arXiv subjects

Keith Moore

Publications and source records attributed to Keith Moore.

4 recordsLinked to original sources

arXiv2025

Not Quite Anything: Overcoming SAMs Limitations for 3D Medical Imaging

Foundation segmentation models such as SAM and SAM-2 perform well on natural images but struggle with brain MRIs where structures like the caudate and thalamus lack sharp boundaries and have low contrast. Rather than fine tune these models (for example MedSAM), we propose a compositional alternative where the foundation model output is treated as an additional input channel and passed alongside the MRI to highlight regions of interest. We generate SAM-2 prompts by using a lightweight 3D U-Net that was previously trained on MRI segmentation. The U-Net may have been trained on a different dataset, so its guesses are often imprecise but usually in the correct region. The edges of the resulting foundation model guesses are smoothed to improve alignment with the MRI. We also test prompt free segmentation using DINO attention maps in the same framework. This has-a architecture avoids modifying foundation weights and adapts to domain shift without retraining the foundation model. It reaches about 96 percent volume accuracy on basal ganglia segmentation, which is sufficient for our study of longitudinal volume change. The approach is fast, label efficient, and robust to out of distribution scans. We apply it to study inflammation linked changes in sudden onset pediatric OCD.

eess.IV↗

arXiv2025

Ask WhAI:Probing Belief Formation in Role-Primed LLM Agents

We present Ask WhAI, a systems-level framework for inspecting and perturbing belief states in multi-agent interactions. The framework records and replays agent interactions, supports out-of-band queries into each agent's beliefs and rationale, and enables counterfactual evidence injection to test how belief structures respond to new information. We apply the framework to a medical case simulator notable for its multi-agent shared memory (a time-stamped electronic medical record, or EMR) and an oracle agent (the LabAgent) that holds ground truth lab results revealed only when explicitly queried. We stress-test the system on a multi-specialty diagnostic journey for a child with an abrupt-onset neuropsychiatric presentation. Large language model agents, each primed with strong role-specific priors ("act like a neurologist", "act like an infectious disease specialist"), write to a shared medical record and interact with a moderator across sequential or parallel encounters. Breakpoints at key diagnostic moments enable pre- and post-event belief queries, allowing us to distinguish entrenched priors from reasoning or evidence-integration effects. The simulation reveals that agent beliefs often mirror real-world disciplinary stances, including overreliance on canonical studies and resistance to counterevidence, and that these beliefs can be traced and interrogated in ways not possible with human experts. By making such dynamics visible and testable, Ask WhAI offers a reproducible way to study belief formation and epistemic silos in multi-agent scientific reasoning.

cs.AI↗

arXiv2018

Quantitative theoretical analysis of lifetimes and decay rates relevant in laser cooling BaH

Tiny radiative losses below the 0.1% level can prove ruinous to the effective laser cooling of a molecule. In this paper the laser cooling of a hydride is studied with rovibronic detail using ab initio quantum chemistry in order to document the decays to all possible electronic states (not just the vibrational branching within a single electronic transition) and to identify the most populated final quantum states. The effect of spin-orbit and associated couplings on the properties of the lowest excited states of BaH are analysed in detail. The lifetimes of the A$^2{\Pi}_{1/2}$, H$^2{\Delta}_{3/2}$ and E$^2{\Pi}_{1/2}$ states are calculated (136 ns, 5.8 {\mu}s and 46 ns respectively) for the first time, while the theoretical value for B$^2{\Sigma}^+_{1/2}$ is in good agreement with experiments. Using a simple rate model the numbers of absorption-emission cycles possible for both one- and two-colour cooling on the competing electronic transitions are determined, and it is clearly demonstrated that the A$^2{\Pi}$ - X$^2{\Sigma}^+$ transition is superior to B$^2{\Sigma}^+$ - X$^2{\Sigma}^+$, where multiple tiny decay channels degrade its efficiency. Further possible improvements to the cooling method are proposed.

physics.chem-ph↗

arXiv2015

Towards a spectroscopically accurate set of potentials for heavy hydride laser cooling candidates: effective core potential calculations of BaH

BaH (and its isotopomers) is an attractive molecular candidate for laser cooling to ultracold temperatures and a potential precursor for the production of ultracold gases of hydrogen and deuterium. The theoretical challenge is to simulate the laser cooling cycle as reliably as possible and this paper addresses the generation of a highly accurate ab initio $^{2}\Sigma^+$ potential for such studies. The performance of various basis sets within the multi-reference configuration-interaction (MRCI) approximation with the Davidson correction (MRCI+Q) is tested and taken to the complete basis set limit. It is shown that the calculated molecular constants using a 46 electron Effective Core-Potential (ECP), the augmented polarized core-valence quintuplet basis set (aug-pCV5Z-PP) but only including three active electrons in the MRCI calculation are in close agreement with the available experimental values. The predicted dissociation energy D$_e$ for the X$^2\Sigma^+$ state (extrapolated to the complete basis set (CBS) limit) is 16895.12 cm$^{-1}$ (2.094 eV), which agrees within 0.1$\%$ of a revised experimental value of $<$16910.6 cm$^{-1}$, while the calculated r$_e$ is within 0.03 pm of the experimental result.

physics.atom-ph↗