Dual-perspective AI model achieves high accuracy in early diagnosis of lung cancer

Lung cancer remains the leading cause of cancer-related deaths worldwide, accounting for nearly one in five cancer deaths and claiming approximately 1.8 million lives each year. One of the main reasons is the delay in diagnosis. In its early stages, the disease appears as very small nodules that even experienced radiologists have difficulty distinguishing from healthy tissue.

For physicians, this means always maintaining a balance between what is seen and what may be missed. Subtle differences in scans can also determine whether cancer is caught early or missed entirely.
Researchers are now exploring how artificial intelligence (AI) can help solve this challenge by providing doctors with a more reliable way to analyze complex medical images.

See both details and context at the same time

To improve lung cancer detection, researchers have developed a system that learns to analyze computed tomography (CT) scans in a way that closely resembles the way doctors work, but without the need to switch perspectives.

One part of the model focuses on small details, such as small specks and textures in the lungs, while another part looks at the entire image to understand the larger context. ”

Inzamam Mashood Nasir, Kaunas University of Technology (KTU) researcher and system developer

This dual approach addresses an important limitation of existing systems, which often capture only part of the information (either the details or the overall structure), but not both at the same time.

In practice, radiologists constantly switch between these two views, zooming in on suspicious areas and then returning to understand how they relate to the whole lung. However, the AI ​​system performs both tasks simultaneously.

“You can think of it as a magnifying glass and a complete view of the scan at the same time,” Nasir explains.

The model was trained using CT scans of both healthy individuals and cancer patients and learned to recognize patterns that distinguish between normal, benign, and malignant cases.

The results show clear performance improvements. The system achieved over 96% accuracy, outperforming existing approaches and maintaining stable performance across a variety of tests. “This level of progress is important, especially in medical applications where even small differences can have significant consequences,” said the KTU PhD student.

Applicable to diseases other than lung cancer, such as brain tumors and breast cancer

In clinical practice, this system could change the way lung cancer is diagnosed.

“This is meant to support clinicians. The system provides a second opinion, helps ensure that important details are not overlooked, and reduces the time required per patient, especially in high-load environments”, emphasizes the KTU researchers.

For patients, the impact is even more significant. Lung cancer is often detected late because treatment options are limited. Early detection dramatically improves survival rates. “Early diagnosis means treatment can start sooner, and outcomes are generally much better,” Nasir says.

This system is designed to improve both sides of the problem. This means fewer cases are missed, while also reducing the number of false alarms that lead to unnecessary stress and procedures.

However, the researchers note that the current model was trained on a relatively limited dataset and needs to be tested on a larger and more diverse group of patients. “In a real-world situation, there are many variables, such as different scanners, imaging protocols, and patient populations, and we need to ensure that the system works reliably for all of them,” Nasir explains.

Future steps include clinical validation, testing in hospital settings, and integration into existing healthcare systems.

In the future, the same approach may be applied beyond lung cancer. “Medical imaging tasks that require both detailed analysis and big-picture understanding could benefit from this type of model,” Nasir said, pointing to areas such as brain tumors, breast cancer, and eye diseases.