Classification of Mouse Lung Metastatic Tumor with Deep Learning

Area of Science:

• Computational pathology
• Digital pathology
• Artificial intelligence in medicine

Background:

• Pathological examination of tissue sections for lesion detection is labor-intensive and subject to inter-observer variability.
• Advancements in computer vision and deep learning offer potential for automating the analysis of microscopic medical images.
• There is a need for improved diagnostic tools to assist pathologists in detecting pathological lesions.

Purpose of the Study:

• To develop and evaluate a deep learning model for the automated detection of metastatic tumors in mouse lung tissue using whole slide imaging (WSI).
• To assess the accuracy of the deep learning model in distinguishing between tumor and normal lung tissue.
• To demonstrate the potential of automated analysis for improving the efficiency and accuracy of tissue evaluation.

Main Methods:

• Utilized the Inception-v3 deep learning architecture for image classification.
• Applied the model to whole slide images (WSI) of mouse lung tissue, cropping images to 151x151 pixels.
• Divided the dataset into training (53.8%) and testing (46.2%) sets, comprising 21,017 and 18,016 images, respectively.

Main Results:

• The deep learning model achieved a 98.76% accuracy in detecting lung tissue containing metastatic tumors.
• The model demonstrated a 99.87% accuracy in identifying normal lung tissue (no tumor).
• The model effectively differentiated between metastatic lesions and normal lung tissue.

Conclusions:

• The developed deep learning model accurately detects metastatic tumors in mouse lung tissue using WSI.
• This automated approach shows promise for rapid and accurate analysis of various tissue types, potentially reducing pathologist workload.
• The findings support the integration of deep learning into digital pathology workflows for enhanced diagnostic capabilities.