Accurate identification of the glomerular ultrastructure is critical for the diagnosis of chronic kidney diseases,but the high cost of acquiring high-quality annotated data limits the application of fully-supervised learning.Therefore,a multi-class semi-supervised semantic segmentation framework based on segment anything model(MC4S-SAM)is proposed.After improving the mask decoder of segment anything model to enable multi-class semantic segmentation without requiring prompt information,the improved model is used to generate and refine pseudo-labels through a self-training strategy,and multi-level consistency regularization constraints are incorporated to enhance the model's performance.Experimental results show that,in the task of segmenting the glomerular mesangial ultrastructure,MC4S-SAM outperformes the fully-supervised model by 11.72%in mean intersection over union(mIoU)and 11.45%in mean Dice similarity coefficient(mDSC)when the labeled data accountes for 1/16 of the total.When the labeled data proportion is 1/4,the mIoU and mDSC reach 68.91%and 78.73%,respectively,demonstrating its significant potential for aiding the diagnosis of chronic kidney diseases.

Accurate identification of the glomerular ultrastructure is critical for the diagnosis of chronic kidney diseases,but the high cost of acquiring high-quality annotated data limits the application of fully-supervised learning.Therefore,a multi-class semi-supervised semantic segmentation framework based on segment anything model(MC4S-SAM)is proposed.After improving the mask decoder of segment anything model to enable multi-class semantic segmentation without requiring prompt information,the improved model is used to generate and refine pseudo-labels through a self-training strategy,and multi-level consistency regularization constraints are incorporated to enhance the model's performance.Experimental results show that,in the task of segmenting the glomerular mesangial ultrastructure,MC4S-SAM outperformes the fully-supervised model by 11.72%in mean intersection over union(mIoU)and 11.45%in mean Dice similarity coefficient(mDSC)when the labeled data accountes for 1/16 of the total.When the labeled data proportion is 1/4,the mIoU and mDSC reach 68.91%and 78.73%,respectively,demonstrating its significant potential for aiding the diagnosis of chronic kidney diseases.

Objective To develop a multi-modal deep learning method for automatic classification of immune-mediated glomerular diseases based on images of optical microscopy(OM),immunofluorescence microscopy(IM),and transmission electron microscopy(TEM).Methods We retrospectively collected the pathological images from 273 patients and constructed a multi-modal multi-instance model for classification of 3 immune-mediated glomerular diseases,namely immunoglobulin A nephropathy(IgAN),membranous nephropathy(MN),and lupus nephritis(LN).This model adopts an instance-level multi-instance learning(I-MIL)method to select the TEM images for multi-modal feature fusion with the OM images and IM images of the same patient.By comparing this model with unimodal and bimodal models,we explored different combinations of the 3 modalities and the optimal methods for modal feature fusion.Results The multi-modal multi-instance model combining OM,IM,and TEM images had a disease classification accuracy of(88.34±2.12)%,superior to that of the optimal unimodal model[(87.08±4.25)%]and that of the optimal bimodal model[(87.92±3.06)%].Conclusion This multi-modal multi-instance model based on OM,IM,and TEM images can achieve automatic classification of immune-mediated glomerular diseases with a good classification accuracy.

Objective To develop a multi-modal deep learning method for automatic classification of immune-mediated glomerular diseases based on images of optical microscopy(OM),immunofluorescence microscopy(IM),and transmission electron microscopy(TEM).Methods We retrospectively collected the pathological images from 273 patients and constructed a multi-modal multi-instance model for classification of 3 immune-mediated glomerular diseases,namely immunoglobulin A nephropathy(IgAN),membranous nephropathy(MN),and lupus nephritis(LN).This model adopts an instance-level multi-instance learning(I-MIL)method to select the TEM images for multi-modal feature fusion with the OM images and IM images of the same patient.By comparing this model with unimodal and bimodal models,we explored different combinations of the 3 modalities and the optimal methods for modal feature fusion.Results The multi-modal multi-instance model combining OM,IM,and TEM images had a disease classification accuracy of(88.34±2.12)%,superior to that of the optimal unimodal model[(87.08±4.25)%]and that of the optimal bimodal model[(87.92±3.06)%].Conclusion This multi-modal multi-instance model based on OM,IM,and TEM images can achieve automatic classification of immune-mediated glomerular diseases with a good classification accuracy.