AI system can learn the “language of cancer” and improve diagnosis

A groundbreaking computer system that harnesses the power of AI to understand the "language of cancer" could revolutionize cancer diagnosis, according to its developers. This promising system, developed by an international team led by researchers from the University of Glasgow and New York University, promises faster and more accurate diagnoses and predictions of patient outcomes.

How does it work?

Known as histomorphological phenotype learning (HPL), the system was developed by collecting thousands of high-resolution images of tissue samples from lung adenocarcinoma patients stored in the United States National Cancer Institute's Cancer Genome Atlas database. Using a training process called self-supervised deep learning, the algorithm analyzed these images, breaking them down into thousands of tiny tiles representing small amounts of human tissue. A deep neural network then scrutinized these tiles, teaching itself to recognize and classify visual features shared across the cells in each tissue sample.

Dr. Ke Yuan from the University of Glasgow's School of Computing Science, who supervised the research, explained that the algorithm learned to spot recurring visual elements corresponding to textures, cell properties, and tissue architectures called phenotypes.

"By comparing those visual elements across the whole series of images, it recognized phenotypes which often appeared together, independently picking out the architectural patterns that human pathologists had already identified," he said.

When the HPL system was tested on slides from squamous cell lung cancer, it correctly distinguished between their features with 99% accuracy. The system also analyzed links between the classified phenotypes and clinical outcomes, predicting the likelihood and timing of cancer recurrence with 72% accuracy, compared to 64% for human pathologists.

Why does it matter?

This AI-driven system could significantly augment the work of pathologists, offering an unbiased second opinion and enhancing diagnostic accuracy.

Professor John Le Quesne from the University of Glasgow's School of Cancer Sciences highlighted the potential impact: "It takes many years to train human pathologists to identify cancer subtypes and draw conclusions about patient outcomes. This algorithm, which taught itself to recognize complex patterns in cancer slides, could provide a valuable tool to aid pathologists in the future."

By integrating AI analysis with human expertise, the HPL system could lead to faster, more accurate cancer diagnoses and better-tailored patient care.

"The insight provided by human expertise and AI analysis working together could improve monitoring and treatment outcomes," Le Quesne added.

The context

The development of the HPL system comes at a time when the need for efficient and accurate cancer diagnostics is more critical than ever. With cancer being a leading cause of death worldwide, innovations that can enhance diagnostic processes and treatment plans are vital. The research, published in the journal Nature Communications, also saw contributions from researchers at University College London and the Karolinska Institute in Sweden, underscoring the global effort to combat this disease.

In summary, the HPL system represents a significant advancement in cancer diagnostics, offering a powerful tool to complement pathologists' expertise and potentially improve patient outcomes through more precise and timely diagnoses.


💡Did you know?

You can take your DHArab experience to the next level with our Premium Membership.
👉 Click here to learn more