Search

Background: Pancreatic cancer fine delineation in medical images by physicians is a major challenge due to the vast volume of medical images and the variability of patients. Purpose: A semi‐automatic fine delineation scheme was designed to assist doctors in accurately and quickly delineating the cancer target region to improve the delineation accuracy of pancreatic cancer in computed tomography (CT) images and effectively reduce the workload of doctors. Methods: A target delineation scheme in image blocks was also designed to provide more information for the deep learning delineation model. The start and end slices of the image block were manually delineated by physicians, and the cancer in the middle slices were accurately segmented using a three‐dimensional Res U‐Net model. Specifically, the input of the network is the CT image of the image block and the delineation of the cancer in the start and end slices, while the output of the network is the cancer area in the middle slices of the image block. Meanwhile, the model performance of pancreatic cancer delineation and the workload of doctors in different image block sizes were studied. Results: We used 37 3D CT volumes for training, 11 volumes for validating and 11 volumes for testing. The influence of different image block sizes on doctors' workload was compared quantitatively. Experimental results showed that the physician's workload was minimal when the image block size was 5, and all cancer could be accurately delineated. The Dice similarity coefficient was 0.894 ± 0.029, the 95% Hausdorff distance was 3.465 ± 0.710 mm, the normalized surface Dice was 0.969 ± 0.019. By completing the accurate delineation of all the CT images, the speed of the new method is 2.16 times faster than that of manual sketching. Conclusion: Our proposed 3D semi‐automatic delineative method based on the idea of block prediction could accurately delineate CT images of pancreatic cancer and effectively deal with the challenges of class imbalance, background distractions, and non‐rigid geometrical features. This study had a significant advantage in reducing doctors' workload, and was expected to help doctors improve their work efficiency in clinical application. [ABSTRACT FROM AUTHOR]

Fast and accurate lesion localization is an important step in medical image analysis. The current supervised deep learning methods have obvious limitations in the application of radiology, as they require a large number of manually annotated images. In response to the above issues, we introduced a deep reinforcement learning (DRL)-based method to locate nasopharyngeal carcinoma lesions in 3D-CT scans. The proposed method uses prior knowledge to guide the agent to reasonably reduce the search space and promote the convergence rate of the model. Furthermore, the multi-scale processing technique is also used to promote the localization of small objects. We trained the proposed model with 3D-CT scans of 50 patients and evaluated it with 3D-CT scans of 30 patients. The experimental results showed that the proposed model has strong robustness, and its accuracy was improved by more than 1 mm on average under the premise of using a smaller dataset compared with the DQN models in recent studies. The proposed model could effectively locate the lesion area of nasopharyngeal carcinoma in 3D-CT scans. [ABSTRACT FROM AUTHOR]

Background: Residual lymphadenopathy and detectable Epstein–Barr virus (EBV) DNA after radiotherapy (RT) are known negative prognostic factors for nasopharyngeal carcinoma (NPC). However, there is a need to distinguish between patients with residual disease that will metastasize and those who will not. Purpose: To develop a prognostic model to improve the risk stratification of NPC patients after RT. Study Type: Retrospective. Population: Three hundred eighty‐seven NPC patients treated with RT between January 2010 and January 2013. Field Strength/Sequence: T1‐, T2‐weighted and enhanced T1‐weighted imaging at 1.5 or 3.0 T pretreatment and 3–4 months post‐RT. Assessment: Post‐RT central nodal necrosis (CNN) and other nodal characteristics on MRI were assessed by three radiologists independently. EBV DNA was measured by quantitative polymerase chain reaction. The association between these variables and the primary endpoint (5‐year distant metastasis‐free survival [DMFS], time from the day of diagnosis to any distant metastasis) was analyzed. Nomograms A (pre‐/posttreatment EBV‐DNA + N stage + post‐RT retropharyngeal lymph node [RLN] CNN), B (tumor‐node‐metastasis [TNM] stage + pretreatment EBV‐DNA), and C (TNM stage + post‐RT EBV‐DNA) were developed. Statistical Tests: Univariate and multivariate analyses were performed with the Cox regression model. Nomograms were developed based on the Cox regression model and two prognostic models. The concordance index (C‐index) and calibration curve were used to evaluate the discriminative ability of the nomograms and TNM stage. P‐values < 0.05 were considered statistically significant. Results: Post‐RT RLN CNN was an independent prognostic factor for 5‐year DMFS (hazard ratio, 2.88 [1.48–5.62]). Nomogram A (C‐index 0.728 [0.660–0.797]) demonstrated better risk discrimination than nomogram B (0.638 [0.571–0.705]), nomogram C (0.707 [0.636–0.778]), and the TNM stage (0.587 [0.515–0.659]) for 5‐year DMFS in NPC. Data Conclusion: Nomogram A combining pretreatment EBV‐DNA and N stage with post‐RT EBV‐DNA and RLN CNN improved the prognostic risk stratification for DMFS in NPC. Evidence Level: 4 Technical Efficacy: Stage 2 [ABSTRACT FROM AUTHOR]

Background: Metastatic lymph nodal number (LNN) is associated with the survival of nasopharyngeal carcinoma (NPC); however, counting multiple nodes is cumbersome. Purpose: To explore LNN threshold and evaluate its use in risk stratification and induction chemotherapy (IC) indication. Study Type: Retrospective. Population: A total of 792 radiotherapy‐treated NPC patients (N classification: N0 182, N1 438, N2 113, N3 59; training group: 396, validation group: 396; receiving IC: 390). Field Strength/Sequence: T1‐, T2‐ and postcontrast T1‐weighted fast spin echo MRI at 1.5 or 3.0 T. Assessment: Nomogram with (model B) or without (model A) LNN was constructed to evaluate the 5‐year overall (OS), distant metastasis‐free (DMFS), and progression‐free survival (PFS) for the group as a whole and N1 stage subgroup. High‐ and low‐risk groups were divided (above vs below LNN‐ or model B‐threshold); their response to IC was evaluated among advanced patients in stage III/IV. Statistical Tests: Maximally selected rank, univariate and multivariable Cox analysis identified the optimal LNN threshold and other variables. Harrell's concordance index (C‐index) and 2‐fold cross‐validation evaluated discriminative ability of models. Matched‐pair analysis compared survival outcomes of adding IC or not. A P value < 0.05 was considered statistically significant. Results: Median follow‐up duration was 62.1 months. LNN ≥ 4 was independently associated with decreased 5‐year DMFS, OS, and PFS in entire patients or N1 subgroup. Compared to model A, model B (adding LNN, LNN ≥ 4 vs <4) presented superior C‐indexes in the training (0.755 vs 0.727) and validation groups (0.676 vs 0.642) for discriminating DMFS. High‐risk patients benefited from IC with improved post‐IC response and OS, but low‐risk patients did not (P = 0.785 and 0.690, respectively). Conclusions: LNN ≥ 4 is an independent risk stratification factor of worse survival in entire or N1 staging NPC patients. LNN ≥ 4 or the associated nomogram has potential to identify high‐risk patients requiring IC. Evidence Level: 4 Technical Efficacy: 4 [ABSTRACT FROM AUTHOR]