Accurate tumor segmentation is required for quantitative image analyses, which are increasingly used for evaluation of tumors. We developed a fully automated and high-performance segmentation model ...of triple-negative breast cancer using a self-configurable deep learning framework and a large set of dynamic contrast-enhanced MRI images acquired serially over the patients' treatment course. Among all models, the top-performing one that was trained with the images across different time points of a treatment course yielded a Dice similarity coefficient of 93% and a sensitivity of 96% on baseline images. The top-performing model also produced accurate tumor size measurements, which is valuable for practical clinical applications.
A betavoltaic battery was prepared using radioactive 63Ni attached to a three-dimensional single trenched P–N absorber. The optimum thickness of a 63Ni layer was determined to be approximately 2 μm, ...considering the minimum self-shielding effect of beta particles. Electroplating of radioactive 63Ni on a nickel (Ni) foil was carried out at a current density of 20 mA/cm2. The difference of the short-circuit currents (Isc) between the pre- and postdeposition of 63Ni (16.65 MBq) on the P–N junction was 5.03 nA, as obtained from the I–V characteristics. An improved design with a sandwich structure was provided for enhancing performance.
•A very low rate of 3 ± 3 FPs per patient was maintained for all brain metastasis sizes.•Automated segmentation performance for brain metastases depends on metastasis size.•For metastases ≥6 mm, DSC ...was 87%, sensitivity was 99%, and PPV was 67%.•For metastases ≥3 mm and <6 mm, DSC was 64%, sensitivity was 87%, and PPV was 49%.•For metastases <3 mm, DSC was 17%, sensitivity was 25%, and PPV was 28%.
Brain metastases are manually contoured during stereotactic radiosurgery (SRS) treatment planning, which is time-consuming, potentially challenging, and laborious. The purpose of this study was to develop and investigate a 2-stage deep learning (DL) approach (MetNet) for brain metastasis segmentation in pre-treatment magnetic resonance imaging (MRI).
We retrospectively analyzed postcontrast 3D T1-weighted spoiled gradient echo MRIs from 934 patients who underwent SRS between August 2009 and August 2018. Neuroradiologists manually identified brain metastases in the MRIs. The treating radiation oncologist or physicist contoured the brain metastases. We constructed a 2-stage DL ensemble consisting of detection and segmentation models to segment the brain metastases on the MRIs. We evaluated the performance of MetNet by computing sensitivity, positive predictive value (PPV), and Dice similarity coefficient (DSC) with respect to metastasis size, as well as free-response receiver operating characteristics.
The 934 patients (mean ±standard deviation age 59 ± 13 years, 474 women) were randomly split into 80% training and 20% testing groups (748:186). For patients with metastases 1–52 mm (n = 766), 648 (85%) were detected and segmented with a mean segmentation DSC of 81% ± 15%. Patient-averaged sensitivity was 88% ± 19%, PPV was 58% ± 25%, and DSC was 85% ± 13% with 3 ± 3 false positives (FPs) per patient. When considering only metastases ≥6 mm, patient-averaged sensitivity was 99% ± 5%, PPV was 67% ± 28%, and DSC was 87% ± 13% with 1 ± 2 FPs per patient.
MetNet can segment brain metastases across a broad range of metastasis sizes with high sensitivity, low FPs, and high segmentation accuracy in postcontrast T1-weighted MRI, potentially aiding treatment planning for SRS.
Evaluate deep learning (DL) to improve the image quality of the PROPELLER (Periodically Rotated Overlapping Parallel Lines with Enhanced Reconstruction technique) for 3 T magnetic resonance imaging ...of the female pelvis.
Three radiologists prospectively and independently compared non-DL and DL PROPELLER sequences from 20 patients with a history of gynecologic malignancy. Sequences with different noise reduction factors (DL 25%, DL 50%, and DL 75%) were blindly reviewed and scored based on artifacts, noise, relative sharpness, and overall image quality. The generalized estimating equation method was used to assess the effect of methods on the Likert scales. Quantitatively, the contrast-to-noise ratio and signal-to-noise ratio (SNR) of the iliac muscle were calculated, and pairwise comparisons were performed based on a linear mixed model. P values were adjusted using the Dunnett method. Interobserver agreement was assessed using the κ statistic. P value was considered statistically significant at less than 0.05.
Qualitatively, DL 50 and DL 75 were ranked as the best sequences in 86% of cases. Images generated by the DL method were significantly better than non-DL images (P < 0.0001). Iliacus muscle SNR on DL 50 and DL 75 was significantly better than non-DL images (P < 0.0001). There was no difference in contrast-to-noise ratio between the DL and non-DL techniques in the iliac muscle. There was a high percent agreement (97.1%) in terms of DL sequences' superior image quality (97.1%) and sharpness (100%) relative to non-DL images.
The utilization of DL reconstruction improves the image quality of PROPELLER sequences with improved SNR quantitatively.
The purpose of this study was to assess the feasibility of a short protocol for screening breast MRI that is noninferior to standard-of-care (SOC) MRI in image quality that complies with American ...College of Radiology accreditation requirements.
In a prospective feasibility trial, 23 women at high risk underwent both an initial SOC MRI examination that included axial iterative decomposition of water and fat with echo asymmetry and least-squares estimation (IDEAL) and T1-weighted volume imaging for breast assessment (VIBRANT) dynamic contrast-enhanced sequences and a separate short breast MRI protocol comprising a fast spin-echo (FSE) triple-echo Dixon T2 sequence for T2-weighted imaging and a 3D dual-echo fast spoiled gradient-echo two-point Dixon sequence for dynamic contrast-enhanced imaging from October 1, 2015, through May 2, 2016. Image quality assessment was performed by three radiologists, who scored the images for fat saturation, artifact severity, and quality of normal anatomic structures. Enhancing lesions were evaluated according to BI-RADS MRI features. Quantitative analysis was performed by measuring the signal intensity of anatomic areas in each patient.
The mean acquisition time for short-protocol breast MRI was 9.42 minutes and for SOC MRI was 22.09 minutes (p < 0.0001). The mean table times were 13.92 and 35.87 minutes (p < 0.0001). Compared with the FSE triple-echo Dixon T2 short-protocol breast MRI sequence, the IDEAL SOC MRI sequence had significantly worse motion artifact (p < 0.01) and fat saturation (p = 0.04). The other parameters did not differ significantly. Quantitative analysis showed that the FSE triple-echo Dixon T2 sequence had more effective fat saturation and higher tissue contrast. All five lesions were given the same assessments by the readers, and at BI-RADS lesion morphologic ranking, identical high image quality scores were assigned to both the VIBRANT and 3D dual-echo fast spoiled gradient-echo 2-point Dixon sequences.
Short-protocol breast MRI comprising a T2-weighted sequence and a fast dynamic sequence with less than 10-minute acquisition time is feasible and has image quality at least equivalent to that of an SOC MRI protocol with a > 20-minute mean acquisition time. Larger studies comparing the cancer detection rate, sensitivity, and specificity of each imaging protocol are needed to determine whether short-protocol breast MRI can replace SOC MRI to screen patients at high breast cancer risk.