Background: Due to the occult anatomic location of the nasopharynx and frequent presence of adenoid hyperplasia, the positive rate for malignancy identification during biopsy is low, thus leading to delayed or missed diagnosis for nasopharyngeal malignancies upon initial attempt. Here, we aimed to develop an artificial intelligence tool to detect nasopharyngeal malignancies under endoscopic examination based on deep learning. Methods: An endoscopic images‐based nasopharyngeal malignancy detection model (eNPM‐DM) consisting of a fully convolutional network based on the inception architecture was developed and fine‐tuned using separate training and validation sets for both classification and segmentation. Briefly, a total of 28,966 qualified images were collected. Among these images, 27,536 biopsy‐proven images from 7951 individuals obtained from January 1st, 2008, to December 31st, 2016, were split into the training, validation and test sets at a ratio of 7:1:2 using simple randomization. Additionally, 1430 images obtained from January 1st, 2017, to March 31st, 2017, were used as a prospective test set to compare the performance of the established model against oncologist evaluation. The dice similarity coefficient (DSC) was used to evaluate the efficiency of eNPM‐DM in automatic segmentation of malignant area from the background of nasopharyngeal endoscopic images, by comparing automatic segmentation with manual segmentation performed by the experts. Results: All images were histopathologically confirmed, and included 5713 (19.7%) normal control, 19,107 (66.0%) nasopharyngeal carcinoma (NPC), 335 (1.2%) NPC and 3811 (13.2%) benign diseases. The eNPM‐DM attained an overall accuracy of 88.7% (95% confidence interval (CI) 87.8%–89.5%) in detecting malignancies in the test set. In the prospective comparison phase, eNPM‐DM outperformed the experts: the overall accuracy was 88.0% (95% CI 86.1%–89.6%) vs. 80.5% (95% CI 77.0%–84.0%). The eNPM‐DM required less time (40 s vs. 110.0 ± 5.8 min) and exhibited encouraging performance in automatic segmentation of nasopharyngeal malignant area from the background, with an average DSC of 0.78 ± 0.24 and 0.75 ± 0.26 in the test and prospective test sets, respectively. Conclusions: The eNPM‐DM outperformed oncologist evaluation in diagnostic classification of nasopharyngeal mass into benign versus malignant, and realized automatic segmentation of malignant area from the background of nasopharyngeal endoscopic images.

Background: Delineating the lesion area is an important task in image-based diagnosis. Pixel-wise classification is a popular approach to segmenting the region of interest. However, at fuzzy boundaries, such methods usually result in glitches, discontinuity or disconnection, inconsistent with the fact that lesions are solid and smooth. Methods: To overcome these problems and to provide an efficient, accurate, robust and concise solution that simplifies the whole segmentation pipeline in AI-assisted applications, we propose the BézierSeg model which outputs Bézier curves encompassing the region of interest. Results: Directly modeling the contour with analytic equations ensures that the segmentation is connected and continuous, and that the boundary is smooth. In addition, it offers sub-pixel accuracy. Without loss of precision, the Bézier contour can be resampled and overlaid with images of any resolution. Moreover, clinicians can conveniently adjust the curve’s control points to refine the result. Conclusions: Our experiments show that the proposed method runs in real time and achieves accuracy competitive with pixel-wise segmentation models.

Objectives: To benchmark the performance of DeepSeek-R1 against three other advanced AI reasoning models (GPT-4o, Qwen3, Grok-3) in automatically extracting T/N staging from esophageal cancer endoscopic ultrasound (EUS) complex medical reports, and to evaluate the impact of language (Chinese/English) and prompting strategy (with/without designed prompt) on model accuracy and robustness. Methods: We retrospectively analyzed 625 EUS reports for T-staging and 579 for N-staging, which were collected from 663 patients at the Sun Yat-sen University Cancer Center between 2018 and 2020. A 2 × 2 factorial design (Language × Prompt) was employed under a zero-shot setting. The performance of the models was evaluated using accuracy, and the odds ratio (OR) was calculated to quantify the comparative performance advantage between models across different scenarios. Results: Performance was evaluated across four scenarios: (1) Chinese with-prompt, (2) Chinese without-prompt, (3) English with-prompt, and (4) English without-prompt. In both T and N-staging tasks, DeepSeek-R1 demonstrated superior overall performance compared to the competitors. For T-staging, the average accuracy was (DeepSeek-R1 vs. GPT-4o vs. Qwen3 vs. Grok-3: 91.4% vs. 84.2% vs. 89.5% vs. 81.3%). For N-staging, the respective average accuracy was 84.2% vs. 65.0% vs. 68.4% vs. 51.9%. Notably, N-staging proved more challenging than T-staging for all models, as indicated by lower accuracy. This superiority was most pronounced in the Chinese without-prompt T-staging scenario, where DeepSeek-R1 achieved significantly higher accuracy than GPT-4o (OR = 7.84, 95% CI [4.62–13.30], p < 0.001), Qwen3 (OR = 5.00, 95% CI [2.85–8.79], p < 0.001), and Grok-3 (OR = 6.47, 95% CI [4.30–9.74], p < 0.001). Conclusions: This study validates the feasibility and effectiveness of large language models (LLMs) for automated T/N staging from EUS reports. Our findings confirm that DeepSeek-R1 possesses strong intrinsic reasoning capabilities, achieving the most robust performance across diverse conditions, with the most pronounced advantage observed in the challenging English without-prompt N-staging task. By establishing a standardized, objective benchmark, DeepSeek-R1 mitigates inter-observer variability, and its deployment provides a reliable foundation for guiding precise, individualized treatment planning for esophageal cancer patients.