Nie Xingyu, Li Guang
Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, United States.
Department of Radiology, University of Kentucky, Lexington, KY, United States.
Front Oncol. 2022 Jun 29;12:898771. doi: 10.3389/fonc.2022.898771. eCollection 2022.
To minimize computation latency using a predictive strategy to retrieve and project tumor volume onto 2D MR beam eye's view (BEV) cine from time-resolved four-dimensional magnetic resonance imaging (TR-4DMRI) libraries (inhalation/exhalation) for personalized MR-guided intensity-modulated radiotherapy (IMRT) or volumetric-modulated arc therapy (VMAT).
Two time-series forecasting algorithms, autoregressive (AR) modeling and deep-learning-based long short-term memory (LSTM), were applied to predict the diaphragm position in the next 2D BEV cine to identify a motion-matched and hysteresis-accounted image to retrieve the tumor volume from the inhalation/exhalation TR-4DMRI libraries. Three 40-s TR-4DMRI (2 Hz, 3 × 80 images) per patient of eight lung cancer patients were used to create patient-specific inhalation/exhalation 4DMRI libraries, extract diaphragmatic waveforms, and interpolate them to 4 and 8 Hz to match 2D cine frame rates. Along a (40•)-timepoint waveform, 30• training timepoints were moved forward to produce 3×(10•-1) predictions. The accuracy of position prediction was assessed against the waveform ground truth. The accuracy of tumor volume projections was evaluated using the center-of-mass difference (∆COM) and Dice similarity index against the TR-4DMRI ground truth for both IMRT (six beam angles, 30° interval) and VMAT (240/480 beam angles, 1.5°/0.75° interval, at 4/8 Hz, respectively).
The accuracy of the first-timepoint prediction is 0.36 ± 0.10 mm (AR) and 0.62 ± 0.21 mm (LSTM) at 4 Hz and 0.06 ± 0.02 mm (AR) and 0.18 ± 0.06 mm (LSTM) at 8 Hz. A 10%-20% random error in prediction-library matching increases the overall uncertainty slightly. For both IMRT and VMAT, the accuracy of projected tumor volume contours on 2D BEV cine is ∆COM = 0.39 ± 0.13 mm and DICE = 0.97 ± 0.02 at 4 Hz and ∆COM = 0.10 ± 0.04 mm and DICE = 1.00 ± 0.00 at 8Hz.
This study demonstrates the feasibility of accurately predicting respiratory motion during 2D BEV cine imaging, identifying a motion-matched and hysteresis-accounted tumor volume, and projecting tumor volume contour on 2D BEV cine for real-time assessment of beam-to-tumor conformality, promising for optimal personalized MR-guided radiotherapy.
使用预测策略将肿瘤体积从时间分辨的四维磁共振成像（TR-4DMRI）库（吸气/呼气）检索并投影到二维MR射野影像（BEV）电影上，以最小化计算延迟，用于个性化MR引导的调强放疗（IMRT）或容积调强弧形放疗（VMAT）。
应用两种时间序列预测算法，自回归（AR）建模和基于深度学习的长短期记忆（LSTM），预测下一个二维BEV电影中的膈肌位置，以识别运动匹配且考虑滞后的图像，从而从吸气/呼气TR-4DMRI库中检索肿瘤体积。八名肺癌患者每人的三个40秒TR-4DMRI（2Hz，3×80幅图像）用于创建患者特异性的吸气/呼气4DMRI库，提取膈肌波形，并将其插值到4Hz和8Hz以匹配二维电影帧率。沿着一个（40个）时间点的波形，将30个训练时间点向前移动以产生3×(10个 - 1) 预测。根据波形真实值评估位置预测的准确性。针对IMRT（六个射野角度，间隔30°）和VMAT（分别在4Hz和8Hz时为240/480个射野角度，间隔1.5°/0.75°），使用质心差异（∆COM）和骰子相似性指数，相对于TR-4DMRI真实值评估肿瘤体积投影的准确性。
在4Hz时，首次时间点预测的准确性为0.36±0.10mm（AR）和0.62±0.21mm（LSTM），在8Hz时为0.06±0.02mm（AR）和0.18±0.06mm（LSTM）。预测 - 库匹配中10%-20%的随机误差会略微增加整体不确定性。对于IMRT和VMAT，二维BEV电影上投影的肿瘤体积轮廓的准确性在4Hz时为∆COM = 0.39±0.13mm，DICE = 0.97±0.02，在8Hz时为∆COM = 0.10±0.04mm，DICE = 1.00±0.00。
本研究证明了在二维BEV电影成像期间准确预测呼吸运动、识别运动匹配且考虑滞后的肿瘤体积以及在二维BEV电影上投影肿瘤体积轮廓以实时评估射束与肿瘤适形性的可行性，有望实现最佳的个性化MR引导放疗。
