Hybrid cnn and lstm model improves biomedical extractive summaries
A Hybrid Hierarchical 1D-CNN-BiLSTM Framework for Extractive Summarization of Biomedical and Clinical Text
Computation and Language
Summary
Summarizing medical texts accurately is tough because making up facts can be dangerous. The authors designed a method that picks important sentences directly from medical documents instead of creating new sentences. They used a mix of two types of neural networks to understand sentences and how they relate inside the text. This approach helps avoid errors and works well on some medical data sets, especially on unstructured text. However, it is less effective on very structured reports that follow strict templates.
What this means in practice
- •For clinical documentation teams: Automatically select key sentences from clinical narratives to create factually accurate summaries without generating new text.
- •For healthcare informatics engineers: Integrate hybrid CNN-LSTM extractive summarization to improve reliability of summary outputs in biomedical text processing pipelines.
Authors
Saad Bin Ather, Muhammad Saif, Ali Hassan Khan, Manzer Abbas, Hajra Waheed
Abstract
Large language models have made abstractive summarization remarkably fluent, but generated summaries can hallucinate facts, posing serious risks in biomedical and clinical domains. We address this by removing generation from the pipeline and framing summarization as extractive sentence selection. Our Hybrid Hierarchical CNN-LSTM Summarizer uses stacked multi-kernel convolutions to compose sentence-level embeddings into richer inter-sentence representations, followed by a bidirectional LSTM to model long-range dependencies across the document. A lightweight scoring head assigns per-sentence importance scores and is trained end-to-end with binary cross-entropy against oracle extractive labels. At inference, a dynamic mean-plus-standard-deviation threshold with a top-3 fallback selects sentences directly from the source and chronologically reorders them into the final summary. Since every output sentence is copied from the input, the model avoids generation-induced factual drift. On PubMed, our architecture outperforms isolated CNN and LSTM baselines, while ablations show that wider convolutional receptive fields improve sentence scoring. On MIMIC-CXR and MIMIC-IV BHC, the model performs well on unstructured narratives but defaults toward positional baselines on highly templated reports. These results suggest that structural constraints can provide a reliable path toward factually grounded summarization systems that are trustworthy by design rather than by correction.