Search

PURPOSE: Radiation dose to cardiac substructures is related to radiation-induced heart disease. However, substructures are not considered in radiation therapy planning (RTP) due to poor visualization on CT. Therefore, we developed a novel deep learning (DL) pipeline leveraging MRI’s soft tissue contrast coupled with CT for state-of-the-art cardiac substructure segmentation requiring a single, non-contrast CT input. MATERIALS/METHODS: Thirty-two left-sided whole-breast cancer patients underwent cardiac T2 MRI and CT-simulation. A rigid cardiac-confined MR/CT registration enabled ground truth delineations of 12 substructures (chambers, great vessels (GVs), coronary arteries (CAs), etc.). Paired MRI/CT data (25 patients) were placed into separate image channels to train a three-dimensional (3D) neural network using the entire 3D image. Deep supervision and a Dice-weighted multi-class loss function were applied. Results were assessed pre/post augmentation and post-processing (3D conditional random field (CRF)). Results for 11 test CTs (seven unique patients) were compared to ground truth and a multi-atlas method (MA) via Dice similarity coefficient (DSC), mean distance to agreement (MDA), and Wilcoxon signed-ranks tests. Three physicians evaluated clinical acceptance via consensus scoring (5-point scale). RESULTS: The model stabilized in ~19 h (200 epochs, training error 0.05) to ground truth. In four cases, DL yielded left main CA contours, whereas MA segmentation failed, and provided improved consensus scores in 44/60 comparisons to MA. DL provided clinically acceptable segmentations for all graded patients for 3/4 chambers. DL contour generation took ~14 s per patient. CONCLUSIONS: These promising results suggest DL poses major efficiency and accuracy gains for cardiac substructure segmentation offering high potential for rapid implementation into RTP for improved cardiac sparing.

Purpose: A recent study by Korreman et al (IJROBP, vol 65, 1375–1380, 2006) demonstrates possible >85% relative reduction of cardiac‐pulmonary complications for tangential breast irradiation by using deep inspiration breath‐hold or gating techniques. The aim of this work is to minimize the risks by correcting the breast displacements using real‐time stereovision guidance. Materials/Methods: Twenty breast‐cancer patients were accrued over the last seventeen months on an IRB‐approved study. Planning target volume (PTV) was encompassed by the prescribed isodose in tangential beams with intensity modulation. The surgical bed was boosted using an electron field or conformal photon beams. The planning information (include CTimages) were transferred to an in‐house stereovision‐guided program which automatically created the reference surfaceimages from the CT‐based plans. The reference surfaces were matched with the real‐time surfaceimages, captured with 3D cameras mounted in the CT simulation room and treatment vault, to determine the displacements of CT setup markers, daily initial setup isocenter, final target position, and target motion during treatment. Large and significant position error were corrected based the image‐guidance.Images and on‐line adjustments were automatically stored for the analysis. Portal images were taken weekly for the position verification. Results: The 3D isocenter displacements improved from ∼10 mm to ∼ 4 mm after the application of IGRT. Isocenter setup error and simulation‐marking error were detected and corrected. All major changes were confirmed with portal images. The intrafractional motion caused only ∼2.0 mm target displacement. The entire procedure (including the table shifts) took < 5 minutes per day. Conclusions: The clinical results demonstrate improvement for breast target positioning using stereovision guidance. This in‐house IGRT for BC is clinically feasible, efficient, and accurate. Research initially sponsored by NIH SBIR‐1R43CA91690‐01 and CA‐88843 grants.

Purposes: to implement and compare four newly developed image‐guidedradiotherapy systems (Varian's Cone‐Beam CT, BrainLAB ExacTrac, Restitu Ultrasound (U/S)‐Sim and Guide, and in‐house stereovision) in one department. Methods and Materials: The cone‐beam CT(CBCT) and the ultrasound(US) systems provide volumetric images of the target at daily setup. The ExacTrac system acquires the biplanar radiographs at patient setup. Both the US and ExacTrac systems are integrated with infrared‐tracking systems for patient‐couch positioning. The in‐house stereovision system captures 3D surface images of the patient at the instants of daily patient setup and during individual beam irradiation. All of four IGRT systems have used treatment planning volumetric imaging information for target position verification and adjustment. Electronic portal images are routinely used for patient position verification. External markers and possible internal markers such as seeds or small cysts or calcifications can be localized and used for additional verification. Results: Emerging data from several institutional IRB‐approved clinical trials demonstrate that the target reposition error and dose delivery uncertainties can be significantly reduced by using such image‐guided systems, each of which may be most useful in specific clinical situations. Conclusions: Our customized stereovision system, which, like US, involves no radiation exposure, is extremely efficient (