Related Experiment Video
Updated: Sep 7, 2025
08:54
Pathological Analysis of Lung Metastasis Following Lateral Tail-Vein Injection of Tumor Cells
Published on: May 20, 2020
9.2K
Machine learning-based gene alteration prediction model for primary lung cancer using cytologic images
Shuhei Ishii1,2, Manabu Takamatsu1,3, Hironori Ninomiya1,3
1Department of Pathology, Cancer Institute Hospital, Japanese Foundation for Cancer Research, Tokyo, Japan.
Cancer Cytopathology
|June 20, 2022
Summary
This study developed a machine learning model to predict gene alterations in lung cancer cytology images, aiding faster treatment decisions. The model accurately identifies key genetic mutations from microscopic images, improving diagnostic efficiency.
Area of Science:
- Oncology
- Computational Pathology
- Genomics
Background:
- Accurate gene alteration status is crucial for lung cancer treatment but conventional testing is slow and expensive.
- Predictive models can help identify potential therapeutic targets early in clinical practice.
Purpose of the Study:
- To develop a cytologic image-based gene alteration prediction model for primary lung cancer using machine learning.
- To assess the model's accuracy in classifying gene alterations (ALK, EGFR, KRAS) from cytologic specimens.
Main Methods:
- Retrospective study involving photomicroscopic images of lung cytology samples.
- Development of two neural network models: one for cancer-positive image selection and another for classifying gene alterations.
- Utilized training and validation datasets for model development and testing.
Main Results:
- High accuracy (0.945) and precision (0.991) in selecting cancer-positive image patches.
- Predictive accuracy for EGFR and KRAS mutations was approximately 0.95.
- The model demonstrated good predictive performance for ALK fusions and other mutations, with correct gene status prediction in high-probability cases.
Conclusions:
- A machine learning model based on cytologic images was successfully developed for predicting gene alterations in lung cancer.
- This approach offers a promising, potentially faster, and cost-effective method for guiding lung cancer treatment strategies.
