As　a　major　cause　of　human　death　and　disability,　neurological　diseases　have　profound　impacts　worldwide,　yet　there　is　still　a　lack　of　effective　research　to　quantify　such　diseases.　As　a　fundamental　but　challenging　task　in　the　field　of　medical　imaging,　segmenting　neuronal　cell　types　measured　by　microscopic　observation　images　to　help　assess　patient　status　plays　a　crucial　role　in　the　accurate　quantification　of　the　disease.　To　this　end,　we　propose　an　instance　segmentation　network　of　Improved　Attention　U-net　for　segmentation　of　different　types　of　neuronal　cells.　Specifically,　we　introduce　adaptive　mechanism　on　the　basis　of　U-net　structure　to　enhance　the　robustness　and　discriminability　of　the　model.　This　measure　can　capture　the　most　important　parts　in　the　field　of　view　to　filter　out　redundant　information　in　the　target　optimization　process.　In　addition,　considering　the　irregular　and　unique　morphology　of　neuronal　cells,　we　introduce　deformable　convolution　kernels　to　adapt　to　the　detection　needs　of　various　types　of　cells　and　lesions.　Experimental　results　show　that　the　algorithm　has　excellent　performance　in　cell　instance　segmentation.　©　2023　IEEE.