Improved SegNet with Hybrid Classifier for Lung Cancer Segmentation and Classification

Abstract

Prompt diagnosis is crucial globally to save lives, underscoring the urgent need in light of lung cancer's status as a leading cause of death. While CT scans serve as a primary imaging tool for LC detection, manual analysis is laborious and prone to inaccuracies. Recognizing these challenges, computational techniques, particularly ML and DL algorithms, are being increasingly explored as efficient alternatives to enhance the precise identification of cancerous and noncancerous regions within CT scans, aiming to expedite diagnosis and mitigate errors. The proposed model employs Preprocessing to standardize image features, followed by segmentation using an Improved SegNet framework to delineate cancerous regions. Features like LGXP and MBP are then extracted, facilitating classification with a hybrid classifier which combines LSTM and LinkNet models. Implemented in Python, the model's performance is evaluated against conventional methods, showcasing superior accuracy, sensitivity, and precision. This framework promises to revolutionize LC diagnosis, enabling early intervention and improved patient outcomes.