Acute xerostomia is the most common side effect of radiation therapy (RT) for head and neck (H&N) malignancies. Investigating radiation-induced changes of computed tomography (CT) radiomics in parotid glands (PGs) and saliva amount (SA) can predict acute xerostomia during the RT for nasopharyngeal cancer (NPC). CT and SA data from 35 patients with stages I-IVB were randomly collected from an NPC clinical trial registered on the clinicaltrials.gov (ID: NCT01762514). All patients received radical treatment based on intensity-modulated RT (IMRT) with a prescription dose of 68.1 Gy in 30 fractions. The patients' ages ranged 24-72 years, and each patient had five CT sets acquired at treatment position: at the 0th, 10th, 20th, 30th fractions during the RT, and at 3-month later after the RT. The PGs for each CT set were delineated by a radiation oncologist and verified independently by another. Patients' saliva was collected every other 10 days during the RT. Acute xerostomia was evaluated based on the RTOG acute toxicity scoring and the SA. In total, 1,703 radiomics features were calculated for PGs from each CT set, including feature value at 0th fraction (FV0F), FV10F, and delta FV (ΔFV10F-0F), respectively. Extensive experiments were conducted to achieve the optimal results. RidgeCV and Recursive Feature Elimination (RFE) were used for feature selection, while linear regression was used for predicting SA30F. Four more patients were added for independent testing. Substantial changes in various radiomics metrics of PGs were observed during the RT. Eight normalized feature value (NFV), selected from NFV0F, predicted SA10F with a mean square error (MSE) of 0.9042 and a R2 score of 0.7406. Fourteen NFV, selected from ΔNFV10F-0F, NFV0F, and NFV10F to predict SA30F, showed the best predictive ability with an MSE of 0.0569. The model predicted the level of acute xerostomia with a precision of 0.9220 and a sensitivity of 100%, compared to the clinical observed SA. For the independent test, the MSE of PSA30F was 0.0233. This study demonstrated that radiation-induced acute xerostomia level could be early predicted based on the SA and radiomics changes of the PGs during IMRT delivery. SA, NFV0F, NFV10F, and especially ΔNFV10F-0F provided the best performance on acute xerostomia prediction for individual patient based on RidgeCV_RFE_LinearRegression method of delta radiomics.

Background: An artificial intelligence (AI)–assisted contouring system benefits radiation oncologists by saving time and improving treatment accuracy. Yet, there is much hope and fear surrounding such technologies, and this fear can manifest as resistance from health care professionals, which can lead to the failure of AI projects. Objective: The objective of this study was to develop and test a model for investigating the factors that drive radiation oncologists’ acceptance of AI contouring technology in a Chinese context. Methods: A model of AI-assisted contouring technology acceptance was developed based on the Unified Theory of Acceptance and Use of Technology (UTAUT) model by adding the variables of perceived risk and resistance that were proposed in this study. The model included 8 constructs with 29 questionnaire items. A total of 307 respondents completed the questionnaires. Structural equation modeling was conducted to evaluate the model’s path effects, significance, and fitness. Results: The overall fitness indices for the model were evaluated and showed that the model was a good fit to the data. Behavioral intention was significantly affected by performance expectancy (β=.155; P=.01), social influence (β=.365; P<.001), and facilitating conditions (β=.459; P<.001). Effort expectancy (β=.055; P=.45), perceived risk (β=−.048; P=.35), and resistance bias (β=−.020; P=.63) did not significantly affect behavioral intention. Conclusions: The physicians’ overall perceptions of an AI-assisted technology for radiation contouring were high. Technology resistance among Chinese radiation oncologists was low and not related to behavioral intention. Not all of the factors in the Venkatesh UTAUT model applied to AI technology adoption among physicians in a Chinese context.

Whether PD-L1/PD-1 expression plays a significant role in the prognosis of NPC is still controversial. The present study mainly aimed to investigate the prognostic significance of PD-L1/PD-1 expression in patients with NPC. A systematical research was performed in the PubMed, Web of Science, EMBASE, and the Cochrane Library databases up to January 06, 2019. Eighteen studies met eligible criteria were included in the meta-analysis. Quality assessment of included articles was evaluated by Newcastle–Ottawa quality assessment scale (NOS). Pooled hazard ratios (HRs) and their corresponding 95% confidence intervals (95% CIs) were used to elucidated the primary endpoint, overall survival (OS), and the secondary endpoints. Furthermore, the relationship between clinicopathological features of NPC and PD-L1/PD-1 expression was estimated by relative ratios (RRs) and 95% CIs. A total of 1836 patients from 15 included studies concerning PD-L1 and 678 patients from six studies regarding PD-1 were included in the meta-analysis. Pooled results revealed that PD-L1 expression in NPC did not correlate with OS (HR 1.34 95% CI 0.93–1.93, p = 0.11), DFS (HR 1.82, 95% CI 0.86–3.85, p = 0.12), PFS (HR 1.19, 95% CI 0.46–3.08, p = 0.72), and DMFS (HR 2.26, 95% CI 0.60–8.56, p = 0.23). Meanwhile, no statistically significant differences existed between the expression level of PD-1 in tumor infiltrating lymphocytes (TILs) and the OS in NPC, with the pooled HR 1.29 (95% CI 0.68–2.42, p = 0.44). In subgroup analysis, higher expression of PD-L1 in immune cells correlated with better OS in patients with NPC, with a pooled HR 0.68 (95% CI 0.47–0.99, p = 0.04). Among the clinicopathological features included in our study, we found that the positive expression of PD-L1 in NPC associated with the higher expression of PD-1 (RR 1.25, 95% CI 1.02–1.52, p = 0.03). Our meta-analysis indicated that higher/positive expression of PD-L1/PD-1 may not serve as suitable biomarkers for the prognosis of NPC, which was not in consistent with some previous studies about the prognostic value of PD-L1/PD-1 in other types of tumors. Despite the positive results in subgroup analysis and study about clinicopathological features, it may still need corroboration of prospective and large-scale studies.

Deformable multi-modal medical image registration aligns the anatomical structures of different modalities to the same coordinate system through a spatial transformation. Due to the difficulties of collecting ground-truth registration labels, existing methods often adopt the unsupervised multi-modal image registration setting. However, it is hard to design satisfactory metrics to measure the similarity of multi-modal images, which heavily limits the multi-modal registration performance. Moreover, due to the contrast difference of the same organ in multi-modal images, it is difficult to extract and fuse the representations of different modal images. To address the above issues, we propose a novel unsupervised multi-modal adversarial registration framework that takes advantage of image-to-image translation to translate the medical image from one modality to another. In this way, we are able to use the well-defined uni-modal metrics to better train the models. Inside our framework, we propose two improvements to promote accurate registration. First, to avoid the translation network learning spatial deformation, we propose a geometry-consistent training scheme to encourage the translation network to learn the modality mapping solely. Second, we propose a novel semi-shared multi-scale registration network that extracts features of multi-modal images effectively and predicts multi-scale registration fields in an coarse-to-fine manner to accurately register the large deformation area. Extensive experiments on brain and pelvic datasets demonstrate the superiority of the proposed method over existing methods, revealing our framework has great potential in clinical application.