Predicting Tumor Origin Based on Deep Learning of Lymph Node Puncture Cytology

A Multicenter Study on Predicting Tumor Origin Based on Deep Learning of Lymph Node Puncture Cytology

Status

Recruiting

Phases

Unknown

Study type

Observational

Source

ClinicalTrials.gov

Registry ID

NCT06810349

Enrollment

10000

Registered

2025-02-05

Start date

2024-11-11

Completion date

2025-12-31

Last updated

2025-02-05

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

Lymph Nodes With Tumor Metastasis

Keywords

Primary unknown tumor origin, Deep learning, Cytopathology

Brief summary

In this study, the investigators aimed to construct a deep learning diagnostic model that uses cytological images to predict primary unknown tumor origins in patients with tumors combined with lymph node metastases. After the model is constructed, the model will be validated by a large-scale test set to test the model performance. The investigators also propose to compare the performance of the constructed model in diagnosing cytology smears compared to human pathologists.

Interventions

None listed

Study design

Observational model

COHORT

Time perspective

RETROSPECTIVE

Eligibility

Sex/Gender

ALL

Healthy volunteers

Yes

Inclusion criteria

* From West China Hospital of Sichuan University (October 1, 2008-August 31, 2024) with corresponding clinical data, including age, sex, specimen puncture site, pathologic diagnosis, pathologic type, whether immunocytochemistry was added, clinical diagnosis, lesion site, co-morbidities, history of malignancy, treatment modality, occurrence of postoperative complications, total number of days of hospitalization postoperatively, and survival time; * From the Department of Pathology of the First Affiliated Hospital of Zhengzhou University, the Sichuan Provincial Cancer Hospital, and the Cancer Hospital of the Chinese Academy of Medical Sciences (January 1, 2020-August 31, 2024) with corresponding clinical data, including age, sex, specimen puncture site, pathologic diagnosis, pathologic type, whether immunocytochemistry was added, clinical diagnosis, lesion site, co-morbidities, history of malignancy, treatment modality, occurrence of postoperative complications, total number of days of hospitalization postoperatively, and survival time.

Exclusion criteria

* Images lacking any supporting clinical or pathologic evidence to support a primary origin and its corresponding clinical information; * Blank, poorly focused, and low-quality images containing severe artifacts and their corresponding clinical information.

Design outcomes

Primary

Measure	Time frame	Description
Model performance metrics	1 year	Model performance was evaluated by Positive Predictive Value (PPV), Negative Predictive Value (NPV), Accuracy, Sensitivity and Specificity.

Countries

China

Contacts

Primary ContactJianyong Lei

guosiyin2000@163.com02885423822

Outcome results