Performance evaluation of two different software programs for automated ASPECTS scoring in patients with suspected stroke.

BACKGROUND

The range of software available to radiologists has increased enormously with the advancement of AI. A good example of this is software to determine ASPECTS in the treatment of potential stroke patients.

OBJECTIVE

In this study, two software packages (eASPECTS from Brainomix and VIA_ASPECTS from Siemens) were tested and compared for their performance in the daily clinical routine of a maximum care provider with a 24/7 stroke unit.

METHODS

A total of 637 noncontrast CT images were obtained from consecutive patients with suspected stroke, of whom 73 were finally diagnosed with MCA infarction. Differences in agreement and quantification of agreement were analysed, as well as the correlation and sensitivity, specificity and accuracy compared to raters.

RESULTS

Compared to VIA_ASPECTS, eASPECTS shows good agreement and strong correlation with the raters. VIA_ASPECTS has lower accuracy and low specificity than eASPECTS but a higher sensitivity.

CONCLUSION

Both software products have the potential to be decision support tools for radiologists. There are, however, differences between the two software products in terms of their intended use.