ConvNeXt-backbone HoVerNet for nuclei segmentation and classification
This is an incremental improvement for medical image analysis, specifically in histopathology for nuclei segmentation and classification tasks.
The paper tackles nuclei segmentation and classification in histopathology images by replacing the ResNet backbone with ConvNeXt and using HED space conversion for label smoothing, resulting in a 0.04 improvement in mPQ+ and 0.0144 in multi r2 on the validation set.
This manuscript gives a brief description of the algorithm used to participate in CoNIC Challenge 2022. After the baseline was made available, we follow the method in it and replace the ResNet baseline with ConvNeXt one. Moreover, we propose to first convert RGB space to Haematoxylin-Eosin-DAB(HED) space, then use Haematoxylin composition of origin image to smooth semantic one hot label. Afterwards, nuclei distribution of train and valid set are explored to select the best fold split for training model for final test phase submission. Results on validation set shows that even with channel of each stage smaller in number, HoVerNet with ConvNeXt-tiny backbone still improves the mPQ+ by 0.04 and multi r2 by 0.0144