Treatment planning plays an important role in the process of radiotherapy (RT). The quality of the treatment plan directly and significantly affects patient treatment outcomes. In the past decades, technological advances in computer and software have promoted the development of RT treatment planning systems with sophisticated dose calculation and optimization algorithms. Treatment planners now have greater flexibility in designing highly complex RT treatment plans in order to mitigate the damage to healthy tissues better while maximizing radiation dose to tumor targets. Nevertheless, treatment planning is still largely a time-inefficient and labor-intensive process in current clinical practice. Artificial intelligence, including machine learning (ML) and deep learning (DL), has been recently used to automate RT treatment planning and has gained enormous attention in the RT community due to its great promises in improving treatment planning quality and efficiency. In this article, we reviewed the historical advancement, strengths, and weaknesses of various DL-based automated RT treatment planning techniques. We have also discussed the challenges, issues, and potential research directions of DL-based automated RT treatment planning techniques.

Lung cancer is the leading cause of cancer-related mortality for males and females. Radiation therapy (RT) is one of the primary treatment modalities for lung cancer. While delivering the prescribed dose to tumor targets, it is essential to spare the tissues near the targets—the so-called organs-at-risk (OARs). An optimal RT planning benefits from the accurate segmentation of the gross tumor volume and surrounding OARs. Manual segmentation is a time-consuming and tedious task for radiation oncologists. Therefore, it is crucial to develop automatic image segmentation to relieve radiation oncologists of the tedious contouring work. Currently, the atlas-based automatic segmentation technique is commonly used in clinical routines. However, this technique depends heavily on the similarity between the atlas and the image segmented. With significant advances made in computer vision, deep learning as a part of artificial intelligence attracts increasing attention in medical image automatic segmentation. In this article, we reviewed deep learning based automatic segmentation techniques related to lung cancer and compared them with the atlas-based automatic segmentation technique. At present, the auto-segmentation of OARs with relatively large volume such as lung and heart etc. outperforms the organs with small volume such as esophagus. The average Dice similarity coefficient (DSC) of lung, heart and liver are over 0.9, and the best DSC of spinal cord reaches 0.9. However, the DSC of esophagus ranges between 0.71 and 0.87 with a ragged performance. In terms of the gross tumor volume, the average DSC is below 0.8. Although deep learning based automatic segmentation techniques indicate significant superiority in many aspects compared to manual segmentation, various issues still need to be solved. We discussed the potential issues in deep learning based automatic segmentation including low contrast, dataset size, consensus guidelines, and network design. Clinical limitations and future research directions of deep learning based automatic segmentation were discussed as well.

Patient-specific quality assurance (PSQA) of volumetric modulated arc therapy (VMAT) to assure accurate treatment delivery is resource-intensive and time-consuming. Recently, machine learning has been increasingly investigated in PSQA results prediction. However, the classification performance of models at different criteria needs further improvement and clinical validation (CV), especially for predicting plans with low gamma passing rates (GPRs). In this study, we developed and validated a novel multi-task model called autoencoder based classification-regression (ACLR) for VMAT PSQA. The classification and regression were integrated into one model, both parts were trained alternatively while minimizing a defined loss function. The classification was used as an intermediate result to improve the regression accuracy. Different tasks of GPRs prediction and classification based on different criteria were trained simultaneously. Balanced sampling techniques were used to improve the prediction accuracy and classification sensitivity for the unbalanced VMAT plans. Fifty-four metrics were selected as inputs to describe the plan modulation-complexity and delivery-characteristics, while the outputs were PSQA GPRs. A total of 426 clinically delivered VMAT plans were used for technical validation (TV), and another 150 VMAT plans were used for CV to evaluate the generalization performance of the model. The ACLR performance was compared with the Poisson Lasso (PL) model and found significant improvement in prediction accuracy. In TV, the absolute prediction error (APE) of ACLR was 1.76%, 2.60%, and 4.66% at 3%/3 mm, 3%/2 mm, and 2%/2 mm, respectively; whereas the APE of PL was 2.10%, 3.04%, and 5.29% at 3%/3 mm, 3%/2 mm, and 2%/2 mm, respectively. No significant difference was found between CV and TV in prediction accuracy. ACLR model set with 3%/3 mm can achieve 100% sensitivity and 83% specificity. The ACLR model could classify the unbalanced VMAT QA results accurately, and it can be readily applied in clinical practice for virtual VMAT QA.

Purpose: To compare plan quality and delivery efficiency of noncoplanar volumetric modulated arc therapy with coplanar volumetric modulated arc therapy, intensity-modulated radiation therapy, and CyberKnife for multiple brain metastases. Methods: For 15 patients with multiple brain metastases, noncoplanar volumetric modulated arc therapy, coplanar volumetric modulated arc therapy, intensity-modulated radiation therapy, and CyberKnife plans with a prescription dose of 30 Gy in 3 fractions were generated. Noncoplanar volumetric modulated arc therapy and coplanar volumetric modulated arc therapy plans consisted of 4 noncoplanar arcs and 2 full coplanar arcs, respectively. Intensity-modulated radiation therapy plans consisted of 7 coplanar fields. CyberKnife plans used skull tracking to ensure accurate position. All plans were generated to cover 95% target volume with prescription dose. Gradient index, conformity index, normal brain tissue volume ( V _3Gy − V _24Gy), monitor units, and beam on time were evaluated. Results: Gradient index was the lowest for CyberKnife (3.49 ± 0.65), followed by noncoplanar volumetric modulated arc therapy (4.21 ± 1.38), coplanar volumetric modulated arc therapy (4.87 ± 1.35), and intensity-modulated radiation therapy (5.36 ± 1.98). Conformity index was the largest for noncoplanar volumetric modulated arc therapy (0.87 ± 0.03), followed by coplanar volumetric modulated arc therapy (0.86 ± 0.04), CyberKnife (0.86 ± 0.07), and intensity-modulated radiation therapy (0.85 ± 0.05). Normal brain tissue volume at high-to-moderate dose spreads ( V _24Gy − V _9Gy) was significantly reduced in noncoplanar volumetric modulated arc therapy over that of intensity-modulated radiation therapy and coplanar volumetric modulated arc therapy. Normal brain tissue volume for noncoplanar volumetric modulated arc therapy was comparable with noncoplanar volumetric modulated arc therapy at high-dose level ( V _24Gy − V _15Gy) and larger than CyberKnife at moderate-to-low dose level ( V _12Gy − V _3Gy). Monitor units was highest for CyberKnife (28 733.59 ± 7197.85), followed by intensity-modulated radiation therapy (4128.40 ± 1185.38), noncoplanar volumetric modulated arc therapy (3105.20 ± 371.23), and coplanar volumetric modulated arc therapy (2997.27 ± 446.84). Beam on time was longest for CyberKnife (30.25 ± 7.32 minutes), followed by intensity-modulated radiation therapy (2.95 ± 0.85 minutes), noncoplanar volumetric modulated arc therapy (2.61 ± 0.07 minutes), and coplanar volumetric modulated arc therapy (2.30 ± 0.23 minutes). Conclusion: For brain metastases far away from organs-at-risk, noncoplanar volumetric modulated arc therapy generated more rapid dose falloff and higher conformity compared to intensity-modulated radiation therapy and coplanar volumetric modulated arc therapy. Noncoplanar volumetric modulated arc therapy provided a comparable dose falloff with CyberKnife at high-dose level and a slower dose falloff than CyberKnife at moderate-to-low dose level. Noncoplanar volumetric modulated arc therapy plans had less monitor units and shorter beam on time than CyberKnife plans.

Background: A preliminary assessment was conducted of the feasibility, efficacy, and morbidity of ¹²⁵I seed implantation for recurrent head and neck carcinoma after surgery and external beam radiotherapy. Methods: Nineteen patients with recurrent head and neck carcinomas underwent ¹²⁵I seed implantation under ultrasound or computed tomography guidance. The actuarial D90 of ¹²⁵I seed implantation ranged from 90 to 160 Gy (median, 131 Gy). The follow-up period ranged from 3 to 44 months (median, 11 months). Results: The median local control was 24 months (95% confidence interval, 10.2 to 37.8). The one- year, two-year and three-year local controls were 73.3%, 27.5% and 27.5%, respectively, whereas the one-year, two-year and three-year survival rates were 53.0%, 18.2% and 18.2%, respectively, and the median survival was 13 months (95% confidence interval, 6.6 to 19.4). A total of 26.3% of patients (5/19) died of local recurrence and 21.1% of patients (4/19) died of metastases. One suffered from a grade 1 skin reaction. Conclusions: ¹²⁵I seed implantation is feasible and safe as a salvage treatment for patients with recurrent head and neck cancers. The high local control results and low morbidity merits further investigation.