Improved accuracy of auto-segmentation of organs at risk in radiotherapy planning for nasopharyngeal carcinoma based on fully convolutional neural network deep learning.

OBJECTIVE: We examined a modified encoder-decoder architecture-based fully convolutional neural network, OrganNet, for simultaneous auto-segmentation of 24 organs at risk (OARs) in the head and neck, followed by validation tests and evaluation of clinical application. MATERIALS AND METHODS: Computed tomography (CT) images from 310 radiotherapy plans were used as the experimental data set, of which 260 and 50 were used as the training and test sets, respectively. An improved U-Net architecture was established by introducing a batch normalization layer, residual squeeze-and-excitation layer, and unique organ-specific loss function for deep learning training. The performance of the trained network model was evaluated by comparing the manual-delineation and the STAPLE contour of 10 physicians from different centers. RESULTS: Our model achieved good segmentation in all 24 OARs in nasopharyngeal cancer radiotherapy plan CT images, with an average Dice similarity coefficient of 83.75%. Specifically, the mean Dice coefficients in large-volume organs (brainstem, spinal cord, left/right parotid glands, left/right temporal lobes, and left/right mandibles) were 84.97% - 95.00%, and in small-volume organs (pituitary, lens, optic nerve, and optic chiasma) were 55.46% - 91.56%. respectively. Using the STAPLE contours as standard contour, the OrganNet achieved comparable or better DICE in organ segmentation then that of the manual-delineation as well. CONCLUSION: The established OrganNet enables simultaneous automatic segmentation of multiple targets on CT images of the head and neck radiotherapy plans, effectively improves the accuracy of U-Net based segmentation for OARs, especially for small-volume organs.

Comparison of Different Combinations of Irradiation Mode and Jaw Width in Helical Tomotherapy for Nasopharyngeal Carcinoma.

Purpose: To aid in the selection of a suitable combination of irradiation mode and jaw width in helical tomotherapy (HT) for the treatment of nasopharyngeal carcinoma (NPC). Materials and Methods: Twenty patients with NPC who underwent radiotherapy were retrospectively selected. Four plans using a jaw width of 2.5 or 5-cm in dynamic jaw (DJ) or fix jaw (FJ) modes for irradiation were designed (2.5DJ, 2.5FJ, 5.0DJ, and 5.0FJ). The dose parameters of planning target volume (PTV) and organs at risk (OARs) of the plans were compared and analyzed, as well as the beam on time (BOT) and monitor unit (MU). The plans in each group were ranked by scoring the doses received by the OARs and the superity was assessed in combination with the planned BOT and MU. Results: The prescribed dose coverage of PTV met the clinical requirements for all plans in the four groups. The groups using a 2.5-cm jaw width or a DJ mode provided better protection to most OARs, particularly for those at the longitudinal edges of the PTV (P < 0.05). The 2.5DJ group had the best ranking for OAR-dose, followed by the 2.5FJ and 5.0DJ groups with a same score. The BOT and MU of the groups using a 5.0-cm jaw width reduced nearly 45% comparing to those of the 2.5-cm jaw groups. Conclusion: 2.5DJ has the best dose distribution, while 5.0DJ has satisfactory dose distribution and less BOT and MU that related to the leakage dose. Both 2.5DJ or 5DJ were recommended for HT treatment plan for NPC based on the center workload.

Analysis of late toxicity in nasopharyngeal carcinoma patients treated with intensity modulated radiation therapy.

PURPOSE: To observe the late toxicities in nasopharyngeal carcinoma (NPC) patients who achieved long-term survival after intensity modulated radiation therapy (IMRT). METHODS: 208 untreated NPC patients who received IMRT and survived more than five years with locoregional disease control and no metastasis were evaluated in this study. The prescription dose to the gross target volume of nasopharynx (GTVnx), positive neck lymph nodes (GTVnd), clinical target volume 1 (CTV1) and 2 (CTV2) was 68Gy/30f, 60-66Gy/30f, 60 Gy/30f and 54Gy/30f, respectively. The nasopharynx and upper neck targets were irradiated using IMRT, and the lower neck and supraclavicular fossae targets were irradiated using the half-beam technique with conventional irradiation. The late toxicities were evaluated according to the LENT/SOMA criteria of 1995. RESULTS: The median follow-up time was 78 months (60-96 months). The occurrence rates of cervical subcutaneous fibrosis, hearing loss, skin dystrophy, xerostomia, trismus, temporal lobe injury, cranial nerve damage, cataract, and brain stem injury induced by radiotherapy were 89.9%, 67.8%, 47.6%, 40.9%, 7.21%, 4.33%, 2.88%, 1.44%, and 0.48%, respectively. No spinal cord injury and mandible damage were found. Grade 3-4 late injuries were observed as follows: 1 (0.48%) skin dystrophy, 4 (1.92%) cervical subcutaneous fibrosis, 2 (0.96%) hearing loss, 2 (0.96%) cranial nerve palsy, and 1 (0.48%) temporal lobe necrosis. No grade 3-4 late injuries occurred in parotid, temporomandibular joints and eyes. Xerostomia decreased gradually over time and then showed only slight changes after 4 years. The change in the incisor distance stabilised by 1 year after RT, however, the incidence of hearing loss, skin dystrophy, subcutaneous fibrosis and nervous system injuries increased over time after RT. CONCLUSION: The late injuries in most NPC patients who had long-term survivals after IMRT are alleviated. Within the 5 years of follow-up, we found xerostomia decreased gradually; The change in the incisor distance stabilised by 1 year after RT; while hearing loss, nervous system injuries increased over time after RT.

Real-time in vivo dosimetry with MOSFET detectors in serial tomotherapy for head and neck cancer patients.

PURPOSE: A real-time dose verification method using a recently designed metal oxide semiconductor field effect transistor (MOSFET) dosimetry system was evaluated for quality assurance (QA) of intensity-modulated radiation therapy (IMRT). METHODS AND MATERIALS: Following the investigation of key parameters that might affect the accuracy of MOSFET measurements (i.e., source surface distance [SSD], field size, beam incident angles and radiation energy spectrum), the feasibility of this detector in IMRT dose verification was demonstrated by comparison with ion chamber measurements taken in an IMRT QA phantom. Real-time in vivo measurements were also performed with the MOSFET system during serial tomotherapy treatments administered to 8 head and neck cancer patients. RESULTS: MOSFET sensitivity did not change with SSD. For field sizes smaller than 20 × 20 cm(2), MOFET sensitivity varied within 1.0%. The detector angular response was isotropic within 2% over 360°, and the observed sensitivity variation due to changes in the energy spectrum was negligible in 6-MV photons. MOSFET system measurements and ion chamber measurements agreed at all points in IMRT phantom plan verification, within 5%. The mean difference between 48 IMRT MOSFET-measured doses and calculated values in 8 patients was 3.33% and ranged from -2.20% to 7.89%. More than 90% of the total measurements had deviations of less than 5% from the planned doses. CONCLUSION: The MOSFET dosimetry system has been proven to be an effective tool in evaluating the actual dose within individual patients during IMRT treatment.