A visual-language foundation model for computational pathology

The accelerated adoption of digital pathology and advances in deep learning have enabled the development of robust models for various pathology tasks across a diverse array of diseases and patient cohorts. However, model training is often difficult due to label scarcity in the medical domain, and a model's usage is limited by the specific task and disease for which it is trained. Additionally, most models in histopathology leverage only image data, a stark contrast to how humans teach each other and reason about histopathologic entities. We introduce CONtrastive learning from Captions for Histopathology (CONCH), a visual-language foundation model developed using diverse sources of histopathology images, biomedical text and, notably, over 1.17 million image-caption pairs through task-agnostic pretraining. Evaluated on a suite of 14 diverse benchmarks, CONCH can be transferred to a wide range of downstream tasks involving histopathology images and/or text, achieving state-of-the-art performance on histology image classification, segmentation, captioning, and text-to-image and image-to-text retrieval. CONCH represents a substantial leap over concurrent visual-language pretrained systems for histopathology, with the potential to directly facilitate a wide array of machine learning-based workflows requiring minimal or no further supervised fine-tuning.

Clonal mixing, clonal restriction, and specification of cell types in the developing rat olfactory bulb

To understand the clonal relationship of various olfactory bulb (OB) cell types, OB progenitor cells were infected at embryonic day (E) 14, E15, and E17 with retroviral libraries encoding alkaline phosphatase or beta-galactosidase. After survival to postnatal day 10-15, sibling relationships were identified by polymerase chain reaction DNA amplification of distinct sequences in the retroviral constructs. Within the OB, clonal progeny dispersed widely in all directions. In sharp contrast, however, clonal dispersion between the OB and neocortex was not observed, although occasional clonal dispersion between the OB and pyriform and hippocampal regions could not be excluded. Most clones (84%) contained a single cell type, especially after E17 injections, suggesting the existence of either restricted precursors, or multipotential progenitors instructed by a restricted cellular environment. Mixed OB clones (16%) contained multiple cell types in the OB, or occasionally glial or neuronal cells outside the OB, demonstrating the existence of multipotential OB progenitors, likely at a stage before formation of the olfactory rostral migratory stream. Surprisingly, OB glial cells were not labeled, suggesting distinct lineages or perhaps distinct migratory paths for glia and neurons into the OB. A hierarchical cell lineage is proposed that involves a multipotential progenitor that gives rise to potentially more limited progenitors.

Systematic widespread clonal organization in cerebral cortex

Cell lineage analysis in the cortex has revealed two clonal patterns, clustered and widespread clones. To determine the relationship of these patterns, progenitor cells were infected with a retroviral library encoding alkaline phosphatase, and cortical sibling cells were identified using PCR. Clones labeled at E15 consisted of single cells or small cells clusters (52%) or of widespread cells (48%). However, widespread clones consisted of multiple neuronal or glial cell types, spaced systematically at 2-3 mm intervals. The data suggest that migratory multipotential progenitors divide asymmetrically at intervals defined by cell cycle length, producing single cells or clusters of cells in different cortical regions. Transition from multipotentiality to more restricted potential may correspond to changes in migratory behavior.