Peter Adeyemo Adepoju, a Master’s student at the University of Salford, has created an artificial intelligence model that could significantly improve how lung cancer is detected.
His research project led to the development of LungCNET, a deep convolutional neural network designed specifically to identify lung cancer with exceptional accuracy.
Lung cancer remains the deadliest cancer globally, responsible for about 1.8 million deaths each year. Early diagnosis plays a major role in survival, yet traditional methods such as CT scans and biopsies can be influenced by differences in human interpretation.
These inconsistencies sometimes result in delayed or incorrect diagnoses.
LungCNET aims to reduce those gaps. The system was built from scratch to classify lung tissues into three categories: benign, malignant, and normal.
Unlike many AI systems adapted from general image-recognition frameworks, this model was engineered solely for lung cancer detection.
It recorded an impressive 99.09% accuracy rate, setting a new performance standard in radiology-focused AI tools.
Peter explained that his goal was to move beyond theory and create something practical for hospitals and clinics.
According to him, late diagnosis remains one of the biggest obstacles in lung cancer treatment. He believes artificial intelligence can support earlier detection and promote consistent diagnostic decisions.
To measure its performance, LungCNET was compared with well-known AI models such as ResNet50, VGG16, and InceptionV3, which are widely used in medical image analysis.
The comparison showed that LungCNET delivered superior results. In detecting malignant tumors, it achieved 100% precision, recall, and F1-score, meaning no cancer cases were incorrectly classified during testing. For benign tumors, which are often harder to distinguish due to similar structural features, the model achieved a 94% F1-score, outperforming existing alternatives.
Peter’s motivation stems from observing unequal access to advanced diagnostic services, especially in regions with limited medical resources.
Many hospitals lack specialist radiologists and high-end diagnostic systems. He hopes that a reliable AI-powered tool can help reduce diagnostic errors, ease pressure on medical professionals, and improve patient care.
One concern surrounding AI in healthcare is the heavy computational demand of deep learning systems. Many require powerful hardware, which can limit adoption in smaller medical facilities.
LungCNET addresses this challenge through an efficient design that balances speed and accuracy. It processes a single image in 55.76 milliseconds and runs effectively on standard medical equipment, making it suitable for broader clinical use.
Peter emphasized that accessibility was as important as performance. For him, innovation only matters if it can be applied where it is most needed.
Although his thesis has already delivered promising outcomes, he plans further improvements. His next focus includes expanding the dataset to enhance reliability across diverse patient groups, collaborating with hospitals for clinical trials, and refining the system for use in resource-constrained healthcare environments.
Researchers and industry professionals have shown interest in the model’s broader potential. Its adaptable structure could support the detection of other cancers and respiratory conditions, extending its usefulness beyond lung cancer.
For many students, a Master’s thesis marks the completion of an academic requirement. For Peter Adeyemo Adepoju, it represents a practical solution with life-saving potential. His work in artificial intelligence and medical imaging stands among the most encouraging developments in lung cancer screening.
LungCNET’s strong accuracy, speed, and efficiency position it as a valuable tool for hospitals and cancer screening centres worldwide. Faster and more reliable diagnosis can significantly improve treatment outcomes.
Peter’s message is simple: if artificial intelligence helps detect lung cancer earlier and saves even one life, then the effort has truly mattered.
