Lipomas and atypical lipomatous tumors (ALTs) may appear similar on MRI, but accurate differentiation is critical for effective treatment. Even experienced radiologists can struggle to distinguish between these tumors, often requiring a biopsy to confirm diagnosis. In a previous work, we have already shown that we can use AI to differentiate between these two types of tumors. However, a critical question remains: Will these AI tools work across diverse populations?
There are numerous factors that can impact the effectiveness of an AI model—differences in MRI scanner models, magnetic field strengths, image quality, and, most importantly, patient diversity. It’s essential for AI models to be validated on data from various populations to ensure they can benefit a wide range of patients.
That’s precisely what Douwe Spaanderman and his team have accomplished. They conducted a multi-center validation study, collecting data from six hospitals across the globe. Their AI tool not only proved capable of differentiating between lipomas and ALTs, but the team went a step further. They initiated a prospective study, following patients over time to assess how well the tool predicted real-world outcomes.
Read the full study for insights on the future of AI in medical imaging and its potential to revolutionize tumor diagnostics in diverse patient populations: https://doi.org/10.1016/j.eclinm.2024.102802.