knowledge_base/publications_detailed.md

Detailed Publications and Research Contributions

BioFusionNet (2024)

Full Title: "BioFusionNet: Deep Learning-Based Survival Risk Stratification in ER+ Breast Cancer Through Multifeature and Multimodal Data Fusion"

Journal: IEEE Journal of Biomedical and Health Informatics

Key Contributions:

• Novel multimodal fusion architecture combining histopathology, genomics, and clinical data
• Attention-based feature selection for interpretability
• Superior performance compared to existing methods
• Clinical validation on large patient cohorts

Technical Details:

• Uses ResNet-based feature extraction for histopathology images
• Implements cross-attention mechanisms for data fusion
• Employs survival analysis with Cox proportional hazards
• Achieves C-index of 0.78 on validation datasets

Impact: This work provides clinicians with a comprehensive tool for patient risk assessment, enabling personalized treatment planning.

GitHub: raktim-mondol/BioFusionNet

hist2RNA (2023)

Full Title: "hist2RNA: An Efficient Deep Learning Architecture to Predict Gene Expression from Breast Cancer Histopathology Images"

Journal: Cancers

Key Contributions:

• Direct prediction of gene expression from tissue images
• Efficient architecture suitable for clinical deployment
• Identification of morphology-gene expression relationships
• Validation across multiple cancer datasets

Technical Details:

• Custom CNN architecture optimized for gene expression prediction
• Multi-task learning framework
• Attention mechanisms for spatial feature importance
• Correlation analysis with known biological pathways

Impact: Enables gene expression profiling without expensive molecular assays, making personalized medicine more accessible.

GitHub: raktim-mondol/hist2RNA

AFExNet (2021)

Full Title: "AFExNet: An Adversarial Autoencoder for Differentiating Breast Cancer Sub-types and Extracting Biologically Relevant Genes"

Journal: IEEE/ACM Transactions on Computational Biology and Bioinformatics

Key Contributions:

• Adversarial training for robust feature learning
• Automatic biomarker discovery
• Cancer subtype classification
• Biologically interpretable features

Technical Details:

• Adversarial autoencoder architecture
• Gene selection based on reconstruction importance
• Validation on TCGA datasets
• Pathway enrichment analysis

Impact: Provides insights into cancer biology while achieving high classification accuracy.

GitHub: raktim-mondol/breast-cancer-sub-types

Ongoing Research

Multimodal Foundation Models

• Developing foundation models for medical imaging
• Pre-training on large-scale medical datasets
• Transfer learning for rare diseases

Ongoing Research

• Large Language Models (LLMs)
• Retrieval-Augmented Generation (RAG)
• Fine-tuning and domain adaptation

AI Ethics in Healthcare

• Bias detection and mitigation
• Fairness in medical AI
• Regulatory compliance frameworks