, we prompt the use of the MS method for ROI selection in the computerized parenchymal analysis for breast cancer risk assessment.