Reproducibility Analysis of Radiomic Features from T2-weighted MRI after Processing and Segmentation Alternations in Neuroblastoma Tumors.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To evaluate the reproducibility of radiomics features extracted from T2-weighted MRI in patients with neuroblastoma. Materials and Methods A retrospective study included 419 patients (mean (SD) age, 29 (34) years; 220 male, 199 female) with neuroblastic tumors, diagnosed between 2002-2023, within the scope of the PRIMAGE-project, involving 746 MRI T2/T2*-weighted sequences at diagnosis and/or after initial chemotherapy. Images underwent processing steps (denoising, inhomogeneity bias field correction, normalization, and resampling). Tumors were automatically segmented and 107 shape, first-order and second-order radiomic features were extracted, considered as the reference standard. Subsequently, the previous image processing settings were modified, and volumetric masks were applied. New radiomics features were extracted and compared with the reference standard. Reproducibility was assessed using the concordance correlation coefficient (CCC), intrasubject repeatability was measured using the coefficient of variation (CoV). Results When normalization was omitted, only 5% of the radiomics features demonstrated high reproducibility. Statistical analysis revealed significant changes in the normalization and resampling processes (P < .001). Inhomogeneities removal had the least impact on radiomics (83% of parameters remained stable). Shape features remained stable after mask modifications, with a CCC > 0.90. Mask modifications were the most favorable changes for achieving high CCC values, with stability of 70% of the radiomics features. Only 7% of second-order radiomics features showed an excellent CoV of < 0.10. Conclusion Modifications in the T2-weighted MRI preparation process in patients with neuroblastoma resulted in changes in radiomics features, with normalization identified as the most influential factor for reproducibility. Inhomogeneities removal had the least impact on radiomics features. ©RSNA, 2024.

A human neural crest model reveals the developmental impact of neuroblastoma-associated chromosomal aberrations.

Early childhood tumours arise from transformed embryonic cells, which often carry large copy number alterations (CNA). However, it remains unclear how CNAs contribute to embryonic tumourigenesis due to a lack of suitable models. Here we employ female human embryonic stem cell (hESC) differentiation and single-cell transcriptome and epigenome analysis to assess the effects of chromosome 17q/1q gains, which are prevalent in the embryonal tumour neuroblastoma (NB). We show that CNAs impair the specification of trunk neural crest (NC) cells and their sympathoadrenal derivatives, the putative cells-of-origin of NB. This effect is exacerbated upon overexpression of MYCN, whose amplification co-occurs with CNAs in NB. Moreover, CNAs potentiate the pro-tumourigenic effects of MYCN and mutant NC cells resemble NB cells in tumours. These changes correlate with a stepwise aberration of developmental transcription factor networks. Together, our results sketch a mechanistic framework for the CNA-driven initiation of embryonal tumours.

Targeting NRAS via miR-1304-5p or farnesyltransferase inhibition confers sensitivity to ALK inhibitors in ALK-mutant neuroblastoma.

Targeting Anaplastic lymphoma kinase (ALK) is a promising therapeutic strategy for aberrant ALK-expressing malignancies including neuroblastoma, but resistance to ALK tyrosine kinase inhibitors (ALK TKI) is a distinct possibility necessitating drug combination therapeutic approaches. Using high-throughput, genome-wide CRISPR-Cas9 knockout screens, we identify miR-1304-5p loss as a desensitizer to ALK TKIs in aberrant ALK-expressing neuroblastoma; inhibition of miR-1304-5p decreases, while mimics of this miRNA increase the sensitivity of neuroblastoma cells to ALK TKIs. We show that miR-1304-5p targets NRAS, decreasing cell viability via induction of apoptosis. It follows that the farnesyltransferase inhibitor (FTI) lonafarnib in addition to ALK TKIs act synergistically in neuroblastoma, inducing apoptosis in vitro. In particular, on combined treatment of neuroblastoma patient derived xenografts with an FTI and an ALK TKI complete regression of tumour growth is observed although tumours rapidly regrow on cessation of therapy. Overall, our data suggests that combined use of ALK TKIs and FTIs, constitutes a therapeutic approach to treat high risk neuroblastoma although prolonged therapy is likely required to prevent relapse.

Comparative transcriptomics coupled to developmental grading via transgenic zebrafish reporter strains identifies conserved features in neutrophil maturation.

Neutrophils are evolutionarily conserved innate immune cells playing pivotal roles in host defense. Zebrafish models have contributed substantially to our understanding of neutrophil functions but similarities to human neutrophil maturation have not been systematically characterized, which limits their applicability to studying human disease. Here we show, by generating and analysing transgenic zebrafish strains representing distinct neutrophil differentiation stages, a high-resolution transcriptional profile of neutrophil maturation. We link gene expression at each stage to characteristic transcription factors, including C/ebp-ß, which is important for late neutrophil maturation. Cross-species comparison of zebrafish, mouse, and human samples confirms high molecular similarity of immature stages and discriminates zebrafish-specific from pan-species gene signatures. Applying the pan-species neutrophil maturation signature to RNA-sequencing data from human neuroblastoma patients reveals association between metastatic tumor cell infiltration in the bone marrow and an overall increase in mature neutrophils. Our detailed neutrophil maturation atlas thus provides a valuable resource for studying neutrophil function at different stages across species in health and disease.

The molecular basis of tumor metastasis and current approaches to decode targeted migration-promoting events in pediatric neuroblastoma.

Cell motility is a crucial biological process that plays a critical role in the development of multicellular organisms and is essential for tissue formation and regeneration. However, uncontrolled cell motility can lead to the development of various diseases, including neoplasms. In this review, we discuss recent advances in the discovery of regulatory mechanisms underlying the metastatic spread of neuroblastoma, a solid pediatric tumor that originates in the embryonic migratory cells of the neural crest. The highly motile phenotype of metastatic neuroblastoma cells requires targeting of intracellular and extracellular processes, that, if affected, would be helpful for the treatment of high-risk patients with neuroblastoma, for whom current therapies remain inadequate. Development of new potentially migration-inhibiting compounds and standardized preclinical approaches for the selection of anti-metastatic drugs in neuroblastoma will also be discussed.

Single-cell transcriptomics and epigenomics unravel the role of monocytes in neuroblastoma bone marrow metastasis.

Metastasis is the major cause of cancer-related deaths. Neuroblastoma (NB), a childhood tumor has been molecularly defined at the primary cancer site, however, the bone marrow (BM) as the metastatic niche of NB is poorly characterized. Here we perform single-cell transcriptomic and epigenomic profiling of BM aspirates from 11 subjects spanning three major NB subtypes and compare these to five age-matched and metastasis-free BM, followed by in-depth single cell analyses of tissue diversity and cell-cell interactions, as well as functional validation. We show that cellular plasticity of NB tumor cells is conserved upon metastasis and tumor cell type composition is NB subtype-dependent. NB cells signal to the BM microenvironment, rewiring via macrophage mgration inhibitory factor and midkine signaling specifically monocytes, which exhibit M1 and M2 features, are marked by activation of pro- and anti-inflammatory programs, and express tumor-promoting factors, reminiscent of tumor-associated macrophages. The interactions and pathways characterized in our study provide the basis for therapeutic approaches that target tumor-to-microenvironment interactions.

Mathematical Model of Clonal Evolution Proposes a Personalised Multi-Modal Therapy for High-Risk Neuroblastoma.

Neuroblastoma is the most common extra-cranial solid tumour in children. Despite multi-modal therapy, over half of the high-risk patients will succumb. One contributing factor is the one-size-fits-all nature of multi-modal therapy. For example, during the first step (induction chemotherapy), the standard regimen (rapid COJEC) administers fixed doses of chemotherapeutic agents in eight two-week cycles. Perhaps because of differences in resistance, this standard regimen results in highly heterogeneous outcomes in different tumours. In this study, we formulated a mathematical model comprising ordinary differential equations. The equations describe the clonal evolution within a neuroblastoma tumour being treated with vincristine and cyclophosphamide, which are used in the rapid COJEC regimen, including genetically conferred and phenotypic drug resistance. The equations also describe the agents' pharmacokinetics. We devised an optimisation algorithm to find the best chemotherapy schedules for tumours with different pre-treatment clonal compositions. The optimised chemotherapy schedules exploit the cytotoxic difference between the two drugs and intra-tumoural clonal competition to shrink the tumours as much as possible during induction chemotherapy and before surgical removal. They indicate that induction chemotherapy can be improved by finding and using personalised schedules. More broadly, we propose that the overall multi-modal therapy can be enhanced by employing targeted therapies against the mutations and oncogenic pathways enriched and activated by the chemotherapeutic agents. To translate the proposed personalised multi-modal therapy into clinical use, patient-specific model calibration and treatment optimisation are necessary. This entails a decision support system informed by emerging medical technologies such as multi-region sequencing and liquid biopsies. The results and tools presented in this paper could be the foundation of this decision support system.

Independent Validation of a Deep Learning nnU-Net Tool for Neuroblastoma Detection and Segmentation in MR Images.

OBJECTIVES: To externally validate and assess the accuracy of a previously trained fully automatic nnU-Net CNN algorithm to identify and segment primary neuroblastoma tumors in MR images in a large children cohort. METHODS: An international multicenter, multivendor imaging repository of patients with neuroblastic tumors was used to validate the performance of a trained Machine Learning (ML) tool to identify and delineate primary neuroblastoma tumors. The dataset was heterogeneous and completely independent from the one used to train and tune the model, consisting of 300 children with neuroblastic tumors having 535 MR T2-weighted sequences (486 sequences at diagnosis and 49 after finalization of the first phase of chemotherapy). The automatic segmentation algorithm was based on a nnU-Net architecture developed within the PRIMAGE project. For comparison, the segmentation masks were manually edited by an expert radiologist, and the time for the manual editing was recorded. Different overlaps and spatial metrics were calculated to compare both masks. RESULTS: The median Dice Similarity Coefficient (DSC) was high 0.997; 0.944-1.000 (median; Q1-Q3). In 18 MR sequences (6%), the net was not able neither to identify nor segment the tumor. No differences were found regarding the MR magnetic field, type of T2 sequence, or tumor location. No significant differences in the performance of the net were found in patients with an MR performed after chemotherapy. The time for visual inspection of the generated masks was 7.9 ± 7.5 (mean ± Standard Deviation (SD)) seconds. Those cases where manual editing was needed (136 masks) required 124 ± 120 s. CONCLUSIONS: The automatic CNN was able to locate and segment the primary tumor on the T2-weighted images in 94% of cases. There was an extremely high agreement between the automatic tool and the manually edited masks. This is the first study to validate an automatic segmentation model for neuroblastic tumor identification and segmentation with body MR images. The semi-automatic approach with minor manual editing of the deep learning segmentation increases the radiologist's confidence in the solution with a minor workload for the radiologist.

Human repair-related Schwann cells adopt functions of antigen-presenting cells in vitro.

The plastic potential of Schwann cells (SCs) is increasingly recognized to play a role after nerve injury and in diseases of the peripheral nervous system. Reports on the interaction between immune cells and SCs indicate their involvement in inflammatory processes. However, the immunocompetence of human SCs has been primarily deduced from neuropathies, but whether after nerve injury SCs directly regulate an adaptive immune response is unknown. Here, we performed comprehensive analysis of immunomodulatory capacities of human repair-related SCs (hrSCs), which recapitulate SC response to nerve injury in vitro. We used our well-established culture model of primary hrSCs from human peripheral nerves and analyzed the transcriptome, secretome, and cell surface proteins for pathways and markers relevant in innate and adaptive immunity, performed phagocytosis assays, and monitored T-cell subset activation in allogeneic co-cultures. Our findings show that hrSCs are phagocytic, which is in line with high MHCII expression. Furthermore, hrSCs express co-regulatory proteins, such as CD40, CD80, B7H3, CD58, CD86, and HVEM, release a plethora of chemoattractants, matrix remodeling proteins and pro- as well as anti-inflammatory cytokines, and upregulate the T-cell inhibiting PD-L1 molecule upon pro-inflammatory stimulation with IFNγ. In contrast to monocytes, hrSC alone are not sufficient to trigger allogenic CD4+ and CD8+ T-cells, but limit number and activation status of exogenously activated T-cells. This study demonstrates that hrSCs possess features and functions typical for professional antigen-presenting cells in vitro, and suggest a new role of these cells as negative regulators of T-cell immunity during nerve regeneration.

The evolutionary dynamics of extrachromosomal DNA in human cancers.

Oncogene amplification on extrachromosomal DNA (ecDNA) is a common event, driving aggressive tumor growth, drug resistance and shorter survival. Currently, the impact of nonchromosomal oncogene inheritance-random identity by descent-is poorly understood. Also unclear is the impact of ecDNA on somatic variation and selection. Here integrating theoretical models of random segregation, unbiased image analysis, CRISPR-based ecDNA tagging with live-cell imaging and CRISPR-C, we demonstrate that random ecDNA inheritance results in extensive intratumoral ecDNA copy number heterogeneity and rapid adaptation to metabolic stress and targeted treatment. Observed ecDNAs benefit host cell survival or growth and can change within a single cell cycle. ecDNA inheritance can predict, a priori, some of the aggressive features of ecDNA-containing cancers. These properties are facilitated by the ability of ecDNA to rapidly adapt genomes in a way that is not possible through chromosomal oncogene amplification. These results show how the nonchromosomal random inheritance pattern of ecDNA contributes to poor outcomes for patients with cancer.

Comparative Multicentric Evaluation of Inter-Observer Variability in Manual and Automatic Segmentation of Neuroblastic Tumors in Magnetic Resonance Images.

Tumor segmentation is one of the key steps in imaging processing. The goals of this study were to assess the inter-observer variability in manual segmentation of neuroblastic tumors and to analyze whether the state-of-the-art deep learning architecture nnU-Net can provide a robust solution to detect and segment tumors on MR images. A retrospective multicenter study of 132 patients with neuroblastic tumors was performed. Dice Similarity Coefficient (DSC) and Area Under the Receiver Operating Characteristic Curve (AUC ROC) were used to compare segmentation sets. Two more metrics were elaborated to understand the direction of the errors: the modified version of False Positive (FPRm) and False Negative (FNR) rates. Two radiologists manually segmented 46 tumors and a comparative study was performed. nnU-Net was trained-tuned with 106 cases divided into five balanced folds to perform cross-validation. The five resulting models were used as an ensemble solution to measure training (n = 106) and validation (n = 26) performance, independently. The time needed by the model to automatically segment 20 cases was compared to the time required for manual segmentation. The median DSC for manual segmentation sets was 0.969 (±0.032 IQR). The median DSC for the automatic tool was 0.965 (±0.018 IQR). The automatic segmentation model achieved a better performance regarding the FPRm. MR images segmentation variability is similar between radiologists and nnU-Net. Time leverage when using the automatic model with posterior visual validation and manual adjustment corresponds to 92.8%.

Amplification of CDK4 and MDM2: a detailed study of a high-risk neuroblastoma subgroup.

In neuroblastoma, MYCN amplification and 11q-deletion are important, although incomplete, markers of high-risk disease. It is therefore relevant to characterize additional alterations that can function as prognostic and/or predictive markers. Using SNP-microarrays, a group of neuroblastoma patients showing amplification of one or multiple 12q loci was identified. Two loci containing CDK4 and MDM2 were commonly co-amplified, although amplification of either locus in the absence of the other was observed. Pharmacological inhibition of CDK4/6 with ribociclib or abemaciclib decreased proliferation in a broad set of neuroblastoma cell lines, including CDK4/MDM2-amplified, whereas MDM2 inhibition by Nutlin-3a was only effective in p53wild-type cells. Combined CDK4/MDM2 targeting had an additive effect in p53wild-type cell lines, while no or negative additive effect was observed in p53mutated cells. Most 12q-amplified primary tumors were of abdominal origin, including those of intrarenal origin initially suspected of being Wilms' tumor. An atypical metastatic pattern was also observed with low degree of bone marrow involvement, favoring other sites such as the lungs. Here we present detailed biological data of an aggressive neuroblastoma subgroup hallmarked by 12q amplification and atypical clinical presentation for which our in vitro studies indicate that CDK4 and/or MDM2 inhibition also could be beneficial.

Mutational spectrum of ATRX aberrations in neuroblastoma and associated patient and tumor characteristics.

Neuroblastoma is the most common extracranial solid tumor in children. The chromatin remodeler ATRX is frequently mutated in high-risk patients with a poor prognosis. Although many studies have reported ATRX aberrations and the associated clinical characteristics in neuroblastoma, a comprehensive overview is currently lacking. In this study, we extensively characterize the mutational spectrum of ATRX aberrations in neuroblastoma tumors reported in previous studies and present an overview of patient and tumor characteristics. We collected the data of a total of 127 neuroblastoma patients and three cell lines with ATRX aberrations originating from 20 papers. We subdivide the ATRX aberrations into nonsense, missense, and multiexon deletions (MEDs) and show that 68% of them are MEDs. Of these MEDs, 75% are predicted to be in-frame. Furthermore, we identify a missense mutational hotspot region in the helicase domain. We also confirm that all three ATRX mutation types are more often identified in patients diagnosed at an older age, but still approximately 40% of the patients are aged 5 years or younger at diagnosis. Surprisingly, we found that 11q deletions are enriched in neuroblastomas with ATRX deletions compared to a reference cohort, but not in neuroblastomas with ATRX point mutations. Taken together, our data emphasizes a distinct ATRX mutation spectrum in neuroblastoma, which should be considered when studying molecular phenotypes and therapeutic strategies.

GPC2-CAR T cells tuned for low antigen density mediate potent activity against neuroblastoma without toxicity.

Pediatric cancers often mimic fetal tissues and express proteins normally silenced postnatally that could serve as immune targets. We developed T cells expressing chimeric antigen receptors (CARs) targeting glypican-2 (GPC2), a fetal antigen expressed on neuroblastoma (NB) and several other solid tumors. CARs engineered using standard designs control NBs with transgenic GPC2 overexpression, but not those expressing clinically relevant GPC2 site density (∼5,000 molecules/cell, range 1-6 × 103). Iterative engineering of transmembrane (TM) and co-stimulatory domains plus overexpression of c-Jun lowered the GPC2-CAR antigen density threshold, enabling potent and durable eradication of NBs expressing clinically relevant GPC2 antigen density, without toxicity. These studies highlight the critical interplay between CAR design and antigen density threshold, demonstrate potent efficacy and safety of a lead GPC2-CAR candidate suitable for clinical testing, and credential oncofetal antigens as a promising class of targets for CAR T cell therapy of solid tumors.

Comparison of three different methods to detect bone marrow involvement in patients with neuroblastoma.

PURPOSE: Neuroblastoma (NB) is the most frequent extracranial tumor in children. The detection of bone marrow (BM) involvement is crucial for correct staging and risk-adapted treatment. We compared three methods regarding the detection of NB involvement in BM. METHODS: Eighty-one patients with NB were included in this retrospective study. BM samples were obtained at designated time points at study entry and during treatment or follow-up. The diagnostic tools for BM analysis included cytomorphology (CM), flow cytometry (FCM) and automatic immunofluorescence plus fluorescence in situ hybridization (AIPF). RESULTS: We analyzed 369 aspirates in 81 patients in whom AIPF, CM, and FCM were simultaneously available. During the observation period, NB cells were detected in 86/369 (23.3%) cases, by CM in 32/369 (8.7%), by FCM in 52 (14.1%), and by AIPF in 72 (19.5%) samples. AIPF and/or FCM confirmed all positive results obtained in CM and detected 11 additional positive BM aspirates in 294 CM negative samples (p < 0,001). Survival of patients with BM involvement at study entry identified solely by FCM/AIPF was 17.4% versus 0% for patients in whom BM involvement was already identified by CM. CONCLUSION: The combination of AIPF/FCM yielded the highest detection rate of NB cells in BM. AIPF was the single, most sensitive method in detecting these cells. Although CM did not provide any additional positive results, it is still a useful, readily available and cost-effective tool. The prognostic significance of FCM and AIPF should be confirmed in a prospective study with a larger number of patients.

Landscape of Bone Marrow Metastasis in Human Neuroblastoma Unraveled by Transcriptomics and Deep Multiplex Imaging.

While the bone marrow attracts tumor cells in many solid cancers leading to poor outcome in affected patients, comprehensive analyses of bone marrow metastases have not been performed on a single-cell level. We here set out to capture tumor heterogeneity and unravel microenvironmental changes in neuroblastoma, a solid cancer with bone marrow involvement. To this end, we employed a multi-omics data mining approach to define a multiplex imaging panel and developed DeepFLEX, a pipeline for subsequent multiplex image analysis, whereby we constructed a single-cell atlas of over 35,000 disseminated tumor cells (DTCs) and cells of their microenvironment in the metastatic bone marrow niche. Further, we independently profiled the transcriptome of a cohort of 38 patients with and without bone marrow metastasis. Our results revealed vast diversity among DTCs and suggest that FAIM2 can act as a complementary marker to capture DTC heterogeneity. Importantly, we demonstrate that malignant bone marrow infiltration is associated with an inflammatory response and at the same time the presence of immuno-suppressive cell types, most prominently an immature neutrophil/granulocytic myeloid-derived suppressor-like cell type. The presented findings indicate that metastatic tumor cells shape the bone marrow microenvironment, warranting deeper investigations of spatio-temporal dynamics at the single-cell level and their clinical relevance.

Frequency and Prognostic Impact of ALK Amplifications and Mutations in the European Neuroblastoma Study Group (SIOPEN) High-Risk Neuroblastoma Trial (HR-NBL1).

PURPOSE: In neuroblastoma (NB), the ALK receptor tyrosine kinase can be constitutively activated through activating point mutations or genomic amplification. We studied ALK genetic alterations in high-risk (HR) patients on the HR-NBL1/SIOPEN trial to determine their frequency, correlation with clinical parameters, and prognostic impact. MATERIALS AND METHODS: Diagnostic tumor samples were available from 1,092 HR-NBL1/SIOPEN patients to determine ALK amplification status (n = 330), ALK mutational profile (n = 191), or both (n = 571). RESULTS: Genomic ALK amplification (ALKa) was detected in 4.5% of cases (41 out of 901), all except one with MYCN amplification (MNA). ALKa was associated with a significantly poorer overall survival (OS) (5-year OS: ALKa [n = 41] 28% [95% CI, 15 to 42]; no-ALKa [n = 860] 51% [95% CI, 47 to 54], [P < .001]), particularly in cases with metastatic disease. ALK mutations (ALKm) were detected at a clonal level (> 20% mutated allele fraction) in 10% of cases (76 out of 762) and at a subclonal level (mutated allele fraction 0.1%-20%) in 3.9% of patients (30 out of 762), with a strong correlation between the presence of ALKm and MNA (P < .001). Among 571 cases with known ALKa and ALKm status, a statistically significant difference in OS was observed between cases with ALKa or clonal ALKm versus subclonal ALKm or no ALK alterations (5-year OS: ALKa [n = 19], 26% [95% CI, 10 to 47], clonal ALKm [n = 65] 33% [95% CI, 21 to 44], subclonal ALKm (n = 22) 48% [95% CI, 26 to 67], and no alteration [n = 465], 51% [95% CI, 46 to 55], respectively; P = .001). Importantly, in a multivariate model, involvement of more than one metastatic compartment (hazard ratio [HR], 2.87; P < .001), ALKa (HR, 2.38; P = .004), and clonal ALKm (HR, 1.77; P = .001) were independent predictors of poor outcome. CONCLUSION: Genetic alterations of ALK (clonal mutations and amplifications) in HR-NB are independent predictors of poorer survival. These data provide a rationale for integration of ALK inhibitors in upfront treatment of HR-NB with ALK alterations.

Multimodal analysis of cell-free DNA whole-genome sequencing for pediatric cancers with low mutational burden.

Sequencing of cell-free DNA in the blood of cancer patients (liquid biopsy) provides attractive opportunities for early diagnosis, assessment of treatment response, and minimally invasive disease monitoring. To unlock liquid biopsy analysis for pediatric tumors with few genetic aberrations, we introduce an integrated genetic/epigenetic analysis method and demonstrate its utility on 241 deep whole-genome sequencing profiles of 95 patients with Ewing sarcoma and 31 patients with other pediatric sarcomas. Our method achieves sensitive detection and classification of circulating tumor DNA in peripheral blood independent of any genetic alterations. Moreover, we benchmark different metrics for cell-free DNA fragmentation analysis, and we introduce the LIQUORICE algorithm for detecting circulating tumor DNA based on cancer-specific chromatin signatures. Finally, we combine several fragmentation-based metrics into an integrated machine learning classifier for liquid biopsy analysis that exploits widespread epigenetic deregulation and is tailored to cancers with low mutation rates. Clinical associations highlight the potential value of cfDNA fragmentation patterns as prognostic biomarkers in Ewing sarcoma. In summary, our study provides a comprehensive analysis of circulating tumor DNA beyond recurrent genetic aberrations, and it renders the benefits of liquid biopsy more readily accessible for childhood cancers.

Evaluation of Deep Learning Architectures for Complex Immunofluorescence Nuclear Image Segmentation.

Separating and labeling each nuclear instance (instance-aware segmentation) is the key challenge in nuclear image segmentation. Deep Convolutional Neural Networks have been demonstrated to solve nuclear image segmentation tasks across different imaging modalities, but a systematic comparison on complex immunofluorescence images has not been performed. Deep learning based segmentation requires annotated datasets for training, but annotated fluorescence nuclear image datasets are rare and of limited size and complexity. In this work, we evaluate and compare the segmentation effectiveness of multiple deep learning architectures (U-Net, U-Net ResNet, Cellpose, Mask R-CNN, KG instance segmentation) and two conventional algorithms (Iterative h-min based watershed, Attributed relational graphs) on complex fluorescence nuclear images of various types. We propose and evaluate a novel strategy to create artificial images to extend the training set. Results show that instance-aware segmentation architectures and Cellpose outperform the U-Net architectures and conventional methods on complex images in terms of F1 scores, while the U-Net architectures achieve overall higher mean Dice scores. Training with additional artificially generated images improves recall and F1 scores for complex images, thereby leading to top F1 scores for three out of five sample preparation types. Mask R-CNN trained on artificial images achieves the overall highest F1 score on complex images of similar conditions to the training set images while Cellpose achieves the overall highest F1 score on complex images of new imaging conditions. We provide quantitative results demonstrating that images annotated by under-graduates are sufficient for training instance-aware segmentation architectures to efficiently segment complex fluorescence nuclear images.