A convolutional neural network algorithm for automatic segmentation of head and neck organs at risk using deep lifelong learning.

PURPOSE: This study suggests a lifelong learning-based convolutional neural network (LL-CNN) algorithm as a superior alternative to single-task learning approaches for automatic segmentation of head and neck (OARs) organs at risk. METHODS AND MATERIALS: Lifelong learning-based convolutional neural network was trained on twelve head and neck OARs simultaneously using a multitask learning framework. Once the weights of the shared network were established, the final multitask convolutional layer was replaced by a single-task convolutional layer. The single-task transfer learning network was trained on each OAR separately with early stoppage. The accuracy of LL-CNN was assessed based on Dice score and root-mean-square error (RMSE) compared to manually delineated contours set as the gold standard. LL-CNN was compared with 2D-UNet, 3D-UNet, a single-task CNN (ST-CNN), and a pure multitask CNN (MT-CNN). Training, validation, and testing followed Kaggle competition rules, where 160 patients were used for training, 20 were used for internal validation, and 20 in a separate test set were used to report final prediction accuracies. RESULTS: On average contours generated with LL-CNN had higher Dice coefficients and lower RMSE than 2D-UNet, 3D-Unet, ST- CNN, and MT-CNN. LL-CNN required ~72 hrs to train using a distributed learning framework on 2 Nvidia 1080Ti graphics processing units. LL-CNN required 20 s to predict all 12 OARs, which was approximately as fast as the fastest alternative methods with the exception of MT-CNN. CONCLUSIONS: This study demonstrated that for head and neck organs at risk, LL-CNN achieves a prediction accuracy superior to all alternative algorithms.

Machine Learning in Radiation Oncology: Opportunities, Requirements, and Needs.

Machine learning (ML) has the potential to revolutionize the field of radiation oncology, but there is much work to be done. In this article, we approach the radiotherapy process from a workflow perspective, identifying specific areas where a data-centric approach using ML could improve the quality and efficiency of patient care. We highlight areas where ML has already been used, and identify areas where we should invest additional resources. We believe that this article can serve as a guide for both clinicians and researchers to start discussing issues that must be addressed in a timely manner.

Dosimetric impact of interfraction catheter movement and organ motion on MRI/CT guided HDR interstitial brachytherapy for gynecologic cancer.

PURPOSE: To determine the dosimetric impact of catheter movement for MRI/CT image guided high dose rate (HDR) interstitial brachytherapy (ISBT) for gynecologic cancers. MATERIALS AND METHODS: Ten patients were treated with HDR ISBT. The CTV and organs at risk were contoured using a postimplant MRI and CT. 5 fractions were delivered twice daily on 3 consecutive days. The first fraction was delivered on day 1 (d1), fraction 2-3 on d2 and fraction 4-5 on d3. MRI/CT was acquired prior to the second and fourth fractions. Four scenarios were modeled. (1) The d1 plan was applied to the d2 and d3 CT, using the updated catheter positions. (2) Replanning was performed for d2 and d3. (3) We applied the dwell positions/times from the d2 replan over the d3 CT and compared with a d3 CT replan. (4) Based on daily MRI, target volumes were recontoured and replanned. Dosimetry was analyzed for each plan and compared to the d1 dose distribution. RESULTS: (1) When using the d1 plan on the d2 and d3 CT with the updated catheter positions, the mean CTV D90 was reduced from 93.4% on d1 to 89.3% (p=0.08) on d2 and to 87.7% (p=0.005) on d3. (2) Replanning on d2 and d3 compensated for catheter movement, mean CTV D90 of 95.4% on d2 and 94.6% (p=0.36) on d3. (3) When compared to the replan of d2 applied on the d3 CT vs the d3 replan, there was no significant difference in coverage, mean CTV D90 of 90.9% (p=0.09). (4) Reoptimization based on daily MRI, significantly improved the CTV coverage for each day. The mean D2cc for the rectum was significantly higher with model 1 vs model 3 59.1±4.7 vs 60.9±4.8 (p=0.04) Gy EQD2. There were no significant differences in D2cc of bladder and sigmoid between models. CONCLUSIONS: Interfraction dosimetric changes significantly decreased the CTV coverage of the third day. Rather than replanning on each day, replanning on the day 2 CT before the second or third fraction would give an optimal solution that would compensate for interfraction catheter displacement.