Search

Deep neural networks have recently been succeessful in the field of medical image segmentation; however, they are typically subject to performance degradation problems when well-trained models are tested in another new domain with different data distributions. Given that annotated cross-domain images may inaccessible, unsupervised domain adaptation methods that transfer learnable information from annotated source domains to unannotated target domains with different distributions have attracted substantial attention. Many methods leverage image-level or pixel-level translation networks to align domain-invariant information and mitigate domain shift issues. However, These methods rarely perform well when there is a large domain gap. A new unsupervised deep consistency learning adaptation network, which adopts input space consistency learning and output space consistency learning to realize unsupervised domain adaptation and cardiac structural segmentation, is introduced in this paper The framework mainly includes a domain translation path and a cross-modality segmentation path. In domain translation path, a symmetric alignment generator network with attention to cross-modality features and anatomy is introduced to align bidirectional domain features. In the segmentation path, entropy map minimization, output probability map minimization and segmentation prediction minimization are leveraged to align the output space features. The model conducts supervised learning to extract source domain features and conducts unsupervised deep consistency learning to extract target domain features. Through experimental testing on two challenging cross-modality segmentation tasks, our method has robust performance compared to that of previous methods. Furthermore, ablation experiments are conducted to confirm the effectiveness of our framework. [ABSTRACT FROM AUTHOR]

Mobile intelligent multimedia presentation systems are limited by various constraints including mobile network characteristics, mobile device capabilities, and user preferences. Those presentation systems which incorporate remotely stored multimedia rely on the bandwidth which is available during actual content transmission on the connecting mobile network. One approach to deal with this is to transcode content, thus reducing its data rate requirement, although this technique is inherently limited by the lowest acceptable quality of that media element. Alternatively, content can be transmoded to different modalities with lower bandwidth requirements. TeleMorph, a cross-modality adaptation control platform is detailed in this paper. The main premise of TeleMorph is that cross-modality adaptation decisions in mobile presentation systems must occur with primary consideration for bandwidth fluctuations. TeleMorph has been implemented as a network-aware fuzzy inference system that controls cross-modality adaptations between multimodal audio-video, audio-images, and images-text presentations, as well as unimodal audio and text presentations. Initially, a brief introduction to Intelligent Multimedia and Mobile Intelligent Multimedia is given, and related systems discussed. TeleTuras, a tourist information application implemented as a testbed for TeleMorph, is utilized for objective and subjective evaluations through its integration of six disparate test scenarios that incorporate various media types and qualities. Positive results for each evaluation technique, based on specific metrics, are also presented. In addition, future work on TeleMorph is also detailed. [ABSTRACT FROM PUBLISHER]

Mobile intelligent multimedia presentation systems are subject to various resource constraints including mobile network characteristics, mobile device capabilities and user preferences. Those presentation systems which incorporate remote multimedia content accessed across HTTP (Hyper Text Transfer Protocol) or RTP (Real Time Transfer Protocol) protocols are particularly reliant on the capabilities of the connecting mobile network (i.e. minimum, average and maximum bandwidth) and in particular on the real-time constraints (i.e. currently available bandwidth, packet loss, bit error rate, latency) which prevail during actual content transmission. One approach to address this is to scale content, thus reducing its datarate requirement, although this technique is inherently limited by the lowest acceptable quality of that media element. Alternatively, content can be converted from one modality to another with a lower resource requirement. TeleMorph, a cross-modality adaptation control platform is detailed here. Initially a brief introduction to Intelligent Multimedia and toMobile IntelligentMultimedia is given, and key systems discussed. The main premise of TeleMorph is that cross-modality adaptations in mobile presentation systems must be controlled in a manner which gives primary consideration to bandwidth fluctuations as well as the constraints listed above. The current prototype of TeleMorph, which uses a fuzzy inference system to control cross-modality adaptations between video and audio, is described. Particular focus is given to the fuzzy inputs, fuzzy control rules and fuzzy outputs which have been utilised in decision making. TeleTuras, a tourist information application which has been implemented as a testbed for TeleMorph, gives promising evaluation results based on multimedia and bandwidth specific test scenarios. TeleMorph is related to other approaches in the area of Mobile Intelligent Multimedia Presentation Systems. TeleMorph differs from other approaches in that it focuses specifically on the challenges posed by controlling bandwidth determined cross-modality adaptations in a mobile network environment. Future work on TeleMorph's output presentation composition will incorporate images and text also, thus allowing for extended adaptation between video, audio, images and text, as well as multimodal combinations of these media elements. [ABSTRACT FROM AUTHOR]