Search for: All records

Biomedical knowledge graphs (KGs) encode rich, structured information critical for drug discovery tasks, but extracting meaningful insights from large-scale KGs remains challenging due to their complex structure. Existing biomedical subgraph retrieval methods are tailored for graph neural networks (GNNs), limiting compatibility with other paradigms, including large language models (LLMs). We introduce K-Paths, a model-agnostic retrieval framework that extracts structured, diverse, and biologically meaningful multi-hop paths from dense biomedical KGs. These paths enable prediction of unobserved drug-drug and drug-disease interactions, including those involving entities not seen during training, thus supporting inductive reasoning. K-Paths is training-free and employs a diversity-aware adaptation of Yen's algorithm to extract the K shortest loopless paths between entities in a query, prioritizing biologically relevant and relationally diverse connections. These paths serve as concise, interpretable reasoning chains that can be directly integrated with LLMs or GNNs to improve generalization, accuracy, and enable explainable inference. Experiments on benchmark datasets show that K-Paths improves zero-shot reasoning across state-of-the-art LLMs. For instance, Tx-Gemma 27B improves by 19.8 and 4.0 F1 points on interaction severity prediction and drug repurposing tasks, respectively. Llama 70B achieves gains of 8.5 and 6.2 points on the same tasks. K-Paths also boosts the training efficiency of EmerGNN, a state-of-the-art GNN, by reducing the KG size by 90% while maintaining predictive performance. Beyond efficiency, K-Paths bridges the gap between KGs and LLMs, enabling scalable and explainable LLM-augmented scientific discovery. We release our code and the retrieved paths as a benchmark for inductive reasoning. 

Abstract PurposeTo determine the feasibility of simultaneous multi‐slice (SMS) real‐time MRI (RT‐MRI) at 0.55T for the evaluation of cardiac function. MethodsCardiac CINE MRI is routinely used to evaluate left‐ventricular (LV) function. The standard is sequential multi‐slice balanced SSFP (bSSFP) over a stack of short‐axis slices using electrocardiogram (ECG) gating and breath‐holds. SMS has been used in CINE imaging to reduce the number of breath‐holds by a factor of 2–4 at 1.5T, 3T, and recently at 0.55T. This work aims to determine if SMS is similarly effective in the RT‐MRI evaluation of cardiac function. We used an SMS bSSFP pulse sequence with golden‐angle spirals at 0.55T with an SMS factor of three. We cover the LV with three acquisitions for SMS, and nine for single‐band (SB). Imaging was performed on 9 healthy volunteers and 1 patient with myocardial fibrosis and sternal wires. A spatio‐temporal constrained reconstruction is used, with regularization parameters selected by a board‐certified cardiologist. Images were quantitatively analyzed with a normalized contrast and an Edge Sharpness (ES) score. ResultsThere was a statistically significant 2‐fold difference in contrast between SMS and SB and no significant difference in ES score. The contrast for SMS and SB were 13.38/29.05 at mid‐diastole and 10.79/22.26 at end‐systole; the ES scores for SMS and SB were 1.77/1.83 at mid‐diastole and 1.50/1.72 at end‐systole. ConclusionsSMS cardiac RT‐MRI at 0.55T is feasible and provides sufficient blood‐myocardium contrast to evaluate LV function in three slices simultaneously without any gating or periodic motion assumptions. 

Abstract PurposeTo develop a small‐tip multidimensional RF pulse design procedure that incorporates linear time‐invariant gradient imperfections and concomitant field effects. This could be particularly important for contemporary low‐field MRI systems with high‐performance gradients. Theory and MethodsWe developed an extension of the small‐tip excitation k‐space formalism, where concomitant fields were approximated as a Bloch‐Siegert shift in the rotating frame. This was evaluated using realistic simulations of 2D selective excitation at various field strengths (0.2T, 0.55T, 1.5T, 3T, and 7T) with single and parallel transmit. Simulated excitation profiles from the original and extended k‐space formalisms were compared. Experimental validations were performed at 0.55T with a single‐channel transmit. ResultsThe extended formalism provides improved 2D excitation profiles in all scenarios simulated, compared against the original formalism. The proposed method corrects the concomitant field effects on 2D selective excitations forB₀ > 0.2T when the magnitude of theB₀is far larger than that of nonrotating concomitant fields. Simulation and phantom experiments at 0.55T match well for both original and proposed methods, with the proposed method providing sharper and more accurate excitation profiles at off‐isocenter distances up to 15 cm. The impact of the proposed method is greatest in scenarios where concomitant fields are substantial, such as low field strengths and off‐isocenter. ConclusionConcomitant fields can be modeled as a Bloch‐Siegert shift in the rotating frame during multidimensional RF pulse design, resulting in improved excitation profiles with sharp edges. This is important to consider for off‐isocenter excitations and imaging at low field strengths with strong gradients. 

Abstract This study investigates the influence of land surface processes on short-spell monsoonal heavy rainfall events under varying soil wetness conditions in India, using the Weather Research and Forecasting Model coupled with two land surface schemes: Noah and SLAB. To represent contrasting soil conditions, four rainfall events are chosen, two in onset (June) and two in active (August) months, during the Indian summer monsoon season. The results indicate that rainfall sensitivity differs notably between onset and active cases. Specifically, in onset, the SLAB overpredicts rainfall to the north of the storm compared to the Noah. The northward displacement of rainfall is attributed to the sensitivity of evapotranspiration to the preferential soil moisture regime in onset. Furthermore, the higher surface air saturation deficit in onset favors plant transpiration, resulting in increased boundary layer moisture. This contributes to enhanced moist static energy, thereby affecting potential vorticity and precipitation. In contrast, evapotranspiration sensitivity is modest during active months, under wet soil and lower surface air saturation deficit conditions. The study reveals the distinct soil moisture feedback mechanisms during the onset and active phases, through variations in evapotranspiration sensitivity. Variations in soil moisture and surface air saturation deficit in these phases play a crucial role in modulating evapotranspiration, which in turn affects precipitation. These findings underscore the importance of land surface initialization and land data assimilation in land–atmosphere interaction studies.