Tumor-Attentive Segmentation-Guided GAN for Synthesizing Breast Contrast-Enhanced MRI Without Contrast Agents

Feasibility of virtual T2-weighted fat-saturated breast MRI images by convolutional neural networks

BACKGROUND

Breast magnetic resonance imaging (MRI) protocols often include T2-weighted fat-saturated (T2w-FS) sequences, which support tissue characterization but significantly increase scan time. This study aims to evaluate whether a 2D-U-Net neural network can generate virtual T2w-FS (VirtuT2w) images from routine multiparametric breast MRI images.

METHODS

This IRB-approved, retrospective study included 914 breast MRI examinations from January 2017 to June 2020. The dataset was divided into training (n = 665), validation (n = 74), and test sets (n = 175). The U-Net was trained using different input protocols consisting of T1-weighted, diffusion-weighted, and dynamic contrast-enhanced sequences to generate VirtuT2. Quantitative metrics were used to evaluate the different input protocols. A qualitative assessment by two radiologists was used to evaluate the VirtuT2w images of the best input protocol.

RESULTS

VirtuT2w images demonstrated the best quantitative metrics compared to original T2w-FS images for an input protocol using all of the available data. A high level of high-frequency error norm (0.87) indicated a strong blurring presence in the VirtuT2 images, which was also confirmed by qualitative reading. Radiologists correctly identified VirtuT2 images with at least 96% accuracy. Significant difference in diagnostic image quality was noted for both readers (p ≤ 0.015). Moderate inter-reader agreement was observed for edema detection on both T2w-FS images (κ = 0.49) and VirtuT2 images (κ = 0.44).

CONCLUSION

The 2D-U-Net generated virtual T2w-FS images similar to real T2w-FS images, though blurring remains a limitation. Investigation of other architectures and using larger datasets is necessary to improve potential future clinical applicability.

RELEVANCE STATEMENT

Generating VirtuT2 images could potentially decrease the examination time of multiparametric breast MRI, but its quality needs to improve before introduction into a clinical setting.

KEY POINTS

Breast MRI T2w-fat-saturated (FS) images can be virtually generated using convolutional neural networks. Image blurring in virtual T2w-FS images currently limits their clinical applicability. Best quantitative performance could be achieved when using full dynamic-contrast-enhanced acquisition and DWI as input of the neural network.

Deep learning empowered gadolinium-free contrast-enhanced abbreviated MRI for diagnosing hepatocellular carcinoma

Background & Aims

By reducing some magnetic resonance imaging (MRI) sequences, abbreviated MRI (aMRI) has shown extensive promise for detecting hepatocellular carcinoma (HCC). We aim to develop deep learning (DL)-based gadolinium-free contrast-enhanced (CE) aMRI protocols (DL-aMRI) for detecting HCC.

Methods

In total, 1,769 patients (913 with HCC) were retrospectively included from three institutions for training, testing, and external validation. Stable diffusion-based DL models were trained to generate CE-MRI, including T1-weighted arterial, portal venous, transitional, and hepatobiliary phase images (AP-syn, VP-syn, TP-syn, and HBP-syn, respectively). Non-contrast-MRI (NC-MRI), including T2-weighted, diffusion-weighted, and pre-contrast T1-weighted (Pre) sequences, along with either actual or DL-synthesized CE-MRI (AP, VP, TP, and HBP or AP-syn, VP-syn, TP-syn, and HBP-syn), were used to create conventional complete MRI (cMRI) and DL-aMRI protocols. An inter-method comparison of image quality between DL-aMRI and cMRI was conducted using a non-inferiority test. The sensitivity and specificity of DL-aMRI and cMRI for detecting HCC were statistically compared using the non-inferiority test and generalized estimating equations models.

Results

DL-aMRI showed a remarkable reduction in acquisition time compared with cMRI (4.1 vs. 28.1 min). The image quality of DL-synthesized CE-MRI was not inferior to that of actual CE-MRI (p <0.001). There was an excellent inter-method agreement between the HCC sizes measured by the two protocols (R2 = 0.9436-0.9683). The pooled sensitivity and specificity of cMRI and DL-aMRI were 0.899 and 0.925 and 0.866 and 0.922, respectively. No significant differences were found between the sensitivity and specificity of the two protocols.

Conclusions

The proposed DL-aMRI could facilitate precise HCC diagnosis with no need for contrast agents, a substantial reduction in acquisition time, and preservation of both NC-MRI and CE-MRI data. DL-aMRI may serve as a valuable tool for HCC diagnosing.

Impact and implications

In this multi-center study involving 1,769 participants, we developed a generative deep learning-based abbreviated MRI (DL-aMRI) strategy that provides an efficient, contrast-agent-free alternative for detecting HCC with accuracy comparable to that of conventional complete MRI, significantly reducing acquisition time from 28.1 min to just 4.1 min. This strategy is valuable for clinicians who face significant workloads resulting from long MRI scanning times and the potential adverse effects of contrast agents, as well as for researchers focused on developing cost-effective and accessible diagnostic tools for HCC detection. The proposed DL-aMRI protocol has practical implications for clinical settings, enhancing diagnostic efficiency while maintaining high image quality, eliminating the need for contrast agents and ultimately benefiting patients and healthcare providers.

Impact of non-contrast-enhanced imaging input sequences on the generation of virtual contrast-enhanced breast MRI scans using neural network

OBJECTIVE

To investigate how different combinations of T1-weighted (T1w), T2-weighted (T2w), and diffusion-weighted imaging (DWI) impact the performance of virtual contrast-enhanced (vCE) breast MRI.

MATERIALS AND METHODS

The IRB-approved, retrospective study included 1064 multiparametric breast MRI scans (age: 52 ± 12 years) obtained from 2017 to 2020 (single site, two 3-T MRI). Eleven independent neural networks were trained to derive vCE images from varying input combinations of T1w, T2w, and multi-b-value DWI sequences (b-value = 50-1500 s/mm2). Three readers evaluated the vCE images with regard to qualitative scores of diagnostic image quality, image sharpness, satisfaction with contrast/signal-to-noise ratio, and lesion/non-mass enhancement conspicuity. Quantitative metrics (SSIM, PSNR, NRMSE, and median symmetrical accuracy) were analyzed and statistically compared between the input combinations for the full breast volume and both enhancing and non-enhancing target findings.

RESULTS

The independent test set consisted of 187 cases. The quantitative metrics significantly improved in target findings when multi-b-value DWI sequences were included during vCE training (p < 0.05). Non-significant effects (p > 0.05) were observed for the quantitative metrics on the full breast volume when comparing input combinations including T1w. Using T1w and DWI acquisitions during vCE training is necessary to achieve high satisfaction with contrast/SNR and good conspicuity of the enhancing findings. The input combination of T1w, T2w, and DWI sequences with three b-values showed the best qualitative performance.

CONCLUSION

vCE breast MRI performance is significantly influenced by input sequences. Quantitative metrics and visual quality of vCE images significantly benefit when multi b-value DWI is added to morphologic T1w-/T2w sequences as input for model training.

KEY POINTS

Question How do different MRI sequences impact the performance of virtual contrast-enhanced (vCE) breast MRI? Findings The input combination of T1-weighted, T2-weighted, and diffusion-weighted imaging sequences with three b-values showed the best qualitative performance. Clinical relevance While in the future neural networks providing virtual contrast-enhanced images might further improve accessibility to breast MRI, the significant influence of input data needs to be considered during translational research.

BentRay-NeRF: Bent-Ray Neural Radiance Fields for Robust Speed-of-Sound Imaging in Ultrasound Computed Tomography

Ultrasound computed tomography (USCT) is a promising technique for breast cancer detection because of its quantitative imaging capability of the speed of sound (SOS) of soft tissues and the fact that malignant breast lesions often have a higher SOS compared to healthy tissues in the human breast. Compared to waveform inversion-based USCT, bent-ray tracing USCT is relatively less computationally expensive, which particularly suits breast cancer screening in a large population. However, SOS image reconstruction using bent-ray tracing in USCT is a highly ill-conditioned problem, making it susceptible to measurement errors. This presents significant challenges in achieving stable and high-quality reconstructions. In this study, we show that using implicit neural representation (INR), the ill-conditioned problem can be well mitigated, and stable reconstruction is achievable. This INR approach uses a multilayer perceptron (MLP) with hash encoding to model the slowness map as a continuous function, to better regularize the inverse problem and has been shown more effective than classical approaches of solely adding regularization terms in the loss function. Thereby, we propose the bent-ray neural radiance fields (BentRay-NeRF) method for SOS image reconstruction to address the aforementioned challenges in classical SOS image reconstruction methods, such as the Gauss-Newton method. In silico and in vitro experiments showed that BentRay-NeRF has remarkably improved performance compared to the classical method in many aspects, including the image quality and the robustness of the inversion to different acquisition settings in the presence of measurement errors.

Evaluating Virtual Contrast-Enhanced Magnetic Resonance Imaging in Nasopharyngeal Carcinoma Radiation Therapy: A Retrospective Analysis for Primary Gross Tumor Delineation

PURPOSE

To investigate the potential of virtual contrast-enhanced magnetic resonance imaging (VCE-MRI) for gross-tumor-volume (GTV) delineation of nasopharyngeal carcinoma (NPC) using multi-institutional data.

METHODS AND MATERIALS

This study retrospectively retrieved T1-weighted (T1w), T2-weighted (T2w) MRI, gadolinium-based contrast-enhanced MRI (CE-MRI), and planning computed tomography (CT) of 348 biopsy-proven NPC patients from 3 oncology centers. A multimodality-guided synergistic neural network (MMgSN-Net) was trained using 288 patients to leverage complementary features in T1w and T2w MRI for VCE-MRI synthesis, which was independently evaluated using 60 patients. Three board-certified radiation oncologists and 2 medical physicists participated in clinical evaluations in 3 aspects: image quality assessment of the synthetic VCE-MRI, VCE-MRI in assisting target volume delineation, and effectiveness of VCE-MRI-based contours in treatment planning. The image quality assessment includes distinguishability between VCE-MRI and CE-MRI, clarity of tumor-to-normal tissue interface, and veracity of contrast enhancement in tumor invasion risk areas. Primary tumor delineation and treatment planning were manually performed by radiation oncologists and medical physicists, respectively.

RESULTS

The mean accuracy to distinguish VCE-MRI from CE-MRI was 31.67%; no significant difference was observed in the clarity of tumor-to-normal tissue interface between VCE-MRI and CE-MRI; for the veracity of contrast enhancement in tumor invasion risk areas, an accuracy of 85.8% was obtained. The image quality assessment results suggest that the image quality of VCE-MRI is highly similar to real CE-MRI. The mean dosimetric difference of planning target volumes was less than 1 Gy.

CONCLUSIONS

The VCE-MRI is highly promising to replace the use of gadolinium-based CE-MRI in tumor delineation of NPC patients.

AI Applications to Breast MRI: Today and Tomorrow

In breast imaging, there is an unrelenting increase in the demand for breast imaging services, partly explained by continuous expanding imaging indications in breast diagnosis and treatment. As the human workforce providing these services is not growing at the same rate, the implementation of artificial intelligence (AI) in breast imaging has gained significant momentum to maximize workflow efficiency and increase productivity while concurrently improving diagnostic accuracy and patient outcomes. Thus far, the implementation of AI in breast imaging is at the most advanced stage with mammography and digital breast tomosynthesis techniques, followed by ultrasound, whereas the implementation of AI in breast magnetic resonance imaging (MRI) is not moving along as rapidly due to the complexity of MRI examinations and fewer available dataset. Nevertheless, there is persisting interest in AI-enhanced breast MRI applications, even as the use of and indications of breast MRI continue to expand. This review presents an overview of the basic concepts of AI imaging analysis and subsequently reviews the use cases for AI-enhanced MRI interpretation, that is, breast MRI triaging and lesion detection, lesion classification, prediction of treatment response, risk assessment, and image quality. Finally, it provides an outlook on the barriers and facilitators for the adoption of AI in breast MRI. LEVEL OF EVIDENCE: 5 TECHNICAL EFFICACY: Stage 6.

Reducing Gadolinium Contrast With Artificial Intelligence

Gadolinium contrast is an important agent in magnetic resonance imaging (MRI), particularly in neuroimaging where it can help identify blood-brain barrier breakdown from an inflammatory, infectious, or neoplastic process. However, gadolinium contrast has several drawbacks, including nephrogenic systemic fibrosis, gadolinium deposition in the brain and bones, and allergic-like reactions. As computer hardware and technology continues to evolve, machine learning has become a possible solution for eliminating or reducing the dose of gadolinium contrast. This review summarizes the clinical uses of gadolinium contrast, the risks of gadolinium contrast, and state-of-the-art machine learning methods that have been applied to reduce or eliminate gadolinium contrast administration, as well as their current limitations, with a focus on neuroimaging applications. EVIDENCE LEVEL: 3 TECHNICAL EFFICACY: Stage 1.