skip to main content

Title: Star cluster classification in the PHANGS– HST survey: Comparison between human and machine learning approaches
ABSTRACT When completed, the PHANGS–HST project will provide a census of roughly 50 000 compact star clusters and associations, as well as human morphological classifications for roughly 20 000 of those objects. These large numbers motivated the development of a more objective and repeatable method to help perform source classifications. In this paper, we consider the results for five PHANGS–HST galaxies (NGC 628, NGC 1433, NGC 1566, NGC 3351, NGC 3627) using classifications from two convolutional neural network architectures (RESNET and VGG) trained using deep transfer learning techniques. The results are compared to classifications performed by humans. The primary result is that the neural network classifications are comparable in quality to the human classifications with typical agreement around 70 to 80 per cent for Class 1 clusters (symmetric, centrally concentrated) and 40 to 70 per cent for Class 2 clusters (asymmetric, centrally concentrated). If Class 1 and 2 are considered together the agreement is 82 ± 3 per cent. Dependencies on magnitudes, crowding, and background surface brightness are examined. A detailed description of the criteria and methodology used for the human classifications is included along with an examination of systematic differences between PHANGS–HST and LEGUS. The distribution of data points in a colour–colour diagram is used as a ‘figure of merit’ to further test the relative performances of the different methods. The effects on science results (e.g. determinations of mass and age functions) of using different cluster classification methods are examined and found to be minimal.  more » « less
Award ID(s):
Author(s) / Creator(s):
; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; more » ; ; ; ; « less
Date Published:
Journal Name:
Monthly Notices of the Royal Astronomical Society
Page Range / eLocation ID:
5294 to 5317
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ABSTRACT We present the results of a proof-of-concept experiment that demonstrates that deep learning can successfully be used for production-scale classification of compact star clusters detected in Hubble Space Telescope(HST) ultraviolet-optical imaging of nearby spiral galaxies ($D\lesssim 20\, \textrm{Mpc}$) in the Physics at High Angular Resolution in Nearby GalaxieS (PHANGS)–HST survey. Given the relatively small nature of existing, human-labelled star cluster samples, we transfer the knowledge of state-of-the-art neural network models for real-object recognition to classify star clusters candidates into four morphological classes. We perform a series of experiments to determine the dependence of classification performance on neural network architecture (ResNet18 and VGG19-BN), training data sets curated by either a single expert or three astronomers, and the size of the images used for training. We find that the overall classification accuracies are not significantly affected by these choices. The networks are used to classify star cluster candidates in the PHANGS–HST galaxy NGC 1559, which was not included in the training samples. The resulting prediction accuracies are 70 per cent, 40 per cent, 40–50 per cent, and 50–70 per cent for class 1, 2, 3 star clusters, and class 4 non-clusters, respectively. This performance is competitive with consistency achieved in previously published human and automated quantitative classification of star cluster candidate samples (70–80 per cent, 40–50 per cent, 40–50 per cent, and 60–70 per cent). The methods introduced herein lay the foundations to automate classification for star clusters at scale, and exhibit the need to prepare a standardized data set of human-labelled star cluster classifications, agreed upon by a full range of experts in the field, to further improve the performance of the networks introduced in this study. 
    more » « less

    We have identified 189 candidate z > 1.3 protoclusters and clusters in the LSST Deep Drilling Fields. This sample will enable the measurement of the metal enrichment and star formation history of clusters during their early assembly period through the direct measurement of the rate of supernovae identified through the LSST. The protocluster sample was selected from galaxy overdensities in a Spitzer/IRAC colour-selected sample using criteria that were optimized for protocluster purity using a realistic light-cone. Our tests reveal that $60\!-\!80~{{\ \rm per\ cent}}$ of the identified candidates are likely to be genuine protoclusters or clusters, which is corroborated by a ∼4σ stacked X-ray signal from these structures. We provide photometric redshift estimates for 47 candidates which exhibit strong peaks in the photo-z distribution of their candidate members. However, the lack of a photo-z peak does not mean a candidate is not genuine, since we find a stacked X-ray signal of similar significance from both the candidates that exhibit photo-z peaks and those that do not. Tests on the light-cone reveal that our pursuit of a pure sample of protoclusters results in that sample being highly incomplete ($\sim 4~{{\ \rm per\ cent}}$) and heavily biased towards larger, richer, more massive, and more centrally concentrated protoclusters than the total protocluster population. Most ($\sim 75~{{\ \rm per\ cent}}$) of the selected protoclusters are likely to have a maximum collapsed halo mass of between 1013 and 1014 M⊙, with only $\sim 25~{{\ \rm per\ cent}}$ likely to be collapsed clusters above 1014 M⊙. However, the aforementioned bias ensures our sample is $\sim 50~{{\ \rm per\ cent}}$ complete for structures that have already collapsed into clusters more massive than 1014 M⊙.

    more » « less

    We analyse Gaia EDR3 and re-calibrated HST proper motion data from the core-collapsed and non-core-collapsed globular clusters NGC 6397 and NGC 3201, respectively, with the Bayesian mass-orbit modelling code MAMPOSSt-PM. We use Bayesian evidence and realistic mock data sets constructed with Agama to select between different mass models. In both clusters, the velocities are consistent with isotropy within the extent of our data. We robustly detect a dark central mass (DCM) of roughly $1000\, \rm M_\odot$ in both clusters. Our MAMPOSSt-PM fits strongly prefer an extended DCM in NGC 6397, while only presenting a mild preference for it in NGC 3201, with respective sizes of a roughly one and a few per cent of the cluster effective radius. We explore the astrophysics behind our results with the CMC Monte Carlo N-body code, whose snapshots best matching the phase space observations lead to similar values for the mass and size of the DCM. The internal kinematics are thus consistent with a population of hundreds of massive white dwarfs in NGC 6397, and roughly 100 segregated stellar-mass black holes in NGC 3201, as previously found with CMC. Such analyses confirm the accuracy of both mass-orbit modelling and Monte Carlo N-body techniques, which together provide more robust predictions on the DCM of globular clusters (core-collapsed or not). This opens possibilities to understand a vast range of interesting astrophysical phenomena in clusters, such as fast radio bursts, compact object mergers, and gravitational waves.

    more » « less

    We study environmental quenching using the spatial distribution of current star formation and stellar population ages with the full SAMI Galaxy Survey. By using a star formation concentration index [C-index, defined as log10(r50, H α/r50, cont)], we separate our sample into regular galaxies (C-index ≥−0.2) and galaxies with centrally concentrated star formation (SF-concentrated; C-index <−0.2). Concentrated star formation is a potential indicator of galaxies currently undergoing ‘outside-in’ quenching. Our environments cover ungrouped galaxies, low-mass groups (M200 ≤ 1012.5M⊙), high-mass groups (M200 in the range 1012.5–14 M⊙) and clusters (M200 > 1014M⊙). We find the fraction of SF-concentrated galaxies increases as halo mass increases by 9 ± 2 per cent, 8 ± 3 per cent, 19 ± 4 per cent, and 29 ± 4 per cent for ungrouped galaxies, low-mass groups, high-mass groups, and clusters, respectively. We interpret these results as evidence for ‘outside-in’ quenching in groups and clusters. To investigate the quenching time-scale in SF-concentrated galaxies, we calculate light-weighted age (AgeL) and mass-weighted age (AgeM) using full spectral fitting, as well as the Dn4000 and HδA indices. We assume that the average galaxy age radial profile before entering a group or cluster is similar to ungrouped regular galaxies. At large radius (1–2 Re), SF-concentrated galaxies in high-mass groups have older ages than ungrouped regular galaxies with an age difference of 1.83 ± 0.38 Gyr for AgeL and 1.34 ± 0.56 Gyr for AgeM. This suggests that while ‘outside-in’ quenching can be effective in groups, the process will not quickly quench the entire galaxy. In contrast, the ages at 1–2 Re of cluster SF-concentrated galaxies and ungrouped regular galaxies are consistent (difference of 0.19 ± 0.21 Gyr for AgeL, 0.40 ± 0.61 Gyr for AgeM), suggesting the quenching process must be rapid.

    more » « less
  5. null (Ed.)
    Introduction: Vaso-occlusive crises (VOCs) are a leading cause of morbidity and early mortality in individuals with sickle cell disease (SCD). These crises are triggered by sickle red blood cell (sRBC) aggregation in blood vessels and are influenced by factors such as enhanced sRBC and white blood cell (WBC) adhesion to inflamed endothelium. Advances in microfluidic biomarker assays (i.e., SCD Biochip systems) have led to clinical studies of blood cell adhesion onto endothelial proteins, including, fibronectin, laminin, P-selectin, ICAM-1, functionalized in microchannels. These microfluidic assays allow mimicking the physiological aspects of human microvasculature and help characterize biomechanical properties of adhered sRBCs under flow. However, analysis of the microfluidic biomarker assay data has so far relied on manual cell counting and exhaustive visual morphological characterization of cells by trained personnel. Integrating deep learning algorithms with microscopic imaging of adhesion protein functionalized microfluidic channels can accelerate and standardize accurate classification of blood cells in microfluidic biomarker assays. Here we present a deep learning approach into a general-purpose analytical tool covering a wide range of conditions: channels functionalized with different proteins (laminin or P-selectin), with varying degrees of adhesion by both sRBCs and WBCs, and in both normoxic and hypoxic environments. Methods: Our neural networks were trained on a repository of manually labeled SCD Biochip microfluidic biomarker assay whole channel images. Each channel contained adhered cells pertaining to clinical whole blood under constant shear stress of 0.1 Pa, mimicking physiological levels in post-capillary venules. The machine learning (ML) framework consists of two phases: Phase I segments pixels belonging to blood cells adhered to the microfluidic channel surface, while Phase II associates pixel clusters with specific cell types (sRBCs or WBCs). Phase I is implemented through an ensemble of seven generative fully convolutional neural networks, and Phase II is an ensemble of five neural networks based on a Resnet50 backbone. Each pixel cluster is given a probability of belonging to one of three classes: adhered sRBC, adhered WBC, or non-adhered / other. Results and Discussion: We applied our trained ML framework to 107 novel whole channel images not used during training and compared the results against counts from human experts. As seen in Fig. 1A, there was excellent agreement in counts across all protein and cell types investigated: sRBCs adhered to laminin, sRBCs adhered to P-selectin, and WBCs adhered to P-selectin. Not only was the approach able to handle surfaces functionalized with different proteins, but it also performed well for high cell density images (up to 5000 cells per image) in both normoxic and hypoxic conditions (Fig. 1B). The average uncertainty for the ML counts, obtained from accuracy metrics on the test dataset, was 3%. This uncertainty is a significant improvement on the 20% average uncertainty of the human counts, estimated from the variance in repeated manual analyses of the images. Moreover, manual classification of each image may take up to 2 hours, versus about 6 minutes per image for the ML analysis. Thus, ML provides greater consistency in the classification at a fraction of the processing time. To assess which features the network used to distinguish adhered cells, we generated class activation maps (Fig. 1C-E). These heat maps indicate the regions of focus for the algorithm in making each classification decision. Intriguingly, the highlighted features were similar to those used by human experts: the dimple in partially sickled RBCs, the sharp endpoints for highly sickled RBCs, and the uniform curvature of the WBCs. Overall the robust performance of the ML approach in our study sets the stage for generalizing it to other endothelial proteins and experimental conditions, a first step toward a universal microfluidic ML framework targeting blood disorders. Such a framework would not only be able to integrate advanced biophysical characterization into fast, point-of-care diagnostic devices, but also provide a standardized and reliable way of monitoring patients undergoing targeted therapies and curative interventions, including, stem cell and gene-based therapies for SCD. Disclosures Gurkan: Dx Now Inc.: Patents & Royalties; Xatek Inc.: Patents & Royalties; BioChip Labs: Patents & Royalties; Hemex Health, Inc.: Consultancy, Current Employment, Patents & Royalties, Research Funding. 
    more » « less