Lifelong Text-Audio Sentiment Analysis learning.

Sentiment analysis refers to the mining of textual context, which is conducted with the aim of identifying and extracting subjective opinions in textual materials. However, most existing methods neglect other important modalities, e.g., the audio modality, which can provide intrinsic complementary knowledge for sentiment analysis. Furthermore, much work on sentiment analysis cannot continuously learn new sentiment analysis tasks or discover potential correlations among distinct modalities. To address these concerns, we propose a novel Lifelong Text-Audio Sentiment Analysis (LTASA) model to continuously learn text-audio sentiment analysis tasks, which effectively explores intrinsic semantic relationships from both intra-modality and inter-modality perspectives. More specifically, a modality-specific knowledge dictionary is developed for each modality to obtain shared intra-modality representations among various text-audio sentiment analysis tasks. Additionally, based on information dependence between text and audio knowledge dictionaries, a complementarity-aware subspace is developed to capture the latent nonlinear inter-modality complementary knowledge. To sequentially learn text-audio sentiment analysis tasks, a new online multi-task optimization pipeline is designed. Finally, we verify our model on three common datasets to show its superiority. Compared with some baseline representative methods, the capability of the LTASA model is significantly boosted in terms of five measurement indicators.

Cross-modality (CT-MRI) prior augmented deep learning for robust lung tumor segmentation from small MR datasets.

PURPOSE: Accurate tumor segmentation is a requirement for magnetic resonance (MR)-based radiotherapy. Lack of large expert annotated MR datasets makes training deep learning models difficult. Therefore, a cross-modality (MR-CT) deep learning segmentation approach that augments training data using pseudo MR images produced by transforming expert-segmented CT images was developed. METHODS: Eighty-one T2-weighted MRI scans from 28 patients with non-small cell lung cancers (nine with pretreatment and weekly MRI and the remainder with pre-treatment MRI scans) were analyzed. Cross-modality model encoding the transformation of CT to pseudo MR images resembling T2w MRI was learned as a generative adversarial deep learning network. This model was used to translate 377 expert segmented non-small cell lung cancer CT scans from the Cancer Imaging Archive into pseudo MRI that served as additional training set. This method was benchmarked against shallow learning using random forest, standard data augmentation, and three state-of-the art adversarial learning-based cross-modality data (pseudo MR) augmentation methods. Segmentation accuracy was computed using Dice similarity coefficient (DSC), Hausdorff distance metrics, and volume ratio. RESULTS: The proposed approach produced the lowest statistical variability in the intensity distribution between pseudo and T2w MR images measured as Kullback-Leibler divergence of 0.069. This method produced the highest segmentation accuracy with a DSC of (0.75 ± 0.12) and the lowest Hausdorff distance of (9.36 mm ± 6.00 mm) on the test dataset using a U-Net structure. This approach produced highly similar estimations of tumor growth as an expert (P = 0.37). CONCLUSIONS: A novel deep learning MR segmentation was developed that overcomes the limitation of learning robust models from small datasets by leveraging learned cross-modality information using a model that explicitly incorporates knowledge of tumors in modality translation to augment segmentation training. The results show the feasibility of the approach and the corresponding improvement over the state-of-the-art methods.