Co-registered histological-polarimetric dataset for tumor cell content estimation in brain tissue

Imaging Mueller polarimetry enables label-free contrast between tumor and healthy brain tissue, but its link to histologically validated tumor cell content remains underexplored. In this work, we imaged formalin-fixed whole-brain thick sections using a wide-field reflection-mode polarimeter and aligned the data to the images of H&E and immunohistochemically stained sections using a validated multimodal pipeline. This allowed the construction of a co-registered dataset linking polarimetric parameters to quantitative local tumor cell content maps. Machine learning models were trained for tissue classification and tumor cell content prediction. This approach supports the development of real-time, label-free tools for intraoperative assessment of tumor margins in brain surgery.