Elecsium MVP
Improvements
- This end-to-end framework forms the foundation of Elecsium's MVP, enabling comprehensive analysis of multimodal healthcare and life sciences (HCLS) data through AWS services like HealthOmics, HealthLake, and HealthImaging, as well as machine learning and analytics tools such as SageMaker, Athena, and QuickSight. It seamlessly ingests raw HCLS data formats, including VCF, FHIR, and DICOM, supporting a zero-ETL architecture for scalable data analysis on AWS. The framework is designed to store, transform, and analyze linked genomic, clinical, and medical imaging data, demonstrated on a synthetic patient dataset with diverse disease scenarios from MITRE. Additionally, it includes an ML model to predict patient outcomes and an interactive dashboard to visualize data summaries and model results, customizable for specific user needs. Reference
- This architecture will be extended with a knowledge-graph based retrieval augmented generation (GraphRAG) platform.
AWS HealthLake, HealthOmics, and HealthImaging
Improvements
- Elecsium MVP is built on top of AWS HealthLake, HealthOmics, and HealthImaging.
Multimodal Radiological Diagnosis Decision Support Platform
Improvements
- Elecsium is creating a multimodal radiological diagnosis decision support platform powered by Meta's Llama 3.2 90 billion parameters model.
- Llama 3.2’s ability to handle multimodal inputs means it can integrate and analyze data from various sources simultaneously—like combining patient medical histories and imaging data to enhance diagnostic accuracy.
- User: What do images A, B, and C represent?
- Llama-3.2-90B-Vision-Instruct-Turbo: Images A, B, and C represent different imaging modalities used to visualize the same anatomical structure. Image A is a radiograph (X-ray) of the abdomen, which shows the skeletal structure and some soft tissue details. The arrow points to a small, round, radiopaque object in the right upper quadrant of the abdomen. Image B is an ultrasound image of the same area, which provides more detailed information about the soft tissues. The arrow points to a similar object as in Image A, but with better resolution and texture. Image C is a computed tomography (CT) scan of the same area, which offers even more detailed information about the internal structures. The arrow points to the same object as in Images A and B, but with higher resolution and contrast. Based on the location and appearance of the object in all three images, it is likely that it represents a gallstone in the gallbladder. Gallstones are small, hard deposits that can form in the gallbladder and cause symptoms such as pain, nausea, and vomiting. They can be diagnosed using various imaging modalities, including X-rays, ultrasound, and CT scans.
Knowledge Graph based RAG platform for Cancer Biology
Improvements
- Elecsium is creating a comprehensive knowledge graph based Retrieval Augmented Generation (RAG) platform for cancer biology.
- Our knowledge graph based RAG platform is powered by Neo4j's AuraDB.
