MRI-Only Radiotherapy Planning for Nasopharyngeal Carcinoma Using Deep Learning

BackgroundRadical radiotherapy is the main treatment modality for early and locally advanced nasopharyngeal carcinoma (NPC). Magnetic resonance imaging (MRI) has the advantages of no ionizing radiation and high soft-tissue resolution compared to computed tomography (CT), but it does not provide electron density (ED) information for radiotherapy planning. Therefore, in this study, we developed a pseudo-CT (pCT) generation method to provide necessary ED information for MRI-only planning in NPC radiotherapy.MethodsTwenty patients with early-stage NPC who received radiotherapy in our hospital were investigated. First, 1433 sets of paired T1 weighted magnetic resonance (MR) simulation images and CT simulation images were rigidly registered and preprocessed. A 16-layer U-Net was used to train the pCT generative model and a “pix2pix” generative adversarial network (GAN) was also trained to compare with the pure U-Net regrading pCT quality. Second, the contours of all target volumes and organs at risk in the original CT were transferred to the pCT for planning, and the beams were copied back to the original CT for reference dose calculation. Finally, the dose distribution calculated on the pCT was compared with the reference dose distribution through gamma analysis and dose-volume indices.ResultsThe average time for pCT generation for each patient was 7.90 ± 0.47 seconds. The average mean (absolute) error was −9.3 ± 16.9 HU (102.6 ± 11.4 HU), and the mean-root-square error was 209.8 ± 22.6 HU. There was no significant difference between the pCT quality of pix2pix GAN and that of pure U-Net (p > 0.05). The dose distribution on the pCT was highly consistent with that on the original CT. The mean gamma pass rate (2 mm/3%, 10% low dose threshold) was 99.1% ± 0.3%, and the mean absolute difference of nasopharyngeal PGTV D99% and PTV V95% were 0.4% ± 0.2% and 0.1% ± 0.1%.ConclusionThe proposed deep learning model can accurately predict CT from MRI, and the generated pCT can be employed in precise dose calculations. It is of great significance to realize MRI-only planning in NPC radiotherapy, which can improve structure delineation and considerably reduce additional imaging dose, especially when an MR-guided linear accelerator is adopted for treatment.

Download Full-text

A dosimetric evaluation on applying RTOG-based and CT/MRI-based delineation methods to brachial plexus in radiotherapy of nasopharyngeal carcinoma treated with helical tomotherapy

British Journal of Radiology ◽

10.1259/bjr.20170881 ◽

2019 ◽

Vol 92 (1102) ◽

pp. 20170881 ◽

Cited By ~ 1

Author(s):

Chi-Him Li ◽

Vincent WC Wu ◽

George Chiu

Keyword(s):

Nasopharyngeal Carcinoma ◽

Brachial Plexus ◽

Helical Tomotherapy ◽

Ct Images ◽

Radiotherapy Planning ◽

Significant Difference ◽

Plan B ◽

The Mean ◽

Dose Constraints ◽

Planning Ct

Objective: In radiotherapy of nasopharyngeal carcinoma (NPC) patients, the brachial plexus (BP) situated at both sides of the neck is often irradiated to high dose. This study was to evaluate different BP delineation methods and analyse the dosimetric consequences when applying BP dose constraints in radiotherapy planning of NPC. Methods: 15 NPC cases radically treated with helical tomotherapy were recruited. Apart from the original treatment plan (Plan A), two new plans (Plans B and C) with additional BP dose constraints were computed using the same planning CT images, structures and planning parameters. Plan B consisted of BP contours based on Radiation Therapy Oncology Group (RTOG)-endorsed atlas; while those in Plan C were based on MR images registered with the planning CT images. Results: The mean BP volume by RTOG method was 19.04 ± 3.50 cm3 vs 10.44 ± 2.00 cm3 by CT/MRI method. The mean BP overlapping volume between the two contouring methods was 1.9 cm3 (0.38–4.03 cm3). There was significant difference between two methods (p < 0.001). The average Dmax, Dmean, D5%, D10% and D15% of both sides of BP in Plan A were significantly higher than those in both Plan B and Plan C. There were no significant dose differences in the targets and organs at risk (OARs) after applying dose constraints in Plan B and Plan C. Conclusion: RTOG method was recommended since larger BP volume provided better protection. Applying BP dose constraints during tomotherapy plan optimisation for NPC patients could significantly reduce the BP dose (p < 0.05) without compromising the doses to the targets and other OARs. Advances in knowledge: This is the first study comparing the delineation method based on RTOG-endorsed atlas with the conventional CT/MRI delineation method for BP in tomotherapy of NPC patients. Our results showed that BP dose could be significantly reduced after applying the dose constraints without compromising the doses to the target volumes and other OARs. The RTOG method was more favoured as it gave a relatively larger BP volume and therefore offered better organ sparing.

Download Full-text

A deep-learning-based prognostic nomogram integrating microscopic digital pathology and macroscopic magnetic resonance images in nasopharyngeal carcinoma: a multi-cohort study

Therapeutic Advances in Medical Oncology ◽

10.1177/1758835920971416 ◽

2020 ◽

Vol 12 ◽

pp. 175883592097141

Author(s):

Fan Zhang ◽

Lian-Zhen Zhong ◽

Xun Zhao ◽

Di Dong ◽

Ji-Jin Yao ◽

...

Keyword(s):

Deep Learning ◽

Magnetic Resonance ◽

Nasopharyngeal Carcinoma ◽

Treatment Failure ◽

Digital Pathology ◽

Imaging Features ◽

Multi Scale ◽

External Test ◽

Whole Slide Images ◽

Independent Cohort

Background: To explore the prognostic value of radiomics-based and digital pathology-based imaging biomarkers from macroscopic magnetic resonance imaging (MRI) and microscopic whole-slide images for patients with nasopharyngeal carcinoma (NPC). Methods: We recruited 220 NPC patients and divided them into training ( n = 132), internal test ( n = 44), and external test ( n = 44) cohorts. The primary endpoint was failure-free survival (FFS). Radiomic features were extracted from pretreatment MRI and selected and integrated into a radiomic signature. The histopathological signature was extracted from whole-slide images of biopsy specimens using an end-to-end deep-learning method. Incorporating two signatures and independent clinical factors, a multi-scale nomogram was constructed. We also tested the correlation between the key imaging features and genetic alternations in an independent cohort of 16 patients (biological test cohort). Results: Both radiomic and histopathologic signatures presented significant associations with treatment failure in the three cohorts (C-index: 0.689–0.779, all p < 0.050). The multi-scale nomogram showed a consistent significant improvement for predicting treatment failure compared with the clinical model in the training (C-index: 0.817 versus 0.730, p < 0.050), internal test (C-index: 0.828 versus 0.602, p < 0.050) and external test (C-index: 0.834 versus 0.679, p < 0.050) cohorts. Furthermore, patients were stratified successfully into two groups with distinguishable prognosis (log-rank p < 0.0010) using our nomogram. We also found that two texture features were related to the genetic alternations of chromatin remodeling pathways in another independent cohort. Conclusion: The multi-scale imaging features showed a complementary value in prognostic prediction and may improve individualized treatment in NPC.

Download Full-text

Oropharyngeal primary tumor segmentation for radiotherapy planning on magnetic resonance imaging using deep learning

Physics and Imaging in Radiation Oncology ◽

10.1016/j.phro.2021.06.005 ◽

2021 ◽

Vol 19 ◽

pp. 39-44

Author(s):

Roque Rodríguez Outeiral ◽

Paula Bos ◽

Abrahim Al-Mamgani ◽

Bas Jasperse ◽

Rita Simões ◽

...

Keyword(s):

Magnetic Resonance Imaging ◽

Deep Learning ◽

Magnetic Resonance ◽

Primary Tumor ◽

Radiotherapy Planning ◽

Tumor Segmentation ◽

Resonance Imaging

Download Full-text

The influence of overall treatment time to the efficiency of chemo-radiotherapy for locally advanced cervical cancer

Journal of Radiotherapy in Practice ◽

10.1017/s1460396917000371 ◽

2017 ◽

Vol 17 (1) ◽

pp. 124-130

Author(s):

Ekkasit Tharavichitkul ◽

Panupat Rugpong ◽

Nisa Chawapun ◽

Razvan M. Galalae

Keyword(s):

Cervical Cancer ◽

Treatment Time ◽

Locally Advanced ◽

Disease Free Survival ◽

Intracavitary Brachytherapy ◽

Treatment Results ◽

Overall Treatment Time ◽

Free Survival ◽

Significant Difference ◽

The Mean

AbstractPurposeThis study aims to clarify the influence of overall treatment time (OTT) on the efficiency of combined chemo-radiotherapy in cervical cancer.Material and methodsThis retrospective study enrolled 122 cervical cancer patients who had squamous cell carcinoma and had undergone definitive chemo-radiotherapy from 2009 to 2013. All patients received whole pelvic radiotherapy (WPRT) with the dose of 50 Gy in 25 fractions (with central shielding after 44 Gy) plus intracavitary brachytherapy with the dose of 28 Gy in four fractions. During WPRT, all patients received concurrent chemotherapy with weekly platinum-based regimen. The data of patient characteristics, OTT, treatment results and toxicities were collected and evaluated.ResultsThe mean follow-up time was 36 months. The mean age of patients was 52 years old; 68% of patients were stage IIB related to International Federation of Gynaecology and Obstetrics staging. Pelvic control (PC), distant metastasis-free survival (DMFS), disease-free survival (DFS) and overall survival (OS) rates did not differ significantly in the data-derived cut points of 55·8 and 53 days. No statistically significant difference in treatment results between the two groups of OTT<49 and OTT≥62 days was observed.ConclusionsIn our data-derived cut point, OTT did not influence to PC, DMFS, DFS and OS. The influence of OTT on treatment results may be found in longer periods.

Download Full-text

Sexual Dimorphism in Laryngeal Volumetric Measurements Using Magnetic Resonance Imaging

Ear Nose & Throat Journal ◽

10.1177/0145561319840568 ◽

2019 ◽

Vol 99 (2) ◽

pp. 132-136 ◽

Cited By ~ 1

Author(s):

Abdul-Latif Hamdan ◽

Elie Khalifee ◽

Georges Ziade ◽

Sahar Semaan

Keyword(s):

Magnetic Resonance Imaging ◽

Magnetic Resonance ◽

Statistical Significance ◽

Resonance Imaging ◽

Men And Women ◽

Mean Width ◽

Significant Difference ◽

The Mean ◽

The Right ◽

Volumetric Measurements

The objective of this study is to investigate the dimensional and volumetric measurements in the thyroarytenoid (TA) muscle in men and women using magnetic resonance imaging (MRI). The hypothesis is that there is a gender-related difference in these measurements. A retrospective chart review of 76 patients who underwent MRI of the neck at the American University of Beirut Medical Center was conducted. The dimension and volume of the right and left TA muscle were measured on axial and coronal planes short tau inversion recovery images. Male and female groups were compared with respect to demographic data and MRI findings using parametric and nonparametric tests. The mean length of the thyro-arytenoid muscle in males was larger than that in females on the right (males 2.44 [0.29] cm vs females 1.70 [0.22] cm) and on the left (males 2.50 [0.28] cm vs females 1.72 [0.24] cm) reaching statistical significance ( P < .001). The mean width of the thyro-arytenoid muscle in males was larger than that in females on the right (males 0.68 [0.13] cm vs females 0.59 [0.11] cm) and on the left (males 0.68 [0.12] cm vs females 0.57 [0.12] cm) reaching statistical significance ( P < .001). The mean height of the thyro-arytenoid muscle in males was larger than that in females on the right (males 1.05 [0.21] cm vs females 0.95 [0.12] cm) and on the left (males 1.05 [0.21] cm vs females 0.95 [0.12] cm) reaching statistical significance ( P < .01 on the right and P < .05 on the left). The volume of the thyroarytenoid muscle in males was larger than that in females on the right (males 0.86 [0.25] mL vs females 0.48 [0.15] mL) and on the left (males 0.89 [0.27] mL vs females 0.48 [0.17] mL) reaching statistical significance ( P < .001). The results of this investigation clearly indicate a significant difference in these measurements between men and women.

Download Full-text

Deep Learning for Locally Advanced Nasopharyngeal Carcinoma Prognostication Based on Pre- and Post-treatment MRI

10.21203/rs.3.rs-103654/v1 ◽

2020 ◽

Author(s):

Song Li ◽

Yu-Qin Deng ◽

Hong-Li Hua ◽

Sheng-Lan Li ◽

Xi-Xiang Chen ◽

...

Keyword(s):

Deep Learning ◽

Nasopharyngeal Carcinoma ◽

Locally Advanced ◽

Tnm Staging ◽

Post Treatment ◽

Mr Images ◽

Mr Image ◽

Heat Maps ◽

Prognosis Prediction ◽

Better Than

Abstract Background: Although it has been reported by several studies that using AI to predict the prognosis of nasopharyngeal carcinoma (NPC) based on magnetic resonance (MR) image, the information around the tumor was not valued and the post-treatment MR images were ignored. Herein we aimed to predict the prognosis of advanced NPC (stage Ⅲ-Ⅳa) using pre- and post-treatment MR images based on deep learning (DL).Methods: A total of 206 patients with primary NPC who were diagnosed and treated at the Renmin Hospital of Wuhan University between June 2012 and January 2018 were retrospectively selected. A rectangular region of interest (ROI), which included the tumor area, surrounding tissues and organs, was delineated on each pre- and post-treatment MR image. Two InceptionResnetV2-based transfer learning models, named pre-model and post-model, were trained with the Pre-dataset and the Post-dataset, respectively. In addition, an ensemble learning model based on the pre-model and post-models was trained. The three established models were evaluated by receiver operating characteristic (ROC) analysis, confusion matrix, and Harrell’s concordance indices (C-index) after the model test. High-risk-related heat maps were developed according to the DL models.Results: The pre-model, post-model, and ensemble models displayed a C-index of 0.717 (95% CI: 0.639 to 0.795), 0.811 (95% CI: 0.745–0.877), 0.830 (95% CI: 0.767–0.893), and AUC of 0.745 (95% CI: 0.592–0.897), 0.820 (95% CI: 0.687–0.953), and 0.841 (95% CI: 0.715–0.968) for the test cohort, respectively. In comparison with the models, the post-model performance was better than the pre-model, which indicated the importance of post-treatment images for prognosis prediction. All three DL models performed better than the TNM staging system. The captured features presented on heat maps showed that the areas around the tumor and lymph nodes were related to the prognosis of the tumor.Conclusions: The three established DL models based on pre- and post-treatment MR images have a better performance than TNM staging. Post-treatment MR images are of great significance for prognosis prediction and could contribute to clinical decision-making.

Download Full-text

Short-term efficacy comparison between helical tomotherapy and intensity-modulated radiotherapy in patients with locally advanced nasopharyngeal carcinoma

10.21203/rs.3.rs-21288/v1 ◽

2020 ◽

Author(s):

Zhen Cui ◽

Jia Liu ◽

Qiaoyu Sun ◽

Chaoge Wang ◽

Meifang Fang ◽

...

Keyword(s):

Nasopharyngeal Carcinoma ◽

Planning Target Volume ◽

Helical Tomotherapy ◽

Locally Advanced ◽

Intensity Modulated Radiotherapy ◽

Target Volume ◽

Short Term ◽

Intensity Modulated ◽

Significant Difference ◽

Better Than

Abstract Background: To evaluate short-term safety and efficacy of helical tomotherapy (HT) versus intensity-modulated radiotherapy (IMRT) in patients with nasopharyngeal carcinoma (NPC).Methods: Retrospective analysis of locally advanced nasopharyngeal carcinoma treated with radiotherapy and concurrent platinum based neoadjuvant chemotherapy (cisplatin 80 mg/m2 every 3 weeks for 1 cycle) in our hospital from February 2017 to October 2019, including 70 patients in HT group and 70 in IMRT group. The target area of the tumor was delineated by magnetic resonance (MRI) imaging. The prescription doses delivered to the gross tumor volume (pGTVnx) and positive lymph nodes (pGTVnd), the high risk planning target volume (PTV1), and the low risk planning target volume (PTV2), were 69.96 Gy, 66-70 Gy, 60 Gy and 50-54 Gy, in 33 fractions, respectively. Acute reactions were evaluated according to the RTOG/EORTC criteria, whereas the therapeutic efficacy was assessed according to RECTST version 1.1 criteria in a 3-months period.Results: The CI of PGTVnx, PGTVnd, PTV1 and PTV2, and HI of PGTVnx, PTV1 and PTV2 in HT group was significantly better than those in IMRT group. The OAR Dmax and Dmean in HT group were better than those in IMRT group with a significant difference (all p <0.05). Patients in the HT group were significantly better than those in the IMRT group in the protection of acute parotid gland injury and hearing damage (p <0.05), but not other acute adverse reaction. No significant difference was found on the short-term efficacy illustrated by ORR between HT group and IMRT group (x2 = 0.119; p = 0.730).Conclusions: Compared with IMRT, HT has better radiophysical-related dosimetric advantages in radiotherapy for locally advanced nasopharyngeal carcinoma. Despite similar on short-term effects, HT has lower incidence of adverse reactions than IMRT.

Download Full-text

Dosimetric evaluation of three commercial radiotherapy planning systems for lung cancer and nasopharyngeal carcinoma cases

10.21203/rs.3.rs-29559/v1 ◽

2020 ◽

Author(s):

Wan-jia Zheng ◽

Ming-li Wang ◽

Jun Zhang ◽

Yi-mei Liu ◽

Li Chen ◽

...

Keyword(s):

Lung Cancer ◽

Spinal Cord ◽

Nasopharyngeal Carcinoma ◽

Treatment Time ◽

Maximum Dose ◽

Organs At Risk ◽

Radiotherapy Planning ◽

Planning Systems ◽

Significant Difference ◽

Better Than

Abstract Purpose: To identify planning systems and techniques suitable for different sites tumors by analyzing dosimetric differences using three commercial radiotherapy planning systems: Tomotherapy, Monaco and Eclipse.Methods: We retrospectively analyzed 20 lung cancer and 8 nasopharyngeal carcinoma (NPC), and each patient plans were designed using the three systems. The dose distribution of the target and organs at risk (OARs) were compared, and monitor unit (MU) and treatment time were also evaluated.Results: For lung cancer, mean dose of PGTV, PTV1 and PTV2 in Monaco and Tomo plans were lower than Eclipse plan. PTV2 CI in Monaco and Eclipse plans were better than Tomo plans (p=0.002, p=0.022). Monaco and Tomo plans were better than Eclipse plan regarding to mean dose and V15Gy of lungs; the lowest lungs V20Gy and V30Gy were provided by Tomo plan. The esophagus, heart and SpinalCord_03 dose were lowest in Monaco plan, and the maximum dose and V45Gy of SpinalCord_03 were 592.1cGy and 1.37% lower than Eclipse plan, respectively. For NPC, mean dose of PGTV, PTV1 and PTV2 in Eclipse plan were superior to Tomo plan (p=0.008, p=0.000, p=0.003); PTV2 V95% in Tomo plan was increased by 1.64% than Eclipse plan. There was no significant difference between Monaco and Eclipse plans. Tomo plan showed better spinal cord and brainstem protection, with spinal cord max dose 249.38cGy lower than Eclipse plan and 555cGy lower than Monaco plan, respectively.Conclusion: Although the three plans reflected their respective advantages in different aspects, in general, the Monaco plan (VMAT) was the best choice for lung cancer, and for the more advanced nasopharyngeal carcinoma, the Tomo plan (HT) was superior to the other two plans.

Download Full-text

Accurate recognition of colorectal cancer with semi-supervised deep learning on pathological images

Nature Communications ◽

10.1038/s41467-021-26643-8 ◽

2021 ◽

Vol 12 (1) ◽

Author(s):

Gang Yu ◽

Kai Sun ◽

Chao Xu ◽

Xing-Hua Shi ◽

Chong Wu ◽

...

Keyword(s):

Artificial Intelligence ◽

Colorectal Cancer ◽

Deep Learning ◽

Supervised Learning ◽

Area Under The Curve ◽

Patient Level ◽

Significant Difference ◽

The Mean ◽

Whole Slide Images ◽

Better Than

AbstractMachine-assisted pathological recognition has been focused on supervised learning (SL) that suffers from a significant annotation bottleneck. We propose a semi-supervised learning (SSL) method based on the mean teacher architecture using 13,111 whole slide images of colorectal cancer from 8803 subjects from 13 independent centers. SSL (~3150 labeled, ~40,950 unlabeled; ~6300 labeled, ~37,800 unlabeled patches) performs significantly better than the SL. No significant difference is found between SSL (~6300 labeled, ~37,800 unlabeled) and SL (~44,100 labeled) at patch-level diagnoses (area under the curve (AUC): 0.980 ± 0.014 vs. 0.987 ± 0.008, P value = 0.134) and patient-level diagnoses (AUC: 0.974 ± 0.013 vs. 0.980 ± 0.010, P value = 0.117), which is close to human pathologists (average AUC: 0.969). The evaluation on 15,000 lung and 294,912 lymph node images also confirm SSL can achieve similar performance as that of SL with massive annotations. SSL dramatically reduces the annotations, which has great potential to effectively build expert-level pathological artificial intelligence platforms in practice.

Download Full-text

Advanced magnetic resonance techniques in early differentiation of pseudoprogression versus progression in the patients with glioblastoma multiforme

Vojnosanitetski pregled ◽

10.2298/vsp170114108m ◽

2019 ◽

Vol 76 (5) ◽

pp. 510-517

Author(s):

Jelena Mihailovic ◽

Marko Dakovic

Keyword(s):

Glioblastoma Multiforme ◽

Magnetic Resonance ◽

Mr Spectroscopy ◽

Progressive Disease ◽

Response To Therapy ◽

Imaging Method ◽

Dynamic Susceptibility ◽

Significant Difference ◽

The Mean ◽

Mean Time

Background/Aim. The diagnosis of glioblastoma multiforme progression may be confounded by a phenomena termed pseudoprogression (PSP) and pseudoresponse (RCT) which has become more common with the adoption of radiation therapy with concurrent and adjuvant application of temozolomide (CRT). Distinguishing of these phenomena is based on the follow-up scans since no single imaging method or technique is yet capable of performing their discrimination. In this study, we evaluated the dynamic susceptibility contrast (DSC perfusion) imaging and magnetic resonance (MR) spectroscopy to predict the prognosis and time to progression in the patients with glioblastoma multiforme. Methods. Fourty patients with primary glioblastoma multiforme were included in the analysis. The patients were examined in 3rd week after surgery and 10th week after the beginning of CRT. The MR exams were performed using the 1.5 T MR scanner (Avanto; Siemens, Erlangen, Germany). The maps of perfusion parameters and time-to-peak (TTP) parameter were calculated using the DPTools v3.79 software. The 3D CSI PRESS MR spectroscopy was performed in the area corresponding to the contrast enhancement on the T1W images. Results. Thirty-two of the 40 patients had progressive disease and 8 had pseudoprogression. Progressive disease showed the mean time of the peak values of 33 ? 7 s in 3rd and 30 ? 5 s in 10th week with no statistical significance between these two periods (p > 0.05). The patients with pseudoprogression showed the mean time of the peak values of 32 ? 8 s in 3rd week and 43 ? 9 s in 10th week; it was statistically significant difference (p < 0.05) was which favors better response to therapy. The spectroscopy results showed presence of glycine peak at 3.56 ppm in 6 patients with progressive disease which was not seen on spectra with pseudoprogression. Conclusion. The observed significant differences in the TTP values for PSP and RCT can provide basis for distinguishing two entities. The presence of glycine peak in the MR spectra could be a marker of RCT.

Download Full-text