WinSPIRE: Science in the NewsWhen 69-year-old Dianne Covey went to the doctor with a persistent cough, a standard chest X-ray was taken. To the human eye, early-stage lung abnormalities can be incredibly difficult to diagnose. However, her scan was analyzed by an Artificial Intelligence (AI) tool called Annalise.ai, which flagged a microscopic area of concern. The tool caught her lung cancer at Stage 1. “I never really understood much about artificial intelligence,” Covey later shared “but now I think that it might have saved my life”. While most people associate AI with brainstorming ideas, answering everyday questions, or creating digital images, its most medically revolutionary role is unfolding in laboratories and hospitals. By analyzing complex cellular data and clinical images, AI allows medical researchers and doctors to diagnose deadly diseases and monitor their progression faster than ever before. However, while this innovation is beneficial to modern cancer diagnosis and tracking, it is not used without limitations, including bias in the samples used to train these systems.
Revolutionizing Cancer Detection Through Cellular Imaging
One of the most critical applications of AI in modern medicine is its integration into cellular imaging for disease diagnosis. Professor Kevin Tsia, program director of the Biomedical Engineering program at the University of Hong Kong, is one of the primary innovators of this technology. Professor Tsia developed an AI-driven imaging tool specifically engineered to speed up and enhance the analysis of cellular data in cancer patients. In traditional pathology, manually examining extensive tissue samples to find patterns and markers is an exhausting task. Tsia’s AI imaging tool alleviates this burden by correcting cell imaging inconsistencies, automatically enhancing cell images, and extracting microscopic, previously undetectable information. By highlighting these deeply hidden cellular abnormalities, the technology gives doctors a much higher success rate in identifying early signs of cancer.
Accelerating Disease Tracking via Bodily Scans
Beyond cellular data, AI is reshaping how physicians monitor disease progression through advanced diagnostic imaging. Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) scans create highly detailed, 2D images of a patient’s internal anatomy. Analyzing these scans to monitor disease progression is extremely time consuming and requires a highly trained data analyst. To solve this issue, the medical AI software company Thirona designed an advanced software package that automates scan evaluations, cutting hours off the traditional workflow. Although originally created to identify lung abnormalities related to Cystic Fibrosis (CF)—such as abnormal airways and collapsed lung tissue—as noted by Imaging Technology News, Thirona is actively expanding this software to track other chronic respiratory diseases like asthma. This new use of AI could improve patient care by providing faster, more accurate, and detailed analyses of medical scans to monitor and diagnose disease progression.
These new uses of AI are not used without limitations. Studies at Harvard University have proven that AI analyzes samples with biases related to self-reported gender, age, and race. Harvard claimed that there may be many explanations for the biased results. In the end they concluded that the models are trained on unequal sample sizes, making it harder to diagnose certain cancer types in underrepresented minorities. All hope is not lost for AI cancer detection, however–Harvard professor Kun-Hsing Yu is determined to find a solution for this problem. He states that “there’s hope that if we are more aware of and careful about how we design AI systems, we can build models that perform well in every population.” All in all, AI cancer detection is a fairly new system that’s still in development. With better manufacturing and system training, it can become more accurate and improve health outcomes across demographics.
Written by: Karla Caracheo Gonzalez and Mia McGonnel Jasionowski
Edited by: Katie Holmes, Hazel Milla, and Lauren Griffith
