Lung Disease Prediction using Multi-Level Atrous Spatial Pyramid Pooling Based Convolutional Neural Network

Lung cancer is a malignant tumor caused by the uncontrolled proliferation of cells, leading to tumor formation.Computed Tomography (CT) scans are commonly used to predict lung diseases like cancer, because of their high-resolution imaging that identifies a tumor and determine its location and size.However, identifying complex patterns precisely is challenging due to the diverse and tenuous nature of abnormalities, which lead to overlap with other conditions.In this research, the MultiLevel Atrous Spatial Pyramid pooling-based Convolutional Neural Network (MLASPP-CNN) is proposed to predict lung diseases accurately.In traditional CNN, the MLASPP is incorporated to capture contextual information across various scales.This enhances the model's ability to detect finegrained details and preserves spatial resolution, which results in robust performance.The median filter and Contrast Limited Adaptive Histogram Equalization (CLAHE) are used in the pre-processing stage to remove noise and enhance contrast in the image.The proposed MLASPP-CNN achieves a high accuracy of 99.81% using IQ-OTH/NCCD compared to existing methods such as Transfer learning of fine-tuned EfficientNet (Lung EffNet) and Optimizing double-layered CNN.