Keywords: Pediatric Tumours, Generative AI, Pathology, Language Models, Retrieval-Augmented Generation
Abstract: Generative Artificial Intelligence (AI), particularly large language models (LLMs), is redefining medical knowledge representation and access. In the field of pathology, especially pediatric tumour pathology, the structured and complex nature of diagnostic criteria offers a compelling use case for domain-specific generative AI solutions.
This work explores the development of AI-powered assistants tailored for pediatric tumour pathology using two complementary approaches: fine-tuning a small language model (SLM) and implementing a retrieval-augmented generation (RAG) framework. A curated knowledge base of expert-authored notes on pediatric tumours was prepared in two formats—prompt-completion pairs for fine-tuning and chunked documents for RAG indexing.
The fine-tuning pathway utilizes Gemma 3n, a lightweight and efficient open-source SLM recently introduced for healthcare-aligned generative tasks. This model is optimized to answer diagnostic, classification, and reasoning-based queries with high specificity. In parallel, the RAG-based system integrates a vector database to retrieve context-relevant information and generate responses using the same underlying model, offering traceability and flexibility without retraining.
Both approaches support applications in pathology education, diagnostic assistance, and structured knowledge retrieval. Fine-tuning ensures embedded expertise and rapid recall, while RAG allows for real-time updates and greater interpretability. These solutions can enhance accessibility to complex pediatric pathology knowledge, especially in training environments or low-resource settings.
This presentation outlines the architecture, implementation, and comparative strengths of both methods, emphasizing their clinical relevance, scalability, and potential for safe deployment in digital pathology ecosystems.
Biography: Dr. Atul Tiwari is an Assistant Professor of Pathology at Government Medical College, Chittorgarh (India), with a special interest in Artificial Intelligence in healthcare. He holds additional degrees in Artificial Intelligence and Computer Applications. He has delivered multiple talks and workshops across India and has also published several articles on AI applications in medical diagnostics, and actively develops real-world solutions integrating deep learning, LLMs, and workflow automation for clinical pathology.
#UCJournals #GenerativeAI #ArtificialIntelligence #MachineLearning #DeepLearning #NaturalLanguageProcessing #GenerativeLanguageModels #AIMedicine #AIinHealthcare #MedicalAI #DigitalPathology #ComputationalPathology #AIDrivenResearch #DataScienceInMedicine #PrecisionDiagnostics #ClinicalAI #PediatricOncology #PediatricPathology #ChildhoodCancer #PediatricTumors #CancerResearch #OncologyResearch #TumorPathology #Histopathology #MedicalPathology #CancerDiagnostics #OncologyAI #PediatricCancerCare #CancerAwareness #OncoPathology #OncologyInnovation #MedicalResearch #HealthTech #Bioinformatics #TranslationalMedicine #MedicalInnovation #HealthcareTechnology #AIForGood #FutureOfMedicine #DigitalHealth #SmartHealthcare #AIRevolutionInMedicine #ClinicalInnovation #HealthAI #MedicalTechnology #BiomedicalResearch #OpenAccessResearch #AcademicPublishing #MedicalJournals #HealthcareResearch #PathologyResearch #OncologyJournals #AIResearch #ResearchInnovation #GlobalHealthResearch
