Convolutional neural network application for automated lung cancer detection on chest CT using Google AI Studio

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Oncology

Background:

• Lung cancer diagnosis relies heavily on computed tomography (CT) imaging.
• Artificial intelligence (AI) models are being developed to enhance diagnostic accuracy in medical imaging.
• Evaluating AI performance in detecting lung cancer is crucial for clinical integration.

Purpose of the Study:

• To assess the diagnostic performance of an AI-enhanced model for lung cancer detection on chest CT images.
• To evaluate the AI model's accuracy, sensitivity, and specificity in classifying normal, benign, and malignant lung conditions.
• To analyze the interpretative consistency and behavior of the AI model.

Main Methods:

• Utilized the publicly available IQ-OTH/NCCD dataset (110 CT cases).
• Fine-tuned a pre-trained convolutional neural network (CNN) using 25 training images.
• Performed quantitative evaluation of diagnostic accuracy and qualitative analysis of AI-generated reports.

Main Results:

• The AI model achieved 75.5% overall accuracy, 74.5% sensitivity, and 76.4% specificity.
• Area Under the ROC Curve (AUC) was 0.824, indicating strong discriminative power.
• Malignant cases showed high performance (AUC=0.902), while benign cases were more challenging (AUC=0.615); AI exhibited oversensitivity to ground-glass opacities.

Conclusions:

• The AI model demonstrates promising potential for detecting lung malignancies on CT scans.
• Limitations in specificity and interpretative errors in non-malignant cases necessitate human oversight.
• AI-enhanced CT interpretation can improve efficiency but should function as a decision-support tool, not a replacement for expert review.