GPDBN

GPDBN integrates genomic and pathological image data using bilinear deep neural networks to predict breast cancer prognosis.

Key Features:

• Integration of Multi-Modal Data: Combines genomic datasets and pathological images to leverage complementary molecular and morphological information.
• Inter-modality Bilinear Encoding: Uses an inter-modality bilinear feature encoding module to model interactions between genomic and pathological modalities.
• Intra-modality Bilinear Encoding: Incorporates two intra-modality bilinear feature encoding modules to capture relations within each modality.
• Unified Multi-layer Deep Neural Network: Aggregates encoded inter- and intra-modality features via a multi-layer deep neural network for prognosis prediction.

Scientific Applications:

• Breast Cancer Prognosis Prediction: Predicts breast cancer prognosis from integrated genomic and pathological image features.
• Personalized Medicine: Informs therapeutic decision-making by modeling patient-specific multi-modal profiles.
• Computational Oncology Research: Facilitates research on multi-modal data integration and prognostic modeling in oncology.

Methodology:

Encode features from genomic and pathological datasets using bilinear networks; model inter-modality relations with an inter-modality bilinear feature encoding module; capture within-modality relations with two intra-modality bilinear feature encoding modules; combine encoded features using a multi-layer deep neural network to produce prognosis predictions.