Dataset Information

Discovering Digital Tumor Signatures-Using Latent Code Representations to Manipulate and Classify Liver Lesions.

ABSTRACT: Modern generative deep learning (DL) architectures allow for unsupervised learning of latent representations that can be exploited in several downstream tasks. Within the field of oncological medical imaging, we term these latent representations "digital tumor signatures" and hypothesize that they can be used, in analogy to radiomics features, to differentiate between lesions and normal liver tissue. Moreover, we conjecture that they can be used for the generation of synthetic data, specifically for the artificial insertion and removal of liver tumor lesions at user-defined spatial locations in CT images. Our approach utilizes an implicit autoencoder, an unsupervised model architecture that combines an autoencoder and two generative adversarial network (GAN)-like components. The model was trained on liver patches from 25 or 57 inhouse abdominal CT scans, depending on the experiment, demonstrating that only minimal data is required for synthetic image generation. The model was evaluated on a publicly available data set of 131 scans. We show that a PCA embedding of the latent representation captures the structure of the data, providing the foundation for the targeted insertion and removal of tumor lesions. To assess the quality of the synthetic images, we conducted two experiments with five radiologists. For experiment 1, only one rater and the ensemble-rater were marginally above the chance level in distinguishing real from synthetic data. For the second experiment, no rater was above the chance level. To illustrate that the "digital signatures" can also be used to differentiate lesion from normal tissue, we employed several machine learning methods. The best performing method, a LinearSVM, obtained 95% (97%) accuracy, 94% (95%) sensitivity, and 97% (99%) specificity, depending on if all data or only normal appearing patches were used for training of the implicit autoencoder. Overall, we demonstrate that the proposed unsupervised learning paradigm can be utilized for the removal and insertion of liver lesions at user defined spatial locations and that the digital signatures can be used to discriminate between lesions and normal liver tissue in abdominal CT scans.

SUBMITTER: Kleesiek J

PROVIDER: S-EPMC8269051 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BACKGROUND:Eukaryotes acquired the trait of oxygenic photosynthesis through endosymbiosis of the cyanobacterial progenitor of plastid organelles. Despite recent advances in the phylogenomics of Cyanobacteria, the phylogenetic root of plastids remains controversial. Although a single origin of plastids by endosymbiosis is broadly supported, recent phylogenomic studies are contradictory on whether plastids branch early or late within Cyanobacteria. One underlying cause may be poor fit of evolutionary models to complex phylogenomic data. RESULTS:Using Posterior Predictive Analysis, we show that recently applied evolutionary models poorly fit three phylogenomic datasets curated from cyanobacteria and plastid genomes because of heterogeneities in both substitution processes across sites and of compositions across lineages. To circumvent these sources of bias, we developed CYANO-MLP, a machine learning algorithm that consistently and accurately phylogenetically classifies ("phyloclassifies") cyanobacterial genomes to their clade of origin based on bioinformatically predicted function-informative features in tRNA gene complements. Classification of cyanobacterial genomes with CYANO-MLP is accurate and robust to deletion of clades, unbalanced sampling, and compositional heterogeneity in input tRNA data. CYANO-MLP consistently classifies plastid genomes into a late-branching cyanobacterial sub-clade containing single-cell, starch-producing, nitrogen-fixing ecotypes, consistent with metabolic and gene transfer data. CONCLUSIONS:Phylogenomic data of cyanobacteria and plastids exhibit both site-process heterogeneities and compositional heterogeneities across lineages. These aspects of the data require careful modeling to avoid bias in phylogenomic estimation. Furthermore, we show that amino acid recoding strategies may be insufficient to mitigate bias from compositional heterogeneities. However, the combination of our novel tRNA-specific strategy with machine learning in CYANO-MLP appears robust to these sources of bias with high accuracy in phyloclassification of cyanobacterial genomes. CYANO-MLP consistently classifies plastids as late-branching Cyanobacteria, consistent with independent evidence from signature-based approaches and some previous phylogenetic studies.

Project description:Spitzoid neoplasms are a challenging group of cutaneous melanocytic proliferations that occur in children or adolescents but can also arise in the elderly. They are characterized by epithelioid and/ or spindle-shaped melanocytes with a stromal background of variable amounts of lymphocytes, blood vessels and sclerosis. According to the WHO 2018 classification, Spitzoid melanocytic lesions are classified as benign Spitz nevi (SN), atypical Spitz tumors (AST) or malignant Spitz tumors (MST). The intermediate AST category represents a diagnostically challenging group since on purely histopathological grounds, their benign or malignant character remains unpredictable. This results in uncertainties in patient management and prognosis. The molecular properties of Spitzoid lesions, especially their transcriptomic landscape, remain poorly understood and genomic alterations in melanoma-associated oncogenes are typically absent. The aim of this study was to characterize the transcriptome of Spitzoid melanocytic neoplasms with digital mRNA expression profiling. Formalin-fixed-paraffin-embedded samples (including 27 SN, 10 AST and 14 MST) were analyzed using the RNA NanoString nCounter PanCancer Pathways Gene Expression panel. The number of significantly differentially expressed genes in SN vs. MST, SN vs. AST and AST vs. MST was 68, 167 and 18, respectively. Gene set enrichment analysis revealed an upregulation of pathways related to epithelial mesenchymal transition, immunomodulatory-, angiogenesis- as well as myogenesis associated processes in AST and MST. In addition, a specific gene expression signature of SN vs. MST was discovered based on the top-ranked six most informative gene expression markers: NRAS, NF1, BMP2, EIF2B4, IFNA17 and FZD9. The AST samples showed intermediate levels of the identified signature. This implies that the identified gene expression signature can potentially be used to distinguish high-grade from low-grade AST. This combined histopathological and transcriptomic methodology is promising for diagnostics of Spitzoid neoplasms and patient management in dermatological oncology in the future.

Project description:Heterogeneity of injury severity among children with traumatic brain injury (TBI) classified by the Glasgow Coma Scale (GCS) makes comparisons across research cohorts, enrollment in clinical trials, and clinical predictions of outcomes difficult. The present study uses latent class analysis (LCA) to distinguish severity subgroups from a prospective cohort of 433 children 2.5-15 years of age with TBI who were recruited from two level 1 pediatric trauma centers. Indicator variables available within 48 h post-injury including emergency department (ED) GCS, hospital motor GCS, Abbreviated Injury Score (AIS), Rotterdam Score, hypotension in the ED, and pre-hospital loss of consciousness, intubation, seizures, and sedation were evaluated to define subgroups. To understand whether latent class subgroups were predictive of clinically meaningful outcomes, the Pediatric Injury Functional Outcome Scale (PIFOS) at 6 and 12 months, and the Behavior Rating Inventory of Executive Function at 12 months, were compared across subgroups. Then, outcomes were examined by GCS (primary) and AIS (secondary) classification alone to assess whether LCA provided improved outcome prediction. LCA identified four distinct increasing severity subgroups (1-4). Unlike GCS classification, mean outcome differences on PIFOS at 6 months showed decreasing function across classes. PIFOS differences relative to the lowest latent class (LC1) were: LC2 2.27 (0.83, 3.72), LC3 3.99 (1.88, 6.10), and LC4 11.2 (7.04, 15.4). Differences in 12 month outcomes were seen between the most and least severely injured groups. Differences in outcomes in relation to AIS were restricted to the most and less severely injured at both time points. This study distinguished four latent classes that are clinically meaningful, distinguished a more homogenous severe injury group, and separated children by 6-month functional outcomes better than GCS alone. Systematic reporting of these variables would allow comparisons across research cohorts, potentially improve clinical predictions, and increase sensitivity to treatment effects in clinical trials.

Dataset Information

Discovering Digital Tumor Signatures-Using Latent Code Representations to Manipulate and Classify Liver Lesions.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets