According to Hassan Taher and many other healthcare technology thought leaders, cancer diagnosis ranks quite high among the current growth areas for artificial intelligence (AI) in medicine. In fact, he contends that “the integration of AI in healthcare has reached a pivotal milestone, particularly in the field of oncology.”
Taher points to the recent work of researchers at Johns Hopkins University as evidence of AI’s tremendous promise in cancer diagnosis and treatment. Led by Bloomberg Distinguished Professor of Computational Cognitive Science Alan Yuille, the Johns Hopkins research team has developed synthetic tumors to aid AI models in the detection of early-stage cancer. The overarching final objective of these efforts is one that many oncology researchers share: to develop AI models that can quickly, accurately, and automatically identify cancerous tumors with no need for human help.
The Johns Hopkins research team seeks to address a big problem in the world of oncology technology: an ongoing lack of high-quality data to train AI in the detection of cancer. “This shortage stems from the challenging process of identifying tumors on medical scans, which can be extremely time-consuming, as it often relies on pathology reports and biopsy confirmations,” explains Johns Hopkins Communications Specialist Jaimie Patterson. “For example, there are only about 200 publicly available CT scans with annotated liver tumors-a minuscule amount for training and testing AI models to detect early-stage cancer.”
The solution posed by the Johns Hopkins research team is relatively simple. If real tumors are scarce, why not use machine learning (ML) to train AI to detect artificial tumors? Much like a boxer working out with a punching bag, this AI will then be ready to fight cancer for real when the time is right. Concentrating on computed tomography (CT) scans of synthetic liver tumors, the Johns Hopkins team has used AI platforms and ML processes to generate massive datasets of tremendous research value.
Related: The Role of AI in Transforming Manufacturing Automation
The critical first step toward this achievement was the development of extraordinarily realistic fake tumors. In collaboration with radiologists, Johns Hopkins researchers created extremely realistic synthetic tumors in shapes that reflect the irregular contours of actual tumors. They also added “noise” patterns to simulate the random texture of tumors and chose strategic locations for their synthetic tumors to avoid impact with surrounding blood vessels. They even recreated the tendency of tumors to push on their surroundings, leading to significant changes in their appearance.
By all accounts, the results of these efforts were extremely successful. Jaimie Patterson writes that “the resulting synthetic tumors are hyperrealistic and have passed the Visual Turing Test – that is, even medical professionals usually confuse them with real tumors in a visual examination.” When the research team trained their AI model using these synthetic tumors, the performance outcome was directly comparable to those achieved by AI models that trained on naturally occurring tumors.
“Our method is exciting because, to date, no existing work utilizing synthetic tumors alone has achieved a similar or even comparable performance to AI trained on real tumors,” reports team member Qixin Hu, a visiting researcher from the Huazhong University of Science and Technology. “Furthermore, our method can automatically generate numerous examples of small, or even tiny, synthetic tumors, which has the potential to improve the success rate of AI-powered tumor detection.” The team’s concerted focus on very small tumors is deliberate because identifying tumors of this size is essential for the detection cancer in its earliest stages.
Hassan Taher welcomes the use of artificial tumors to help AI detect real tumors as well as differentiate between benign and malignant tissues. “The AI algorithms analyze vast amounts of data from the artificial tumors,” he writes, “learning to recognize patterns and anomalies that may indicate the presence of cancer.”
A respected technology writer and AI thought leader, Hassan Taher is perhaps best known as the author of the influential books The Future of Work in an AI-Powered World, The Rise of Intelligent Machines, and AI and Ethics: Navigating the Moral Maze. As the head of the consulting firm Taher AI Solutions, he advises clients in industries that range from finance to manufacturing. He singles out the growing use of AI in medicine as the tech trend that excites him the most, recognizing its ability to revolutionize patient care and make a tremendous impact on people’s lives.
When it comes to training healthcare AI to identify artificial tumors, Hassan Taher sees uses that go far beyond those tested by the Johns Hopkins research team, stating bluntly that “the real-world applications of this technology are immense.” In fact, numerous case studies and research trials have demonstrated AI’s unique ability to accurately detect early-stage cancer, often outperforming traditional diagnostic modalities by a considerable measure.
“One notable application is in the screening of breast cancer,” writes Taher. “Traditional mammography, while effective, has limitations in sensitivity and specificity. AI systems, however, can enhance the screening process by providing more accurate readings and reducing false positives and negatives. This leads to more reliable diagnoses and fewer unnecessary procedures for patients.”
Hassan Taher goes on to discuss oncology applications for AI that go beyond diagnostics to impact treatment in meaningful ways. For example, he sees a prominent place for AI in the development of personalized patient treatment plans. “By analyzing data from artificial tumors, AI can predict how a particular cancer might respond to various treatments,” he writes. “This enables oncologists to tailor treatment plans to the individual needs of each patient, improving the overall effectiveness of the therapy.”
In terms of improving patient outcomes, boosting survival rates, and reducing the overall burden of cancer on the global population, the meeting of AI and artificial tumors marks a “new era,” according to Taher. “Looking ahead, the future of AI in cancer detection is promising,” he writes. “Continued advancements in AI and biomedical engineering will likely lead to even more sophisticated artificial tumors and more powerful AI systems. As these technologies evolve, they will become integral parts of routine cancer screening and treatment, transforming the landscape of oncology.”
While the recent achievements of Johns Hopkins researchers might seem a relatively minor advancement in the seemly endless war against cancer, Hassan Taher recognizes it as a landmark achievement worth celebrating. “The journey from artificial tumors to real results is a testament to the innovative spirit of modern science and technology,” he writes. “As we continue to explore and harness the potential of AI, we are not only pushing the boundaries of what is possible but also paving the way for a healthier, cancer-free future.”
Keep Reading: Hassan Taher Reflects on AI’s Future and Ethical Challenges
