Spinal-Multiple-Myeloma-SEG: Segmentation of spinal multiple myeloma lesions in dual-energy CT
2026-08-11
We present a unique dataset comprising dual-energy low-dose CT scans from 67 patients diagnosed with multiple myeloma, with a total of 72 scans. The dataset includes conventional CT images, virtual monoenergetic images (at 40, 80, and 120 keV), calcium-suppressed images (with suppression indices of 25, 50, 75, and 100), as well as segmentation masks of vertebrae (including vertebra type classification) and multiple myeloma lesions in the spine. In total, the dataset contains 576 image series comprising 564,464 axial slices. In addition to image data, the dataset provides supporting non-image information, including basic demographic details of patients (mean age 66 years, range 48–85; 36% female) and selected clinical variables reflecting disease burden and progression (e.g., M-protein characteristics, free light chains, laboratory parameters, staging, bone marrow findings, and treatment response). This dataset holds substantial promise for advancing and objectively evaluating computer-aided detection and diagnostic systems, particularly those based on machine learning and artificial intelligence. It addresses the current lack of publicly available datasets focused on dual-energy CT–based lesion segmentation in patients with multiple myeloma (acquired using dual-layer CT technology, Philips IQon Spectral CT), and may support the development and validation of algorithms for lesion detection, disease monitoring, and assessment of treatment response.