Swedish scientists have created an AI model that can detect the transition to a more severe form of multiple sclerosis up to three years earlier than traditional methods—offering new hope for timely, life-altering treatment adjustments.
Key Points at a Glance
- New AI model detects MS progression with 90% accuracy
- Can identify transition to secondary progressive MS up to 3 years earlier
- Based on data from over 22,000 Swedish patients
- AI provides confidence levels for each diagnosis
- Open-access version available for researchers worldwide
For the over 22,000 people living with multiple sclerosis (MS) in Sweden, a silent shift in the disease’s nature can drastically alter how it should be treated. Yet today, that transition—when MS evolves from the relapsing-remitting type to the more debilitating secondary progressive form—often goes unnoticed until it’s too late to act effectively. But that may soon change, thanks to an artificial intelligence model developed by researchers at Uppsala University.
This groundbreaking tool can detect, with around 90% accuracy, whether a patient has transitioned to secondary progressive MS (SPMS)—and do so up to three years earlier than traditional diagnostics. That time advantage could mean the difference between prolonged stability and avoidable deterioration.
MS is a chronic condition that attacks the central nervous system, often starting in a pattern of flare-ups followed by recovery (known as relapsing-remitting MS, or RRMS). Over time, however, many patients’ symptoms stop fluctuating and begin to progress steadily. This marks the shift to SPMS, which not only changes how the disease behaves but also demands a completely different treatment approach. Alarmingly, most patients are diagnosed with SPMS long after the change has occurred, potentially wasting precious years on medications that are no longer effective.
The team at Uppsala tackled this diagnostic lag with a data-rich approach, feeding their AI model information from the Swedish MS Registry—a treasure trove of clinical data spanning neurological tests, MRI scans, and treatment histories of tens of thousands of patients. By learning patterns hidden within this massive dataset, the model can determine whether a patient is still in the relapsing phase or has transitioned to progressive MS.
What sets this AI apart is not only its high accuracy but also its transparency. Unlike many black-box algorithms, this model openly provides its confidence level with every diagnosis. “This means that the doctor will know how reliable the conclusion is,” explains Kim Kultima, lead researcher on the project. In other words, the AI doesn’t just make a prediction—it tells clinicians how sure it is, helping them make more informed decisions.
During testing, the AI correctly identified the transition to SPMS earlier than medical records in nearly 87% of the cases. Such early warning could give doctors a crucial head start in adjusting therapies, potentially slowing the disease’s progression and improving quality of life. Patients would no longer have to wait for symptoms to worsen visibly before switching to more effective treatments.
Beyond diagnostics, the implications extend into clinical research. By identifying patients earlier, the AI could help streamline clinical trials, ensuring that only the most appropriate participants are enrolled—improving both the efficiency and effectiveness of MS treatment development.
In a nod to scientific openness, the Uppsala team has made an anonymised version of the model publicly accessible via the web platform msp-tracker. This allows researchers worldwide to test and refine the model further, paving the way for even greater accuracy and broader applications.
Ultimately, this innovation exemplifies how AI is reshaping medicine not just through speed or efficiency, but through deeply human gains—earlier diagnoses, better treatment, and lives improved. For MS patients, that transformation can’t come soon enough.
Source: Uppsala University
