Can a spinal cord MRI tell SPG4 from SPG5?

A plain-language summary of a published scientific study. It is educational and does not replace medical advice.

The study: Huang Z et al., Neuroradiology, published online 8 June 2026[1]. A prospective study. Only the abstract is freely available.

What the researchers studied

SPG4 and SPG5 are two types of hereditary spastic paraplegia. The authors write that telling them apart matters because treatment strategies differ, and genetic testing is not easy to obtain everywhere[1]. So they tried another route: radiomics, which extracts hundreds of measurements (shape, texture) from MRI images that machine-learning algorithms then analyse[1].

40 patients with SPG4 and 30 with SPG5 took part. 56 were used to train the models and 14 to test them[1].

What they found

• Models based on the whole spinal cord worked better than those based only on specific areas of the cervical cord[1].
• The best model separated the two groups perfectly in the test set (AUC = 1.00)[1].
• Five measurements mattered most. The most important one (the ratio of the cord's surface to its volume) also correlated with disease severity[1].

What it means for families

The authors see the method as a non-invasive aid alongside genetic testing that could support a more precise diagnosis[1]. SPG4 is still confirmed by genetic testing[2].

What the study does not tell us yet

The test set had only 14 people. A perfect score in such a small group needs to be confirmed in larger groups, at other hospitals and on other MRI scanners. The study compared only two subtypes, so it cannot tell SPG4 apart from other forms of HSP or other conditions.