Explore the vast range of titles available.

Challenges exist with the adoption of Fast Healthcare Interoperability Resources (FHIR) within analytics, including the difficulty in transforming complex data structures, and performance issues when querying large datasets in their native JSON or XML formats. In 2023, an international working group began work on a solution to this problem that would be easier to implement than existing approaches. Over the course of 18 months, the group authored a new specification and validated it through the development and testing of multiple independent implementations. The outcome of this work is a standard, implementation-agnostic method for defining views that produce tabular data from FHIR resources. SQL on FHIR view definitions can be written to cover common use cases and can be executed across a variety of technology platforms. We evaluate the feasibility of this approach by replicating findings from an existing study across multiple view runner and database implementations, demonstrating portability and consistency.

Virtual Ligand Screening (VLS) has become an integral part of the drug discovery process for many pharmaceutical companies. Ligand similarity searches provide a very powerful method of screening large databases of ligands to identify possible hits. If these hits belong to new chemotypes the method is deemed even more successful. eHiTS LASSO uses a new interacting surface point types (ISPT) molecular descriptor that is generated from the 3D structure of the ligand, but unlike most 3D descriptors it is conformation independent. Combined with a neural network machine learning technique, LASSO screens molecular databases at an ultra fast speed of 1 million structures in under 1 min on a standard PC. The results obtained from eHiTS LASSO trained on relatively small training sets of just 2, 4 or 8 actives are presented using the diverse directory of useful decoys (DUD) dataset. It is shown that over a wide range of receptor families, eHiTS LASSO is consistently able to enrich screened databases and provides scaffold hopping ability.

We introduce MedGemma 1.5 4B, the latest model in the MedGemma collection. MedGemma 1.5 expands on MedGemma 1 by integrating additional capabilities: high-dimensional medical imaging (CT/MRI volumes and histopathology whole slide images), anatomical localization via bounding boxes, multi-timepoint chest X-ray analysis, and improved medical document understanding (lab reports, electronic health records). We detail the innovations required to enable these modalities within a single architecture, including new training data, long-context 3D volume slicing, and whole-slide pathology sampling. Compared to MedGemma 1 4B, MedGemma 1.5 4B demonstrates significant gains in these new areas, improving 3D MRI condition classification accuracy by 11% and 3D CT condition classification by 3% (absolute improvements). In whole slide pathology imaging, MedGemma 1.5 4B achieves a 47% macro F1 gain. Additionally, it improves anatomical localization with a 35% increase in Intersection over Union on chest X-rays and achieves a 4% macro accuracy for longitudinal (multi-timepoint) chest x-ray analysis. Beyond its improved multimodal performance over MedGemma 1, MedGemma 1.5 improves on text-based clinical knowledge and reasoning, improving by 5% on MedQA accuracy and 22% on EHRQA accuracy. It also achieves an average of 18% macro F1 on 4 different lab report information extraction datasets (EHR Datasets 2, 3, 4, and Mendeley Clinical Laboratory Test Reports). Taken together, MedGemma 1.5 serves as a robust, open resource for the community, designed as an improved foundation on which developers can create the next generation of medical AI systems. Resources and tutorials for building upon MedGemma 1.5 can be found at https://goo.gle/MedGemma.