Iranian researchers develop new method to detect cancer from medical images

Iranian researchers have developed a new artificial intelligence-based method that improves the detection of different types of cancer from medical images, offering a promising approach for earlier diagnosis and more accurate clinical decision-making.

The study, supported by Iran's National Science Foundation (INSF), was carried out by postdoctoral researcher Yousef Sharafi at Khajeh Nasir Toosi University of Technology.

According to Sharafi, the research combines artificial intelligence, deep learning and transfer learning to extract the most important features from medical images while filtering out unnecessary data.

The approach is designed to improve the accuracy of identifying different cancers, including breast, brain and blood cancers, he added.

He explained that as cancer remains one of the world's leading causes of death, claiming more than 10 million lives each year, the development of more reliable diagnostic technologies is becoming increasingly important.

"Cancer is one of the most significant global health challenges, and advances in diagnosis and prediction can play an effective role in reducing mortality and improving treatment outcomes," Sharafi said.

The researchers first used advanced image-processing techniques and neural networks to identify and segment cancerous tumors in medical images, Sharafi said.

Once the suspicious regions were isolated, he highlighted, the system extracted key geometric, spatial and textural characteristics of cancer cells.

The extracted information was then refined using deep-learning models and optimization techniques to reduce the amount of data while preserving the most meaningful features, according to Sharafi.

Finally, he said, the images were analyzed using fuzzy neural networks and machine-learning algorithms capable of handling uncertainty, particularly in cases where the boundaries between healthy and cancerous tissue are difficult to distinguish.

The team also developed a two-level deep-learning framework specifically for analyzing breast cancer pathology and mammography images, which further improved the system's classification performance, Sharafi noted.

According to the researchers, the new method consistently outperformed conventional approaches by delivering higher accuracy in classifying cancerous tumors.

Eliminating redundant image features enabled the AI models to focus on the most informative patterns, resulting in more reliable diagnoses, Sharafi said.

"The proposed methods achieved higher accuracy and efficiency than conventional techniques in classifying cancerous tumors," Sharafi said.

"By focusing on key image features, the system improves the performance of intelligent models for cancer detection and classification."

The researchers believe the technology could contribute to the development of more advanced AI-powered medical tools for early cancer detection, disease prediction and clinical monitoring.

Sharafi also stressed that the success of AI in healthcare depends heavily on access to high-quality medical data.

"The effectiveness of artificial intelligence systems depends, above all, on standardized, accurate and well-organized medical data," he said, calling on policymakers to invest in national centers dedicated to collecting, validating and integrating cancer-related datasets.

He added that strengthening data infrastructure and training specialists in medical AI would pave the way for smarter diagnostic systems and ultimately improve healthcare outcomes for cancer patients.

Press TV’s website can also be accessed at the following alternate addresses:

www.presstv.ir