No abstract

No abstract

Detection of brain metastases is a paramount task in cancer management due both to the number of high-risk patients and the difficulty of achieving consistent detection. In this study, we aim to improve the accuracy of automated brain metastasis (BM) detection methods using a novel asymmetric UNet (asym-UNet) architecture. An end-to-end asymmetric 3D-UNet architecture, with two down-sampling arms and one up-sampling arm, was constructed to capture the imaging features. The two down-sampling arms were trained using two different kernels (3 × 3 × 3 and 1 × 1 × 3, respectively) with the kernel (1 × 1 × 3) dominating the learning. As a comparison, vanilla single 3D UNets were trained with different kernels and evaluated using the same datasets. Voxel-based Dice similarity coefficient (DSCv), sensitivity (S v), precision (P v), BM-based sensitivity (S BM), and false detection rate (F BM) were used to evaluate model performance. Contrast-enhanced T1 MR images from 195 patients with a total of 1034 BMs were solicited from our institutional stereotactic radiosurgery database. The patient cohort was split into training (160 patients, 809 lesions), validation (20 patients, 136 lesions), and testing (15 patients, 89 lesions) datasets. The lesions in the testing dataset were further divided into two subgroups based on the diameters (small S = 1–10 mm, large L = 11–26 mm). In the testing dataset, there were 72 and 17 BMs in the S and L sub-groups, respectively. Among all trained networks, asym-UNet achieved the highest DSCv of 0.84 and lowest F BM of 0.24. Although vanilla 3D-UNet with a single 1 × 1 × 3 kernel achieved the highest sensitivities for the S group, it resulted in the lowest precision and highest false detection rate. Asym-UNet was shown to balance sensitivity and false detection rate as well as keep the segmentation accuracy high. The novel asym-UNet segmentation network showed overall competitive segmentation performance and more pronounced improvement in hard-to-detect small BMs comparing to the vanilla single 3D UNet.

No abstract

Geometric accuracy of MRI is one of the main concerns for its use as a sole image modality in precision radiation therapy (RT) planning. In a state-of-art scanner, system level geometric distortions are within acceptable levels for precision RT. However, subject-induced B0 inhomogeneity may vary substantially, especially in air-tissue interfaces. Recent studies have shown distortion levels of more than 2 mm near the sinus and ear canal are possible due to subject-induced field inhomogeneity. These distortions can be corrected with the use of accurate B0 inhomogeneity field maps. Most existing methods estimate these field maps from dual gradient-echo (GRE) images acquired at two different echo-times under the assumption that the GRE images are practically undistorted. However distortion that may exist in the GRE images can result in estimated field maps that are distorted in both geometry and intensity, leading to inaccurate correction of clinical images. This work proposes a method for estimating undistorted field maps from GRE acquisitions using an iterative joint estimation technique. The proposed method yields geometrically corrected GRE images and undistorted field maps that can also be used for the correction of images acquired by other sequences. The proposed method is validated through simulation, phantom experiments and applied to patient data. Our simulation results show that our method reduces the root-mean-squared error of the estimated field map from the ground truth by ten-fold compared to the distorted field map. Both the geometric distortion and the intensity corruption (artifact) in the images caused by the B0 field inhomogeneity are corrected almost completely. Our phantom experiment showed improvement in the geometric correction of approximately 1mm at an air-water interface using the undistorted field map compared to using a distorted field map. The proposed method for undistorted field map estimation can lead to improved geometric distortion correction at air-tissue interfaces, especially in low readout-bandwidth acquisitions, thus making them suitable for clinical use in precision RT without increasing the treatment planning margin.