Cross-site prognosis prediction for nasopharyngeal carcinoma from incomplete multi-modal data.

Accurate prognosis prediction for nasopharyngeal carcinoma based on magnetic resonance (MR) images assists in the guidance of treatment intensity, thus reducing the risk of recurrence and death. To reduce repeated labor and sufficiently explore domain knowledge, aggregating labeled/annotated data from external sites enables us to train an intelligent model for a clinical site with unlabeled data. However, this task suffers from the challenges of incomplete multi-modal examination data fusion and image data heterogeneity among sites. This paper proposes a cross-site survival analysis method for prognosis prediction of nasopharyngeal carcinoma from domain adaptation viewpoint. Utilizing a Cox model as the basic framework, our method equips it with a cross-attention based multi-modal fusion regularization. This regularization model effectively fuses the multi-modal information from multi-parametric MR images and clinical features onto a domain-adaptive space, despite the absence of some modalities. To enhance the feature discrimination, we also extend the contrastive learning technique to censored data cases. Compared with the conventional approaches which directly deploy a trained survival model in a new site, our method achieves superior prognosis prediction performance in cross-site validation experiments. These results highlight the key role of cross-site adaptability of our method and support its value in clinical practice.

Time-Concentration Superposition for Linear Viscoelasticity of Polymer Solutions.

The concentration dependence of linear viscoelastic properties of polymer solutions is a well-studied topic in polymer physics. Dynamic scaling theories allow qualitative predictions of polymer solution rheology, but quantitative predictions are still limited to model polymers. Meanwhile, the scaling properties of non-model polymer solutions must be determined experimentally. In present paper, the time-concentration superposition (TCS) of experimental data is shown to be a robust procedure for studying the concentration scaling properties of binary and ternary polymer solutions. TCS can not only identify whether power law scaling may exist or not, and over which concentration range, but also unambiguously estimate the concentration scaling exponents of linear viscoelastic properties for a range of non-model polymer solutions.

Concentration Scaling on Linear Viscoelastic Properties of Cellular Suspensions and Effects of Equilibrium Phase Behavior.

Concentration scaling on linear viscoelastic properties of cellular suspensions has been studied by rheometric characterisation of Phormidium suspensions and human blood in a wide range of volume fraction under small amplitude oscillatory shear experiments. The rheometric characterisation results are analysed by the time-concentration superposition (TCS) principle and show a power law scaling of characteristic relaxation time, plateau modulus and the zero-shear viscosity over the concentration ranges studied. The results show that the concentration effect of Phormidium suspensions on their elasticity is much stronger than that of human blood due to its strong cellular interactions and a high aspect ratio. For human blood, no obvious phase transition could be observed over the range of hematocrits studied here and with respect to a high-frequency dynamic regime, only one concentration scaling exponent could be identified. For Phormidium suspensions with respect to a low-frequency dynamic regime, three concentration scaling exponents in the volume fraction Region I (0.36≤ϕ/ϕref≤0.46), Region II (0.59≤ϕ/ϕref≤2.89) and Region III (3.11≤ϕ/ϕref≤3.44) are identified. The image observation shows that the network formation of Phormidium suspensions occurs as the volume fraction is increased from Region I to Region II; the sol-gel transition takes place from Region II to Region III. In combination with analysis of other nanoscale suspensions and liquid crystalline polymer solutions reported in the literature, it is revealed that such a power law concentration scaling exponent depends on colloidal or molecular interactions mediated with solvent and is sensitive to the equilibrium phase behaviour of complex fluids. The TCS principle is an unambiguous tool to give a quantitative estimation.

Cross-Site Severity Assessment of COVID-19 From CT Images via Domain Adaptation.

Early and accurate severity assessment of Coronavirus disease 2019 (COVID-19) based on computed tomography (CT) images offers a great help to the estimation of intensive care unit event and the clinical decision of treatment planning. To augment the labeled data and improve the generalization ability of the classification model, it is necessary to aggregate data from multiple sites. This task faces several challenges including class imbalance between mild and severe infections, domain distribution discrepancy between sites, and presence of heterogeneous features. In this paper, we propose a novel domain adaptation (DA) method with two components to address these problems. The first component is a stochastic class-balanced boosting sampling strategy that overcomes the imbalanced learning problem and improves the classification performance on poorly-predicted classes. The second component is a representation learning that guarantees three properties: 1) domain-transferability by prototype triplet loss, 2) discriminant by conditional maximum mean discrepancy loss, and 3) completeness by multi-view reconstruction loss. Particularly, we propose a domain translator and align the heterogeneous data to the estimated class prototypes (i.e., class centers) in a hyper-sphere manifold. Experiments on cross-site severity assessment of COVID-19 from CT images show that the proposed method can effectively tackle the imbalanced learning problem and outperform recent DA approaches.