A Deep Learning Model for Diagnosing Lymph Node Metastasis in Nasopharyngeal Carcinoma(NPC)

Официальное название

Development and Validation of a Deep Learning Model for Diagnosing Lymph Node Metastasis in Nasopharyngeal Carcinoma Using Histologic Whole Slide Images and Time-dependent Magnetic Resonance Images

Обзор

(I) AI Model for Diagnosing Lymph Node Metastasis We developed an AI model to help diagnose whether a single lymph node in nasopharyngeal cancer has spread. The model uses MRI images of the lymph node and the area around it. It includes: 1.Automatically identifying the lymph nodes and the primary tumor. 2.Analyzing MRI images of the lymph node and surrounding area. 3.Using MRI scans before and after chemotherapy to track changes in the lymph node. (II) AI Model for Predicting Lymph Node Metastasis We created an AI model that predicts whether a lymph node in a specific area has cancer. This model uses a combination of the primary tumor's pathology and MRI images of both the tumor and lymph node. It also tracks changes in the lymph node over time. The model includes: 1.Analyzing the tumor's pathology to identify specific lymphatic structures. 2.Using MRI scans to predict the likelihood of metastasis in a single lymph node. 3.Examining MRI scans before and after chemotherapy to help determine if the lymph node has metastasized. (III) Verifying and Analyzing the Benefits of the AI Model We are testing the AI model to see how well it works and its potential benefits, including: 1.Checking if the AI can correct past diagnoses of recurrent lymph nodes in nasopharyngeal cancer, which could help guide treatment plans for radiotherapy. 2.Testing the model using biopsy results from head and neck cancer patients to see if it can accurately detect negative lymph nodes. 3.Running clinical trials to test the AI model's safety and effectiveness in guiding radiation treatment for upper neck and single lymph node areas in nasopharyngeal cancer. 4.Analyzing the economic benefits of using the AI model in radiation treatment for nasopharyngeal cancer.

Подробное описание

(I) AI Model for Assisting Diagnosis of Lymph Node Metastasis in Nasopharyngeal Carcinoma Based on MRI Features of the Lymph Node and Surrounding 3mm Area.

An AI model is developed to assist in diagnosing whether a single lymph node in nasopharyngeal carcinoma has metastasized, based on the MRI features of the lymph node itself and its surrounding 3mm area. Specifically, the model includes: 1.Automatic segmentation of the lymph node and primary lesion using a semi-supervised AI model. 2.Construction of a single dual-view AI model based on baseline MRI features of the lymph node itself and its surrounding 3mm area. 3.Development of a dual-time series dual-view AI model based on MRI features of the lymph nodes before and after induction chemotherapy.

(II) Multimodal AI Model for Predicting Lymph Node Metastasis. A multimodal AI model is constructed to predict metastasis of a single station lymph node and diagnose lymph node metastasis "from surface to point." This is based on pathological features of the nasopharyngeal primary lesion, MRI images, lymph node location, and other factors. Specifically, the model includes: 1.Construction of an AI model based on H\&E stained digital pathology of the primary lesion to extract features of the tertiary lymphatic structure. 2.Development of a multimodal AI model integrating pathological features of the nasopharyngeal primary lesion and baseline MRI images of both the primary lesion and lymph node to predict the probability of lymph node metastasis in a single station. 3.Construction of a single or dual-time series multimodal AI model using the probability of lymph node metastasis and baseline MRI or two MRI scans before and after induction chemotherapy to diagnose whether a single lymph node in the station has metastasized.

(III) Verification and Economic Benefit Analysis of AI Models. The AI models are subjected to thorough verification and economic benefit analysis. Specifically, the process includes: 1.Retrospective correction of historical diagnoses of in situ recurrent lymph nodes in nasopharyngeal carcinoma patients using the AI model to verify its potential benefits in guiding prescription doses for single lymph node radiotherapy. 2.Validation of the AI model using pathological results from lymph node dissection in head and neck squamous cell carcinoma patients to assess the detection rate of clinically diagnosed negative lymph nodes. 3.Prospective clinical trials to evaluate the effectiveness and safety of the AI model in guiding prescription doses for upper neck radiotherapy and single lymph node radiotherapy in nasopharyngeal carcinoma patients. 4.Economic benefit analysis to illustrate the economic value of the AI model in guiding upper neck radiotherapy for nasopharyngeal carcinoma.

Первичные конечные точки

• AUC [Срок оценки: through study completion, an average of 2 year]

Вторичные конечные точки (1)

• Sensitivity and Specificity [Срок оценки: through study completion, an average of 2 year]

Критерии участия

Критерии включения

• The primary lesion was pathologically confirmed as nasopharyngeal carcinoma (WHO classification is I, II and III);
• MRI scan was performed at the initial diagnosis (before anti-tumor treatment), and transverse and coronal MRI images before treatment were available, including T1-weighted, T2-weighted and T1-enhanced scanning sequences.
• PET/CT scan was performed at the initial diagnosis (before anti-tumor treatment)
• When MRI and PET/CT were inconsistent in judging the benign or malignant nature of lymph nodes, the patient agreed to undergo cervical lymph node puncture and pathological examination.

Критерии исключения

• The patient has undergone cervical lymph node radiotherapy for any reason
• Combined with other malignant tumors

Критерии приведены из реестра в оригинале (на английском). Окончательную оценку соответствия проводит исследовательский центр.

Здоровые добровольцы: Нет

Дизайн исследования

Модель наблюдения

Когортное

Центры проведения

Китай · 1 центр

• Department of Radiation Oncology, Sun Yat-sen University Cancer Center — Гуанчжоу

Публикации

• Liu Y, Lai F, Lin B, Gu Y, Chen L, Chen G, Xiao H, Luo S, Pang Y, Xiong D, Li B, Peng S, Lv W, Alexander EK, Xiao H. Deep learning to predict cervical lymph node metastasis from intraoperative frozen section of tumour in papillary thyroid carcinoma: a multicentre diagnostic study. EClinicalMedicine. 2023 May 18;60:102007. doi: 10.1016/j.eclinm.2023.102007. eCollection 2023 Jun. PMID 37251623
• Wessels F, Schmitt M, Krieghoff-Henning E, Jutzi T, Worst TS, Waldbillig F, Neuberger M, Maron RC, Steeg M, Gaiser T, Hekler A, Utikal JS, von Kalle C, Frohling S, Michel MS, Nuhn P, Brinker TJ. Deep learning approach to predict lymph node metastasis directly from primary tumour histology in prostate cancer. BJU Int. 2021 Sep;128(3):352-360. doi: 10.1111/bju.15386. Epub 2021 May 5. PMID 33706408
• Chen C, Zhang Y, Wu X, Shen J. The role of tertiary lymphoid structure and B cells in nasopharyngeal carcinoma: Based on bioinformatics and experimental verification. Transl Oncol. 2024 Mar;41:101885. doi: 10.1016/j.tranon.2024.101885. Epub 2024 Jan 31. PMID 38295746
• Wibawa MS, Zhou JY, Wang R, Huang YY, Zhan Z, Chen X, Lv X, Young LS, Rajpoot N. AI-Based Risk Score from Tumour-Infiltrating Lymphocyte Predicts Locoregional-Free Survival in Nasopharyngeal Carcinoma. Cancers (Basel). 2023 Dec 10;15(24):5789. doi: 10.3390/cancers15245789. PMID 38136336
• Haggstrom I, Leithner D, Alven J, Campanella G, Abusamra M, Zhang H, Chhabra S, Beer L, Haug A, Salles G, Raderer M, Staber PB, Becker A, Hricak H, Fuchs TJ, Schoder H, Mayerhoefer ME. Deep learning for [18F]fluorodeoxyglucose-PET-CT classification in patients with lymphoma: a dual-centre retrospective analysis. Lancet Digit Health. 2024 Feb;6(2):e114-e125. doi: 10.1016/S2589-7500(23)00203-0. Epub PMID 38135556
• Zhong Y, Cai C, Chen T, Gui H, Chen C, Deng J, Yang M, Yu B, Song Y, Wang T, Chen Y, Shi H, Xie D, Chen C, She Y. PET/CT-based deep learning grading signature to optimize surgical decisions for clinical stage I invasive lung adenocarcinoma and biologic basis under its prediction: a multicenter study. Eur J Nucl Med Mol Imaging. 2024 Jan;51(2):521-534. doi: 10.1007/s00259-023-06434-7. Epub 2023 Sep PMID 37725128
• Lee JO, Ahn SS, Choi KS, Lee J, Jang J, Park JH, Hwang I, Park CK, Park SH, Chung JW, Choi SH. Added prognostic value of 3D deep learning-derived features from preoperative MRI for adult-type diffuse gliomas. Neuro Oncol. 2024 Mar 4;26(3):571-580. doi: 10.1093/neuonc/noad202. PMID 37855826
• OuYang PY, Zhang BY, Guo JG, Liu JN, Li J, Peng QH, Yang SS, He Y, Liu ZQ, Zhao YN, Li A, Wu YS, Hu XF, Chen C, Han F, You KY, Xie FY. Deep learning-based precise prediction and early detection of radiation-induced temporal lobe injury for nasopharyngeal carcinoma. EClinicalMedicine. 2023 Apr 4;58:101930. doi: 10.1016/j.eclinm.2023.101930. eCollection 2023 Apr. PMID 37090437

Идентификаторы

NCT: NCT06829147 · B2024-769-01

Первоисточники (государственные реестры)

• ClinicalTrials.gov (NIH, США)

Открыть это исследование на ClinicalTrials.gov ↗