Deep Learning Techniques for Early Diagnosis of Lung Tumours Using Medical Images

This study explores the effectiveness of different deep learning models for the detection of lung cancer using CT scan images. Four models were evaluated: a custom convolutional neural network (CNN), DenseNet121, ResNet152, and VGG19. The primary objective was to assess the models based on key performance metrics such as accuracy, precision, recall, F1-score, and the area under the receiver operating characteristic (AUC-ROC) curve. Among these models, the custom CNN emerged as the top performer, surpassing the others in terms of both sensitivity and specificity. It achieved the highest number of True Positives (TP), correctly identifying the largest number of cancerous cases, and had the lowest False Negatives (FN), minimizing the risk of missed diagnoses. The custom CNN also demonstrated fewer False Positives (FP) compared to the other models, ensuring that healthy individuals were less likely to be misclassified. The model’s strong performance was reflected in its precision (0.92), recall (0.90), and F1-score (0.91), which highlighted its ability to strike an optimal balance between correctly detecting cancer cases and minimizing misclassifications. Overall, the study concludes that the custom CNN model is the most reliable and efficient for lung cancer detection, providing significant promise for clinical applications in medical imaging and early cancer diagnosis.