ABSTRACT
Artificial intelligence (AI) holds much promise for enabling highly desired imaging diagnostics improvements. One of the most limiting bottlenecks for the development of useful clinical-grade AI models is the lack of training data. One aspect is the large amount of cases needed and another is the necessity of high-quality ground truth annotation. The aim of the project was to establish and describe the construction of a database with substantial amounts of detail-annotated oncology imaging data from pathology and radiology. A specific objective was to be proactive, that is, to support undefined subsequent AI training across a wide range of tasks, such as detection, quantification, segmentation, and classification, which puts particular focus on the quality and generality of the annotations. The main outcome of this project was the database as such, with a collection of labeled image data from breast, ovary, skin, colon, skeleton, and liver. In addition, this effort also served as an exploration of best practices for further scalability of high-quality image collections, and a main contribution of the study was generic lessons learned regarding how to successfully organize efforts to construct medical imaging databases for AI training, summarized as eight guiding principles covering team, process, and execution aspects.
Subject(s)
Artificial Intelligence , Radiology , Algorithms , Databases, Factual , Female , Humans , RadiographyABSTRACT
In the developing Drosophila central nervous system (CNS), neural progenitor (neuroblast [NB]) selection is gated by lateral inhibition, controlled by Notch signaling and proneural genes. However, proneural mutants still generate many NBs, indicating the existence of additional proneural genes. Moreover, recent studies reveal involvement of key epithelial-mesenchymal transition (EMT) genes in NB selection, but the regulatory interplay between Notch signaling and the EMT machinery is unclear. We find that SoxNeuro (SoxB family) and worniu (Snail family) are integrated with the Notch pathway, and constitute the missing proneural genes. Notch signaling, the proneural, SoxNeuro, and worniu genes regulate key EMT genes to orchestrate the NB selection process. Hence, we uncover an expanded lateral inhibition network for NB selection and demonstrate its link to key players in the EMT machinery. The evolutionary conservation of the genes involved suggests that the Notch-SoxB-Snail-EMT network may control neural progenitor selection in many other systems.
Subject(s)
Drosophila Proteins/metabolism , Epithelial-Mesenchymal Transition , Neural Stem Cells/metabolism , Receptors, Notch/metabolism , SOX Transcription Factors/metabolism , Transcription Factors/metabolism , Animals , Drosophila Proteins/genetics , Drosophila melanogaster , Membrane Proteins/genetics , Membrane Proteins/metabolism , Neural Stem Cells/cytology , Neurogenesis , Receptors, Notch/genetics , SOX Transcription Factors/genetics , Signal Transduction , Transcription Factors/geneticsABSTRACT
During embryonic development, a number of genetic cues act to generate neuronal diversity. While intrinsic transcriptional cascades are well-known to control neuronal sub-type cell fate, the target cells can also provide critical input to specific neuronal cell fates. Such signals, denoted retrograde signals, are known to provide critical survival cues for neurons, but have also been found to trigger terminal differentiation of neurons. One salient example of such target-derived instructive signals pertains to the specification of the Drosophila FMRFamide neuropeptide neurons, the Tv4 neurons of the ventral nerve cord. Tv4 neurons receive a BMP signal from their target cells, which acts as the final trigger to activate the FMRFa gene. A recent FMRFa-eGFP genetic screen identified several genes involved in Tv4 specification, two of which encode components of the U5 subunit of the spliceosome: brr2 (l(3)72Ab) and Prp8. In this study, we focus on the role of RNA processing during target-derived signaling. We found that brr2 and Prp8 play crucial roles in controlling the expression of the FMRFa neuropeptide specifically in six neurons of the VNC (Tv4 neurons). Detailed analysis of brr2 revealed that this control is executed by two independent mechanisms, both of which are required for the activation of the BMP retrograde signaling pathway in Tv4 neurons: (1) Proper axonal pathfinding to the target tissue in order to receive the BMP ligand. (2) Proper RNA splicing of two genes in the BMP pathway: the thickveins (tkv) gene, encoding a BMP receptor subunit, and the Medea gene, encoding a co-Smad. These results reveal involvement of specific RNA processing in diversifying neuronal identity within the central nervous system.