Spatial Pyramid Pooling With 3D Convolution Improves Lung Cancer Detection.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Lung cancer is a leading cause of cancer mortality worldwide.
• Low-dose computed tomography (CT) screening reduces lung cancer mortality but has a high false positive rate.
• Deep learning offers potential to improve lung cancer screening accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate DeepScreener, a deep learning algorithm for predicting lung cancer status from volumetric lung CT scans.
• To assess the performance of DeepScreener on an independent dataset from the National Lung Screening Trial (NLST).

Main Methods:

• DeepScreener utilizes a Spatial Pyramid Pooling model, previously recognized in the Data Science Bowl 2017 competition.
• The algorithm was tested on 1449 low-dose CT scans from the NLST cohort.
• A combination of Spatial Pyramid Pooling and 3D Convolution was employed to enhance performance.

Main Results:

• DeepScreener demonstrated consistent high accuracy in predicting cancer status.
• The algorithm achieved an Area Under the Curve (AUC) of 0.892 when combining Spatial Pyramid Pooling and 3D Convolution.
• This performance surpassed previous state-of-the-art algorithms that used only 3D convolution.

Conclusions:

• DeepScreener shows significant potential for improving lung cancer detection in low-dose CT screening.
• The integration of advanced deep learning techniques can enhance diagnostic accuracy and potentially reduce false positives.
• Further development of AI algorithms can aid in more effective lung cancer screening strategies.