AI and the World's Hard Problems · Medicine & drug discovery · Nature (computational biology) · 2026-10-24
SCLC-SBT: a spectral-bilinear transformer for small cell lung cancer histopathology image classification
To classify small cell lung cancer (SCLC) histopathology images as normal or SCLC, the authors propose SCLC-SBT, a hierarchical Spectral-Bilinear Transformer that captures both high-frequency nuclear texture and long-range tissue dependencies. On a 1,974-image benchmark it reaches 0.9697 accuracy and 0.9971 AUC, with the best point estimate among 11 compared architectures on five of seven metrics. The advantage is statistically significant against five of the eleven baselines but not against the six strongest, so the authors describe it as competitive to superior.