Hyperspectral imaging (HSI) is a three-dimensional imaging technique that integrates spectroscopy and imaging. When combined with microscopy, hyperspectral microscopic imaging (HMI) captures rich spatial-spectral information at the cellular scale, offering new avenues for histopathological analysis. Conventional pathological diagnosis relies on manual inspection of stained slides, which is time-consuming, subjective, and limited in capturing biochemical variations. While machine learning combined with HMI has shown promise in improving diagnostic accuracy and automation, progress remains constrained by the lack of publicly available datasets, especially for lung cancer, one of the most common malignant tumors worldwide. To address this gap, we present HMI-LUSC, the first open HMI dataset for lung squamous cell carcinoma (LUSC). The dataset was acquired using a custom HMI system and includes 62 hyperspectral images from 10 patients, spanning 450-750 nm across 61 spectral bands, with pathologist-provided tumor annotations and refined cell-level labels generated via a semi-automated workflow. HMI-LUSC provides a robust benchmark for spectral analysis and tumor detection, fostering future advances in computational pathology and spectral diagnostic research.
Yan et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: