PanNuke Dataset Extension, Insights and Baselines
read the original abstract
The emerging area of computational pathology (CPath) is ripe ground for the application of deep learning (DL) methods to healthcare due to the sheer volume of raw pixel data in whole-slide images (WSIs) of cancerous tissue slides. However, it is imperative for the DL algorithms relying on nuclei-level details to be able to cope with data from `the clinical wild', which tends to be quite challenging. We study, and extend recently released PanNuke dataset consisting of ~200,000 nuclei categorized into 5 clinically important classes for the challenging tasks of segmenting and classifying nuclei in WSIs. Previous pan-cancer datasets consisted of only up to 9 different tissues and up to 21,000 unlabeled nuclei and just over 24,000 labeled nuclei with segmentation masks. PanNuke consists of 19 different tissue types that have been semi-automatically annotated and quality controlled by clinical pathologists, leading to a dataset with statistics similar to the clinical wild and with minimal selection bias. We study the performance of segmentation and classification models when applied to the proposed dataset and demonstrate the application of models trained on PanNuke to whole-slide images. We provide comprehensive statistics about the dataset and outline recommendations and research directions to address the limitations of existing DL tools when applied to real-world CPath applications.
This paper has not been read by Pith yet.
Forward citations
Cited by 15 Pith papers
-
VitaminP: cross-modal learning enables whole-cell segmentation from routine histology
VitaminP uses paired H&E-mIF data to train a model that transfers molecular boundary information, enabling accurate whole-cell segmentation directly from routine H&E histology across 34 cancer types.
-
SAM 3: Segment Anything with Concepts
SAM 3 introduces promptable concept segmentation that doubles accuracy of prior systems on images and videos while improving standard SAM segmentation performance.
-
MedSIGHT: Towards Grounded Visual Comprehension in Medical Large Vision-Language Models
MedSIGHT unifies medical image comprehension and segmentation in Med-LVLMs via a Region Perceiver module and region codebook, trained progressively on 72K pairs to reach SOTA on both tasks across modalities.
-
Leveraging Spatial Transcriptomics as Alternative to Manual Annotations for Deep Learning-Based Nuclei Analysis
Spatial transcriptomics provides cell-type labels and nuclear masks to train image-based deep learning models for nuclei analysis, achieving better segmentation accuracy and transferability to unseen organs than conve...
-
Multi-Beholder: Biomarker Prediction for Low-Grade Glioma with Multiple Instance Learning and One-Class Classification
Multi-Beholder integrates one-class classification into multiple instance learning to predict LGG biomarker status from histopathology images, reporting AUCs of 0.973 on TCGA-LGG and 0.820 on an external Xiangya cohort.
-
CellDETR: A Detection-Guided Framework for Scalable Cell Representation Learning from Histopathology Images
CellDETR is a detection-guided framework extending Deformable DETR for cell representation learning from WSIs, with contrastive pretraining and cross-dataset transfer shown on PanNuke and Xenium data.
-
SegTME-UNI2: A Foundation Model-Based Framework for Generalisable Multiclass Cell Segmentation and LLM-Driven Tumour Microenvironment Characterisation in Histopathology
SegTME-UNI2 pairs a UNI2-based dual-head segmentation model trained via progressive pseudo-labeling with an LLM to produce multiclass cell maps and narrative TME descriptions from H&E images.
-
Atlas H&E-TME: Scalable AI-Based Tissue Profiling at Expert Pathologist-Level Accuracy
Atlas H&E-TME is a new AI system for cell-level tissue profiling on H&E slides that matches pathologist performance when validated against an IHC-informed consensus and a large multi-cancer H&E annotation set.
-
Shift-Dependent Asymmetry: Orthogonal Inverse Low-Rank Adaptation for Federated Medical Segmentation
Introduces IAT with module-specific personalization and orthogonality regularization to handle appearance and supervision shifts in federated medical segmentation.
-
One Stone, Three Birds: Self-adaptive Optimal Transport for Multi-VLM Selection, Adaptation, and Ensembling
OSTB estimates a consensus sample-to-class transport plan from multiple frozen VLMs to perform model selection by reliability ranking, target adaptation via transport-conditioned classifiers, and ensembling via reliab...
-
PathAR: Structure-First Autoregressive Synthesis of Multimodal Pathology Images
PathAR factorizes structure and appearance tokens via Dual-VQ and IAR transformer for modality-conditioned pathology image synthesis with improved structural consistency.
-
Biological Spatial Priors Regularize Foundation Model Representations for Cross-Site MSI Generalization in Colorectal Cancer
Biological spatial priors based on MSI histology, when injected into TransMIL with foundation model features, improve cross-site generalization for MSI prediction from H&E WSIs, with peripheral distance encoding achie...
-
CellPrior-Net: Prior-Guided Nuclei Detection and Classification for H&E Whole-Slide Images
CellPrior-Net integrates hematoxylin channel prior into a lightweight CNN for nuclei detection and classification in H&E WSIs, claiming comparable accuracy to SOTA with significantly reduced inference time across 10.4...
-
DualGate-Net: A Prior-Gated Dual-Encoder Framework for Histopathology Cell Detection
DualGate-Net combines ConvNeXtV2 local and SegFormer global encoders via a prior-gated fusion module plus auxiliary reconstruction and cellness branches, reporting macro F1 of 0.7722 (val) and 0.7345 (test) on OCELOT.
-
OpenTME: An Open Dataset of AI-powered H&E Tumor Microenvironment Profiles from TCGA
OpenTME provides pre-computed TME profiles with over 4,500 quantitative readouts per slide from 3,634 TCGA H&E images using an AI pipeline based on pathology foundation models.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.