Search

In health, most large language model (LLM) research has focused on clinical tasks. However, mobile and wearable devices, which are rarely integrated into such tasks, provide rich, longitudinal data for personal health monitoring. Here we present Personal Health Large Language Model (PH-LLM), fine-tuned from Gemini for understanding and reasoning over numerical time-series personal health data. We created and curated three datasets that test 1) production of personalized insights and recommendations from sleep patterns, physical activity, and physiological responses, 2) expert domain knowledge, and 3) prediction of self-reported sleep outcomes. For the first task we designed 857 case studies in collaboration with domain experts to assess real-world scenarios in sleep and fitness. Through comprehensive evaluation of domain-specific rubrics, we observed that Gemini Ultra 1.0 and PH-LLM are not statistically different from expert performance in fitness and, while experts remain superior for sleep, fine-tuning PH-LLM provided significant improvements in using relevant domain knowledge and personalizing information for sleep insights. We evaluated PH-LLM domain knowledge using multiple choice sleep medicine and fitness examinations. PH-LLM achieved 79% on sleep and 88% on fitness, exceeding average scores from a sample of human experts. Finally, we trained PH-LLM to predict self-reported sleep quality outcomes from textual and multimodal encoding representations of wearable data, and demonstrate that multimodal encoding is required to match performance of specialized discriminative models. Although further development and evaluation are necessary in the safety-critical personal health domain, these results demonstrate both the broad knowledge and capabilities of Gemini models and the benefit of contextualizing physiological data for personal health applications as done with PH-LLM., Comment: 72 pages

We present the CO(3-2) APEX survey at 6 pc resolution of the bar of the SMC. We aboard the CO analysis in the SMC-Bar comparing the CO(3-2) survey with that of the CO(2-1) of similar resolution. We study the CO(3-2)-to-CO(2-1) ratio (R32) that is very sensitive to the environment properties (e.g., star-forming regions). We analyzed the correlation of this ratio with observational quantities that trace the star formation as the local CO emission, the Spitzer color [70/160], and the total IR surface brightness measured from the Spitzer and Herschel bands. For the identification of the CO(3-2) clouds, we used the CPROPS algorithm, which allowed us to measure the physical properties of the clouds. We analyzed the scaling relationships of such physical properties. We obtained an R32 of 0.65 as a median value for the SMC, with a standard deviation of 0.3. We found that R32 varies from region to region, depending on the star formation activity. In regions dominated by HII and photo-dissociated regions (e.g., N22, N66), R32 tends to be higher than the median values. Meanwhile, lower values were found toward quiescent clouds. We also found that R32 correlates positively with the IR color [70/160] and the total IR surface brightness. This finding indicates that R32 increases with environmental properties like the dust temperature, the total gas density, and the radiation field. We have identified 225 molecular clouds with sizes R > 1.5 pc and signal-to-noise (S/N) ratio > 3 and only 17 well-resolved CO(3-2) clouds increasing the S/N ratio to > 5. These 17 clouds follow consistent scaling relationships to the inner Milky Way clouds but with some departure. The CO(3-2) tends to be less turbulent and less luminous than the inner Milky Way clouds of similar size. Finally, we estimated a median virial-based CO-to-H2 conversion factor of 12.6_{-7}^{+10} Msun/(K km s^{-1} pc^{2}) for the total sample.

High-spatial-resolution HI observations have led to the realisation that the nearby (within few hundreds of parsecs) Galactic atomic filamentary structures are aligned with the ambient magnetic field. Enabled by the high quality data from the Australian Square Kilometre Array Pathfinder (ASKAP) radio telescope for the Galactic ASKAP HI (GASKAP-HI) survey, we investigate the potential magnetic alignment of the $\gtrsim 10\,{\rm pc}$-scale HI filaments in the Small Magellanic Cloud (SMC). Using the Rolling Hough Transform (RHT) technique that automatically identifies filamentary structures, combined with our newly devised ray-tracing algorithm that compares the HI and starlight polarisation data, we find that the HI filaments in the northeastern end of the SMC main body ("Bar" region) and the transition area between the main body and the tidal feature ("Wing" region) appear preferentially aligned with the magnetic field traced by starlight polarisation. Meanwhile, the remaining SMC volume lacks starlight polarisation data of sufficient quality to draw any conclusions. This suggests for the first time that filamentary HI structures can be magnetically aligned across a large spatial volume ($\gtrsim\,{\rm kpc}$) outside of the Milky Way. In addition, we generate maps of the preferred orientation of HI filaments throughout the entire SMC, revealing the highly complex gaseous structures of the galaxy likely shaped by a combination of the intrinsic internal gas dynamics, tidal interactions, and star formation feedback processes. These maps can further be compared with future measurements of the magnetic structures in other regions of the SMC., Comment: 24 pages, MNRAS accepted

The Small Magellanic Cloud (SMC) is the closest low-metallicity galaxy to the Milky Way where the dynamical state of molecular clouds can be analyzed. We present a CO(2-1) survey at 9 pc resolution obtained with the APEX telescope in an extensive region of the SMC and characterize the properties of the molecular clouds. We study the dynamical state and stability of these clouds uniformly. We identify 177 molecular clouds within the SMC by using CPROPS, of which 124 clouds are fully resolved with signal-to-noise ratio $>$ 5. The scaling relationships show that the SMC clouds are (on average) less turbulent and less luminous than their inner Milky Way counterparts of similar size by a factor of 2 and 3, respectively, while for a fixed linewidth, the SMC clouds are over-luminous by a factor of 3.5. Using the virial masses, we derive a CO-to-H2 conversion factor for the SMC CO clouds of 10.5 M$_{\odot}(K km s^{-1} pc^{2})^{-1}$, measured at 9 pc resolution. We also determine a dust-based conversion factor of 28 M$_{\odot}(K km s^{-1} pc^{2})^{-1}$, obtained at 12 pc resolution. We find that the SMC clouds appear to be in approximate gravitational virial equilibrium. We find that the cumulative mass functions based on both the luminous mass and the virial mass are steeper than $\frac{dN}{dM} \propto M^{-2}$, suggesting that most of the molecular mass of the SMC is contained in low-mass clouds., Comment: 22 pages, 18 figures. Submitted to the Astronomy & Astrophysics (A&A) journal

We present results of a wide-field (approximately 60 x 90 pc) ALMA mosaic of CO(2-1) and $^{13}$CO(2-1) emission from the molecular cloud associated with the 30 Doradus star-forming region. Three main emission complexes, including two forming a bowtie-shaped structure extending northeast and southwest from the central R136 cluster, are resolved into complex filamentary networks. Consistent with previous studies, we find that the central region of the cloud has higher line widths at fixed size relative to the rest of the molecular cloud and to other LMC clouds, indicating an enhanced level of turbulent motions. However, there is no clear trend in gravitational boundedness (as measured by the virial parameter) with distance from R136. Structures observed in $^{13}$CO are spatially coincident with filaments and are close to a state of virial equilibrium. In contrast, CO structures vary greatly in virialization, with low CO surface brightness structures outside of the main filamentary network being predominantly unbound. The low surface brightness structures constitute ~10% of the measured CO luminosity; they may be shredded remnants of previously star-forming gas clumps, or alternatively the CO-emitting parts of more massive, CO-dark structures., Comment: 24 pages; published in ApJ on 15 Jun 2022; associated data can be found at https://doi.org/10.13012/B2IDB-1671495_V1

We present the most sensitive and detailed view of the neutral hydrogen (HI) emission associated with the Small Magellanic Cloud (SMC), through the combination of data from the Australian Square Kilometre Array Pathfinder (ASKAP) and Parkes (Murriyang), as part of the Galactic Australian Square Kilometre Array Pathfinder (GASKAP) pilot survey. These GASKAP-HI pilot observations, for the first time, reveal HI in the SMC on similar physical scales as other important tracers of the interstellar medium, such as molecular gas and dust. The resultant image cube possesses an rms noise level of 1.1 K (1.6 mJy/beam) per 0.98 km s$^{-1}$ spectral channel with an angular resolution of 30$''$ ($\sim$10 pc). We discuss the calibration scheme and the custom imaging pipeline that utilizes a joint deconvolution approach, efficiently distributed across a computing cluster, to accurately recover the emission extending across the entire $\sim$25 deg$^2$ field-of-view. We provide an overview of the data products and characterize several aspects including the noise properties as a function of angular resolution and the represented spatial scales by deriving the global transfer function over the full spectral range. A preliminary spatial power spectrum analysis on individual spectral channels reveals that the power-law nature of the density distribution extends down to scales of 10 pc. We highlight the scientific potential of these data by comparing the properties of an outflowing high velocity cloud with previous ASKAP+Parkes HI test observations., Comment: Accepted for publication in PASA, 34 pages, 18 figures, 5 tables

Using the Australian Square Kilometre Array Pathfinder to measure 21-cm absorption spectra toward continuum background sources, we study the cool phase of the neutral atomic gas in the far outer disk, and in the inner Galaxy near the end of the Galactic bar at longitude 340 degrees. In the inner Galaxy the cool atomic gas has a smaller scale height than in the solar neighborhood, similar to the molecular gas and the superthin stellar population in the bar. In the outer Galaxy the cool atomic gas is mixed with the warm, neutral medium, with the cool fraction staying roughly constant with Galactic radius. The mean spin temperature, i.e. the ratio of the emission brightness temperature to the absorption, is roughly constant for velocities corresponding to Galactic radius greater than about twice the solar circle radius. The ratio has a value of about 300 K, but this does not correspond to a physical temperature in the gas. If the gas causing the absorption has kinetic temperature of about 100 K, as in the solar neighborhood, then the value 300 K indicates that the fraction of the gas mass in this phase is one-third of the total HI mass., Comment: 29 pages, 12 figures, submitted to Astrophysical Journal

The Galactic Plane Infrared Polarization Survey (GPIPS) seeks to characterize the magnetic field in the dusty Galactic disk using near-infrared stellar polarimetry. All GPIPS observations were completed using the 1.83 m Perkins telescope and Mimir instrument. GPIPS observations surveyed 76 sq-deg of the northern Galactic plane, from Galactic longitudes 18 to 56 deg and latitudes -1 to +1 deg, in the H-band (1.6 um). Surveyed stars span 7th to 16th mag, resulting in nearly 10 million stars with measured linear polarizations. Of these stars, ones with m_H < 12.5 mag and polarization percentage uncertainties under 2% were judged to be high quality and number over one million. GPIPS data reveal plane-of-sky magnetic field orientations for numerous interstellar clouds for AV values to ~30 mag. The average sky separation of stars with m_H < 12.5 mag is about 30arcsec, or about 60 per Planck polarization resolution element. Matching to Gaia DR2 showed the brightest GPIPS stars are red giants with distances in the 0.6-7.5 kpc range. Polarization orientations are mostly parallel to the Galactic disk, with some zones showing significant orientation departures. Changes in orientations are stronger as a function of Galactic longitude than of latitude. Considered at 10 arcmin angular scales, directions that show the greatest polarization fractions and narrowest polarization position angle distributions are confined to about ten large, coherent structures that are not correlated with star forming clouds. The GPIPS polarimetric and photometric data products (Data Release 4 catalogs and images) are publicly available for over 13million stars., Comment: ApJSupplements, in press

We report the direct detection of cold HI gas in a cloud ejected from the Small Magellanic Cloud (SMC) towards the Magellanic Stream. The cloud is part of a fragmented shell of HI gas on the outskirts of the SMC. This is the second direct detection of cold HI associated with the Magellanic Stream using absorption. The cold gas was detected using 21-cm HI absorption-line observations with the Australia Telescope Compact Array (ATCA) towards the extra-galactic source PMN J0029$-$7228. We find a spin (excitation) temperature for the gas of $68 \pm 20$ K. We suggest that breaking super shells from the Magellanic Clouds may be a source of cold gas to supply the rest of the Magellanic Stream., Comment: 9 pages, 5 figures, accepted for publication in MNRAS

We report the first evidence of molecular gas in two atomic hydrogen (HI) clouds associated with gas outflowing from the Small Magellanic Cloud (SMC). We used the Atacama Pathfinder Experiment (APEX) to detect and spatially resolve individual clumps of CO(2-1) emission in both clouds. CO clumps are compact (~ 10 pc) and dynamically cold (linewidths < 1 km/s). Most CO emission appears to be offset from the peaks of the HI emission, some molecular gas lies in regions without a clear HI counterpart. We estimate a total molecular gas mass of 10^3-10^4 Msun in each cloud and molecular gas fractions up to 30% of the total cold gas mass (molecular + neutral). Under the assumption that this gas is escaping the galaxy, we calculated a cold gas outflow rate of 0.3-1.8 Msun/yr and mass loading factors of 3 -12 at a distance larger than 1 kpc. These results show that relatively weak star-formation-driven winds in dwarf galaxies like the SMC are able to accelerate significant amounts of cold and dense matter and inject it into the surrounding environment., Comment: Accepted for publication in ApJ Letters

We use new high-resolution HI data from the Australian Square Kilometre Array Pathfinder (ASKAP) to investigate the dynamics of the Small Magellanic Cloud (SMC). We model the HI gas component as a rotating disc of non-negligible angular size, moving into the plane of the sky and undergoing nutation/precession motions. We derive a high-resolution (~ 10 pc) rotation curve of the SMC out to R ~ 4 kpc. After correcting for asymmetric drift, the circular velocity slowly rises to a maximum value of Vc ~ 55 km/s at R ~ 2.8 kpc and possibly flattens outwards. In spite of the SMC undergoing strong gravitational interactions with its neighbours, its HI rotation curve is akin to that of many isolated gas-rich dwarf galaxies. We decompose the rotation curve and explore different dynamical models to deal with the unknown three-dimensional shape of the mass components (gas, stars and dark matter). We find that, for reasonable mass-to-light ratios, a dominant dark matter halo with mass M(R<4 kpc) = 1-1.5 x 10^9 solar masses is always required to successfully reproduce the observed rotation curve, implying a large baryon fraction of 30%-40%. We discuss the impact of our assumptions and the limitations of deriving the SMC kinematics and dynamics from HI observations., Comment: Accepted for publication in MNRAS

Data diversity is critical to success when training deep learning models. Medical imaging data sets are often imbalanced as pathologic findings are generally rare, which introduces significant challenges when training deep learning models. In this work, we propose a method to generate synthetic abnormal MRI images with brain tumors by training a generative adversarial network using two publicly available data sets of brain MRI. We demonstrate two unique benefits that the synthetic images provide. First, we illustrate improved performance on tumor segmentation by leveraging the synthetic images as a form of data augmentation. Second, we demonstrate the value of generative models as an anonymization tool, achieving comparable tumor segmentation results when trained on the synthetic data versus when trained on real subject data. Together, these results offer a potential solution to two of the largest challenges facing machine learning in medical imaging, namely the small incidence of pathological findings, and the restrictions around sharing of patient data., Comment: Accepted for 2018 Workshop on Simulation and Synthesis in Medical Imaging - SASHIMI2018

We use the Spitzer SAGE survey of the Magellanic Clouds to evaluate the relationship between the 8-mic PAH emission, 24-mic hot dust emission, and HII region radiative transfer. We confirm that in the higher-metallicity Large Magellanic Cloud, PAH destruction is sensitive to optically thin conditions in the nebular Lyman continuum: objects identified as optically thin candidates based on nebular ionization structure show 6 times lower median 8-mic surface brightness (0.18 mJy arcsec^-2) than their optically thick counterparts (1.2 mJy arcsec^-2). The 24-mic surface brightness also shows a factor of 3 offset between the two classes of objects (0.13 vs 0.44 mJy arcsec^-2, respectively), which is driven by the association between the very small dust grains and higher density gas found at higher nebular optical depths. In contrast, PAH and dust formation in the low-metallicity Small Magellanic Cloud is strongly inhibited such that we find no variation in either 8-mic or 24-mic emission between our optically thick and thin samples. This is attributable to extremely low PAH and dust production together with high, corrosive UV photon fluxes in this low-metallicity environment. The dust mass surface densities and gas-to-dust ratios determined from dust maps using Herschel HERITAGE survey data support this interpretation., Comment: Accepted to ApJ, May 15, 2017. 10 pages, 9 figures