Digital Breast Tomosynthesis: What Have We Learned?

A deep learning approach to estimate x-ray scatter in digital breast tomosynthesis: From phantom models to clinical applications

BACKGROUND

Digital breast tomosynthesis (DBT) has gained popularity as breast imaging modality due to its pseudo-3D reconstruction and improved accuracy compared to digital mammography. However, DBT faces challenges in image quality and quantitative accuracy due to scatter radiation. Recent advancements in deep learning (DL) have shown promise in using fast convolutional neural networks for scatter correction, achieving comparable results to Monte Carlo (MC) simulations.

PURPOSE

To predict the scatter radiation signal in DBT projections within clinically-acceptable times and using only clinically-available data, such as compressed breast thickness and acquisition angle.

METHODS

MC simulations to obtain scatter estimates were generated from two types of digital breast phantoms. One set consisted of 600 realistically-shaped homogeneous breast phantoms for initial DL training. The other set was composed of 80 anthropomorphic phantoms, containing realistic internal tissue texture, aimed at fine tuning the DL model for clinical applications. The MC simulations generated scatter and primary maps per projection angle for a wide-angle DBT system. Both datasets were used to train (using 7680 projections from homogeneous phantoms), validate (using 960 and 192 projections from the homogeneous and anthropomorphic phantoms, respectively), and test (using 960 and 48 projections from the homogeneous and anthropomorphic phantoms, respectively) the DL model. The DL output was compared to the corresponding MC ground truth using both quantitative and qualitative metrics, such as mean relative and mean absolute relative differences (MRD and MARD), and to previously-published scatter-to-primary (SPR) ratios for similar breast phantoms. The scatter corrected DBT reconstructions were evaluated by analyzing the obtained linear attenuation values and by visual assessment of corrected projections in a clinical dataset. The time required for training and prediction per projection, as well as the time it takes to produce scatter-corrected projection images, were also tracked.

RESULTS

The quantitative comparison between DL scatter predictions and MC simulations showed a median MRD of 0.05% (interquartile range (IQR), -0.04% to 0.13%) and a median MARD of 1.32% (IQR, 0.98% to 1.85%) for homogeneous phantom projections and a median MRD of -0.21% (IQR, -0.35% to -0.07%) and a median MARD of 1.43% (IQR, 1.32% to 1.66%) for the anthropomorphic phantoms. The SPRs for different breast thicknesses and at different projection angles were within ± 15% of the previously-published ranges. The visual assessment showed good prediction capabilities of the DL model with a close match between MC and DL scatter estimates, as well as between DL-based scatter corrected and anti-scatter grid corrected cases. The scatter correction improved the accuracy of the reconstructed linear attenuation of adipose tissue, reducing the error from -16% and -11% to -2.3% and 4.4% for an anthropomorphic digital phantom and clinical case with similar breast thickness, respectively. The DL model training took 40 min and prediction of a single projection took less than 0.01 s. Generating scatter corrected images took 0.03 s per projection for clinical exams and 0.16 s for one entire projection set.

CONCLUSIONS

This DL-based method for estimating the scatter signal in DBT projections is fast and accurate, paving the way for future quantitative applications.

Breast cancer screening in women with and without implants: retrospective study comparing digital mammography to digital mammography combined with digital breast tomosynthesis

OBJECTIVES

Compare four groups being screened: women without breast implants undergoing digital mammography (DM), women without breast implants undergoing DM with digital breast tomosynthesis (DM/DBT), women with implants undergoing DM, and women with implants undergoing DM/DBT.

METHODS

Mammograms from February 2011 to March 2017 were retrospectively reviewed after 13,201 were excluded for a unilateral implant or prior breast cancer. Patients had been allowed to choose between DM and DM/DBT screening. Mammography performance metrics were compared using chi-square tests.

RESULTS

Six thousand forty-one women with implants and 91,550 women without implants were included. In mammograms without implants, DM (n = 113,973) and DM/DBT (n = 61,896) yielded recall rates (RRs) of 8.53% and 6.79% (9726/113,973 and 4204/61,896, respectively, p < .001), cancer detection rates per 1000 exams (CDRs) of 3.96 and 5.12 (451/113,973 and 317/61,896, respectively, p = .003), and positive predictive values for recall (PPV1s) of 4.64% and 7.54% (451/9726 and 317/4204, respectively, p < .001), respectively. In mammograms with implants, DM (n = 6815) and DM/DBT (n = 5138) yielded RRs of 5.81% and 4.87% (396/6815 and 250/5138, respectively, p = .158), CDRs of 2.49 and 2.92 (17/6815 and 15/5138, respectively, p > 0.999), and PPV1s of 4.29% and 6.0% (17/396 and 15/250, respectively, p > 0.999), respectively.

CONCLUSIONS

DM/DBT significantly improved recall rates, cancer detection rates, and positive predictive values for recall compared to DM alone in women without implants. DM/DBT performance in women with implants trended towards similar improvements, though no metric was statistically significant.

KEY POINTS

• Digital mammography with tomosynthesis improved recall rates, cancer detection rates, and positive predictive values for recall compared to digital mammography alone for women without implants. • Digital mammography with tomosynthesis trended towards improving recall rates, cancer detection rates, and positive predictive values for recall compared to digital mammography alone for women with implants, but these trends were not statistically significant - likely related to sample size.

Positive Predictive Value of Tomosynthesis-guided Biopsies of Architectural Distortions Seen on Digital Breast Tomosynthesis and without an Ultrasound Correlate

Objective

The objective of the study was to determine the positive predictive value (PPV) of architectural distortions (AD) observed on digital breast tomosynthesis (DBT) and without an ultrasound (US) correlate.

Materials and Methods

In this single-institution, retrospective study, patients who underwent DBT-guided biopsies of AD without any associated findings on digital mammography (DM) or DBT, and without a correlate on targeted US exam, over a 14-month period were included in this study. All patients had DM and DBT and targeted US exams. The PPV was computed along with the exact 95% confidence limits (CL) using simple binomial proportions, with histopathology as the reference standard.

Results

A total of 45 ADs in 45 patients met the inclusion criteria. Histopathology indicated 6/45 (PPV: 13.3%, CL: 5.1-26.8%), ADs were malignant, including one high-risk lesion that was upgraded at surgery. ADs were appreciated only on DBT in 12/45 (26.7%) patients, and on both DBT and DM in 33/45 (73.3%) patients, and the corresponding PPV was 25% (3/12, CL: 5.5-57.2%) and 9.1% (3/33, CL: 1.9-24.3%), respectively. In all analyses, the observed PPV significantly exceeded the 2% probability of malignancy for Breast Imaging Reporting and Data System-3 diagnostic categories (P < 0.004).

Conclusions

The PPV of malignancy in DBT detected AD without an US correlate in our series of 45 cases was 6/45 (13.3%). In the absence of an US correlate, the PPV of AD is lower than that mentioned in prior literature but exceeds the 2% threshold to justify DBT-guided biopsy.