Deep learning based tomosynthesis denoising: a bias investigation across different breast types.

Purpose: High noise levels due to low X-ray dose are a challenge in digital breast tomosynthesis (DBT) reconstruction. Deep learning algorithms show promise in reducing this noise. However, these algorithms can be complex and biased toward certain patient groups if the training data are not representative. It is important to thoroughly evaluate deep learning-based denoising algorithms before they are applied in the medical field to ensure their effectiveness and fairness. In this work, we present a deep learning-based denoising algorithm and examine potential biases with respect to breast density, thickness, and noise level. Approach: We use physics-driven data augmentation to generate low-dose images from full field digital mammography and train an encoder-decoder network. The rectified linear unit (ReLU)-loss, specifically designed for mammographic denoising, is utilized as the objective function. To evaluate our algorithm for potential biases, we tested it on both clinical and simulated data generated with the virtual imaging clinical trial for regulatory evaluation pipeline. Simulated data allowed us to generate X-ray dose distributions not present in clinical data, enabling us to separate the influence of breast types and X-ray dose on the denoising performance. Results: Our results show that the denoising performance is proportional to the noise level. We found a bias toward certain breast groups on simulated data; however, on clinical data, our algorithm denoises different breast types equally well with respect to structural similarity index. Conclusions: We propose a robust deep learning-based denoising algorithm that reduces DBT projection noise levels and subject it to an extensive test that provides information about its strengths and weaknesses.

Impact of Artificial Intelligence Decision Support Using Deep Learning on Breast Cancer Screening Interpretation with Single-View Wide-Angle Digital Breast Tomosynthesis.

Background The high volume of data in digital breast tomosynthesis (DBT) and the lack of agreement on how to best implement it in screening programs makes its use challenging. Purpose To compare radiologist performance when reading single-view wide-angle DBT images with and without an artificial intelligence (AI) system for decision and navigation support. Materials and Methods A retrospective observer study was performed with bilateral mediolateral oblique examinations and corresponding synthetic two-dimensional images acquired between June 2016 and February 2018 with a wide-angle DBT system. Fourteen breast screening radiologists interpreted 190 DBT examinations (90 normal, 26 with benign findings, and 74 with malignant findings), with the reference standard being verified by using histopathologic analysis or at least 1 year of follow-up. Reading was performed in two sessions, separated by at least 4 weeks, with a random mix of examinations being read with and without AI decision and navigation support. Forced Breast Imaging Reporting and Data System (categories 1-5) and level of suspicion (1-100) scores were given per breast by each reader. The area under the receiver operating characteristic curve (AUC) and the sensitivity and specificity were compared between conditions by using the public-domain iMRMC software. The average reading times were compared by using the Wilcoxon signed rank test. Results The 190 women had a median age of 54 years (range, 48-63 years). The examination-based reader-averaged AUC was higher when interpreting results with AI support than when reading unaided (0.88 [95% CI: 0.84, 0.92] vs 0.85 [95% CI: 0.80, 0.89], respectively; P = .01). The average sensitivity increased with AI support (64 of 74, 86% [95% CI: 80%, 92%] vs 60 of 74, 81% [95% CI: 74%, 88%]; P = .006), whereas no differences in the specificity (85 of 116, 73.3% [95% CI: 65%, 81%] vs 83 of 116, 71.6% [95% CI: 65%, 78%]; P = .48) or reading time (48 seconds vs 45 seconds; P = .35) were detected. Conclusion Using a single-view digital breast tomosynthesis (DBT) and artificial intelligence setup could allow for a more effective screening program with higher performance, especially in terms of an increase in cancers detected, than using single-view DBT alone. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Chan and Helvie in this issue.

High-attenuation artifact reduction in breast tomosynthesis using a novel reconstruction algorithm.

PURPOSE: To assess the effect on reducing the out-of-plane artifacts from metal objects in breast tomosynthesis (BT) using a novel artifact-reducing reconstruction algorithm in specimen radiography. METHODS AND MATERIALS: The study was approved by the Regional Ethical Review Board. BT images of 18 partial- and whole mastectomy specimens from women with breast cancer were acquired before and after a needle was inserted close to the lesion. The images were reconstructed using both a standard reconstruction algorithm, and a novel algorithm; the latter uses pre-segmentation to remove highly attenuating artifact-inducing objects from projection images before reconstruction. Images were separately reconstructed with and without segmentation, and combined into an artifact-reduced reconstruction. Standard and artifact-reduced BT-algorithms were compared visually and quantitatively using clinical images of mastectomy specimens and a physical anthropomorphic phantom. Six readers independently assessed the visibility of the lesion with and without artifact-reduction in a side-by-side comparison. A quantitative analysis was performed, comparing the signal-difference to background ratio (SDBR) and artifact spread function (ASF) between the two reconstruction methods. RESULTS: The magnitude of out-of-plane artifacts was clearly reduced with the novel reconstruction compared to BT-images without artifact reduction. Lesion masking by artifacts was largely averted; tumour visibility was comparable to standard BT images without a needle. In 76 ± 8% (standard deviation) of cases overall, readers could confidently state needle location. The same figure was 94 ± 6% for whole mastectomy cases, compared to 62 ± 17% for partial mastectomies. With metal artifact reduction, SDBR increased by 97% in the phantom, and by 69% in the mastectomies. The artifact spread function was substantially narrower. CONCLUSION: Artifact reduction in BT using a novel reconstruction method enables qualitatively and quantitatively improved clinical use of BT when metal artifacts can be a limiting factor such as in tomosynthesis-guided biopsy.

Comparison of contrast-enhanced digital mammography and contrast-enhanced digital breast tomosynthesis for lesion assessment.

Contrast-enhanced digital mammography (CEDM) reveals neovasculature of breast lesions in a two-dimensional contrast enhancement map. Contrast-enhanced digital breast tomosynthesis (CEDBT) provides contrast enhancement in three dimensions, which may improve lesion characterization and localization. We aim to compare CEDM and CEDBT for lesion assessment. Women with breast imaging-reporting and data system 4 or 5 suspicious breast lesion(s) were recruited in our study and were imaged with CEDM and CEDBT in succession under one breast compression. Two radiologists assessed CEDM and CEDBT with both images displayed side-by-side and compared (1) contrast enhancement of lesions and (2) lesion margin using a five-point scale ranging from - 2 (CEDM much better) to + 2 (CEDBT much better). Biopsy identified 19 malignant lesions with contrast enhancement. Our results show that CEDBT provides better lesion margins than CEDM with limited reduction in contrast enhancement. CEDBT delivers less radiation dose compared to CEDM + DBT. Synthetic CEDM can be generated from CEDBT data and provides lesion contrast enhancement comparable to CEDM. CEDBT has potential for clinical applications, such as treatment response monitoring and guidance for biopsy.

Multiple Angulated Mammography Reconstructions in Digital Breast Tomosynthesis for the Diagnosis of Microcalcifications - Added Value to Standard Stack Reconstructions and Synthesized Mammography.

PURPOSE: To compare ratings regarding the depiction, diagnostic accuracy and lesion characterization of conventional synthesized mammography (SM), multiple angulated mammography reconstructions (INSIGHT3D), and standard stack reconstructions in digital breast tomosynthesis for microcalcifications. MATERIALS AND METHODS: This is a retrospective, multicase, multireader study. We included patients with digital breast tomosynthesis (DBT), microcalcifications and histology over a period of four months in our institution and the same number of normal cases. Three radiologists, who were blinded to patient data, independently rated the depiction, distribution, morphology and BI-RADS score of microcalcifications in SM, INSIGHT3D and standard stack reconstructions. Deidentified images were presented in random order. Reading time was measured. Friedman and post hoc Nemenyi tests, Cochrane's Q and post hoc Wilcoxon signed rank tests, Fleiss' kappa and receiver operating characteristics were used for statistical analysis. RESULTS: We included 41 histopathologically proven and 41 normal cases. Depiction of microcalcifications was rated better in INSIGHT3D than in SM and better in stack reconstructions than in INSIGHT3D and SM (P < 0.001). The reading time was lower in SM and INSIGHT3D compared to stack reconstructions (P < 0.001). The diagnostic accuracy and inter-rater correlation were comparable between all tested modes of reconstruction. CONCLUSIONS: INSIGHT3D has higher ratings regarding the depiction of microcalcifications compared to SM while maintaining a short reading time. Our preliminary assessment suggests that INSIGHT3D provides added value to SM. KEY POINTS: · INSIGHT3D depicts microcalcifications better than synthesized mammography while maintaining a low reading time.. · The diagnostic accuracy and inter-rater correlation were comparable between INSIGHT3D and synthesized mammography.. · INSIGHT3D may be a potential successor to synthesized mammography.. CITATION FORMAT: · Neubauer J, Neubauer C, Wicklein J et al. Multiple Angulated Mammography Reconstructions in Digital Breast Tomosynthesis for the Diagnosis of Microcalcifications - Added Value to Standard Stack Reconstructions and Synthesized Mammography. Fortschr Röntgenstr 2018; DOI: 10.1055/s-0044-100726.

Image features for misalignment correction in medical flat-detector CT.

PURPOSE: Misalignment artifacts are a serious problem in medical flat-detector computed tomography. Generally, the geometrical parameters, which are essential for reconstruction, are provided by preceding calibration routines. These procedures are time consuming and the later use of stored parameters is sensitive toward external impacts or patient movement. The method of choice in a clinical environment would be a markerless online-calibration procedure that allows flexible scan trajectories and simultaneously corrects misalignment and motion artifacts during the reconstruction process. Therefore, different image features were evaluated according to their capability of quantifying misalignment. METHODS: Projections of the FORBILD head and thorax phantoms were simulated. Additionally, acquisitions of a head phantom and patient data were used for evaluation. For the reconstruction different sources and magnitudes of misalignment were introduced in the geometry description. The resulting volumes were analyzed by entropy (based on the gray-level histogram), total variation, Gabor filter texture features, Haralick co-occurrence features, and Tamura texture features. The feature results were compared to the back-projection mismatch of the disturbed geometry. RESULTS: The evaluations demonstrate the ability of several well-established image features to classify misalignment. The authors elaborated the particular suitability of the gray-level histogram-based entropy on identifying misalignment artifacts, after applying an appropriate window level (bone window). CONCLUSIONS: Some of the proposed feature extraction algorithms show a strong correlation with the misalignment level. Especially, entropy-based methods showed very good correspondence, with the best of these being the type that uses the gray-level histogram for calculation. This makes it a suitable image feature for online-calibration.