A0703
Title: A deep attention LSTM embedded aggregation network for multiple histopathological images
Authors: Sunghun Kim - Chungnam National University (Korea, South) [presenting]
Eunjee Lee - Chungnam National University (Korea, South)
Abstract: Recent computer vision and neural network advancements have facilitated medical imaging survival analysis for various medical applications. However, challenges arise when patients have multiple images from multiple lesions, as current deep-learning methods can provide multiple survival predictions for each patient, complicating result interpretation. To address this issue, a deep-learning survival model that can provide accurate predictions at the patient level was developed. A Deep Attention Long Short-Term Memory Embedded Aggregation Network (DALAN) is proposed for histopathology images, designed to perform feature extraction and aggregation of lesion images simultaneously. This design enables the model to learn imaging features from lesions efficiently and aggregate lesion-level information to the patient level. DALAN comprises a weight-shared CNN, attention layers, and LSTM layers. The attention layer calculates the significance of each lesion image, while the LSTM layer combines the weighted information to produce an all-encompassing representation of the patient's lesion data. DALAN was evaluated against several naive aggregation methods on simulated and real datasets in terms of the c-index. The results showed that DALAN outperformed the competing methods on the MNIST and Cancer dataset simulations. On the real TCGA dataset, DALAN also achieved a higher c-index of 0.803 compared to the naive methods and the competing models.
