ABSTRACT
AIM: This study evaluated a convolutional neural network (CNN) for automatically delineating the liver on contrast-enhanced or non-contrast-enhanced CT, making comparisons with a commercial automated technique (MIM Maestro®). BACKGROUND: Intensity-modulated radiation therapy requires careful labor-intensive planning involving delineation of the target and organs on CT or MR images to ensure delivery of the effective dose to the target while avoiding organs at risk. MATERIALS AND METHODS: Contrast-enhanced planning CT images from 101 pancreatic cancer cases and accompanying mask images showing manually-delineated liver contours were used to train the CNN to segment the liver. The trained CNN then performed liver segmentation on a further 20 contrast-enhanced and 15 non-contrastenhanced CT image sets, producing three-dimensional mask images of the liver. RESULTS: For both contrast-enhanced and non-contrast-enhanced images, the mean Dice similarity coefficients between CNN segmentations and ground-truth manual segmentations were significantly higher than those between ground-truth and MIM Maestro software (pâ¯<â¯0.001). Although mean CT values of the liver were higher on contrast-enhanced than on non-contrast-enhanced CT, there were no significant differences in the Hausdorff distances of the CNN segmentations, indicating that the CNN could successfully segment the liver on both image types, despite being trained only on contrast-enhanced images. CONCLUSIONS: Our results suggest that a CNN can perform highly accurate automated delineation of the liver on CT images, irrespective of whether the CT images are contrast-enhanced or not.
ABSTRACT
PURPOSE: To determine whether maximum or average intensity projection (MIP or AIP, respectively) reconstructed from 4-dimensional computed tomography (4DCT) is preferred for alignment to cone beam CT (CBCT) images in lung stereotactic body radiation therapy. METHODS AND MATERIALS: Stationary CT and 4DCT images were acquired with a target phantom at the center of motion and moving along the superior-inferior (SI) direction, respectively. Motion profiles were asymmetrical waveforms with amplitudes of 10, 15, and 20 mm and a 4-second cycle. Stationary CBCT and dynamic CBCT images were acquired in the same manner as stationary CT and 4DCT images. Stationary CBCT was aligned to stationary CT, and the couch position was used as the baseline. Dynamic CBCT was aligned to the MIP and AIP of corresponding amplitudes. Registration error was defined as the SI deviation of the couch position from the baseline. In 16 patients with isolated lung lesions, free-breathing CBCT (FBCBCT) was registered to AIP and MIP (64 sessions in total), and the difference in couch shifts was calculated. RESULTS: In the phantom study, registration errors were within 0.1 mm for AIP and 1.5 to 1.8 mm toward the inferior direction for MIP. In the patient study, the difference in the couch shifts (mean, range) was insignificant in the right-left (0.0 mm, ≤1.0 mm) and anterior-posterior (0.0 mm, ≤2.1 mm) directions. In the SI direction, however, the couch position significantly shifted in the inferior direction after MIP registration compared with after AIP registration (mean, -0.6 mm; ranging 1.7 mm to the superior side and 3.5 mm to the inferior side, P=.02). CONCLUSIONS: AIP is recommended as the reference image for registration to FBCBCT when target alignment is performed in the presence of asymmetrical respiratory motion, whereas MIP causes systematic target positioning error.
