Related Experiment Video
Updated: Jul 1, 2026
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
An automatic detection model for spread through air spaces in postoperative pathological sections based on deep
Yu Zhang1, Shuzhe Deng2, Mingyuan Guo1
1Department of Medical Oncology, Harbin Medical University Cancer Hospital, Harbin, Heilongjiang, China.
Computer Assisted Surgery (Abingdon, England)
|May 22, 2026
Summary
An artificial intelligence framework accurately detects Spread Through Air Spaces (STAS) in non-small cell lung cancer (NSCLC) pathology slides. This AI tool aids pathologists in diagnosing STAS, an aggressive cancer pattern linked to poorer survival.
Area of Science:
- Pathology
- Artificial Intelligence
- Oncology
Background:
- Spread Through Air Spaces (STAS) is an aggressive lung cancer invasion pattern linked to poor survival.
- STAS is often overlooked or misdiagnosed in routine pathological diagnoses.
- Accurate STAS detection is crucial for non-small cell lung cancer (NSCLC) patient outcomes.
Purpose of the Study:
- Develop and validate an AI framework for automated STAS detection and quantification in NSCLC.
- Improve the accuracy and efficiency of STAS diagnosis in digital whole-slide images.
- Assess the correlation between AI-derived STAS counts and patient survival outcomes.
Main Methods:
- An AI framework combining tumor region segmentation and object detection was developed.
- The framework processed 129 pathological slides from 91 STAS-positive NSCLC patients.
- Performance was evaluated using Jaccard similarity, precision, recall, AP, and F1 score, with external validation and inter-rater reliability analysis.
Main Results:
- The AI framework achieved high performance in segmenting tumor regions (Jaccard similarity = 0.846) and detecting STAS (AP = 0.784).
- The model demonstrated superior performance compared to other object detection methods, meeting diagnostic requirements.
- AI-derived STAS counts showed substantial agreement with pathologists (ICC = 0.703), and STAS events were significantly associated with disease-free survival in stage I lung adenocarcinoma (LUAD).
Conclusions:
- A deep learning-based AI framework enables automated STAS detection and quantification in NSCLC.
- The AI model shows potential to assist pathologists in achieving more accurate and comprehensive STAS diagnoses.
- Automated STAS analysis can improve prognostic accuracy and guide clinical decision-making for NSCLC patients.
