A Comprehensive Analysis of Implementing Convolutional Neural Networks (CNN) and InceptionV3 for Early-Lung Cancer Detection

This work focuses on the categorization and comparison of early-stage lung cell pictures that are malignant and non-cancerous utilizing Convolutional Neural Networks (CNNs) and Transfer Learning with the InceptionV3 model. Images of lung cells from four different categories— adenocarcinoma, big cell carcinoma, normal, and squamous cell carcinoma—make up the dataset used in this study. This study's main focus is on accurately classifying lung cell pictures into carcinogenic and non-cancerous groups. This is an important step towards the early diagnosis and detection of lung disorders. To improve the features and lower noise, the dataset underwent intensive preprocessing, which included adaptive histogram equalization, noise reduction using Gaussian blur, and histogram equalization. Two deep learning models were used for image classification: The InceptionV3 model and a modified CNN architecture. The models' performance was assessed on distinct validation and test sets after they had been trained on a labelled dataset. The models were evaluated using criteria such F1 score, area under the ROC curve (AUC), accuracy, precision, and recall. The findings show that while both models performed well in terms of accuracy, precision, recall, AUC, and F1 score, the InceptionV3 model performed better. Comprehensive insights into the algorithms' capacity to accurately categorize photos of malignant and non-cancerous lung cells are offered by the confusion matrix and classification report. By showcasing the efficacy of deep learning models in the classification of lung cell pictures, this study advances the field of medical image analysis. The results imply that transfer learning, especially in medical image analysis, can greatly improve classification job accuracy by using pre-trained models such as InceptionV3. According to the data, InceptionV3 has a testing accuracy of 0.9755 compared to CNN's 0.8955, making it a better fit for classifying lung tumors than classic CNN.