An annotated fluorescence image dataset for training nuclear segmentation methods

Abstract

Fully-automated nuclear image segmentation is the prerequisite to ensure statistically significant, quantitative analyses of tissue preparations,applied in digital pathology or quantitative microscopy. The design of segmentation methods that work independently of the tissue type or preparation is complex, due to variations in nuclear morphology, staining intensity, cell density and nuclei aggregations. Machine learning-based segmentation methods can overcome these challenges, however high quality expert-annotated images are required for training. Currently, the limited number of annotated fluorescence image datasets publicly available do not cover a broad range of tissues and preparations. We present a comprehensive, annotated dataset including tightly aggregated nuclei of multiple tissues for the training of machine learning-based nuclear segmentation algorithms. The proposed dataset covers sample preparation methods frequently used in quantitative immunofluorescence microscopy. We demonstrate the heterogeneity of the dataset with respect to multiple parameters such as magnification, modality, signal-to-noise ratio and diagnosis. Based on a suggested split into training and test sets and additional single-nuclei expert annotations, machine learning-based image segmentation methods can be trained and evaluated.

Single-cell landscape of bone marrow metastases in human neuroblastoma unraveled by deep multiplex imaging

“…The deep neuronal network Mask R-CNN 47 , trained on an annotated fluorescence image dataset 48 , allows accurate cell and nucleus segmentation in the processed images. Simultaneous segmentation of the nucleus (based on the nuclear stain propidium iodide) and the cell (based on phase contrast images acquired prior to each staining cycle) allows the elimination of displaced and inaccurately segmented cells by considering only those cells for the analysis, which are present in every cell segmentation mask, but also in the nucleus segmentation mask.…”

“…Furthermore, we only found myelocytes in samples with a high DTC infiltration, which may be attributable to the immunosuppressive and tumor-promoting functions of cells from the myeloid origin as well as their role in inflammation. 71 Our findings were based on single-cell analyses of MELC multiplex imaging data, enabled by the pipeline DeepFLEX, which we developed based on the integration of methods for image processing 45,46 , segmentation 47,48 , normalization 49 and single-cell analysis. 50 DeepFLEX tackles confounding factors of targeted imaging technologies such as unspecific binding and autofluorescence and combines deep-learning based cell and nucleus segmentation, which allow accurate single-cell assessment.…”

“…For accurate nuclei and cell segmentation, annotated datasets 48 of propidium iodide or phase contrast images were created, respectively. We trained the deep learning architecture Mask R-CNN for instance-aware segmentation, as previously described.…”

Evaluation of Deep Learning Architectures for Complex Immunofluorescence Nuclear Image Segmentation

“…We use a recently published dataset [12] consisting of 79 images of IF stained nuclei images containing 7813 nuclei in total. The images are from specimens of different diagnosis, namely human ganglioneuroblastoma (GNB) tumors, human neuroblastoma (NB) tumors, Wilms tumor (Wilms) and a human keratinocyte cell line (HaCaT).…”

“…In particular, the annotation of highly complex images such as tissue sections is time consuming and expensive due to the human resources needed. To enable an evaluation of stateof-the-art segmentation methods on such images, we recently published an expert-annotated dataset consisting of fluorescent nuclear images and annotations of various tissue origins and sample preparation types, acquired using multiple modalities, different levels of magnification and signal-to-noise ratio [12], [13].…”