Shengbo Gao

Shengbo Gao, Ziji Zhang, Jiechao Ma et al.

Semi-supervised learning has become increasingly popular in medical image segmentation due to its ability to leverage large amounts of unlabeled data to extract additional information. However, most existing semi-supervised segmentation methods only focus on extracting information from unlabeled data, disregarding the potential of labeled data to further improve the performance of the model. In this paper, we propose a novel Correlation Aware Mutual Learning (CAML) framework that leverages labeled data to guide the extraction of information from unlabeled data. Our approach is based on a mutual learning strategy that incorporates two modules: the Cross-sample Mutual Attention Module (CMA) and the Omni-Correlation Consistency Module (OCC). The CMA module establishes dense cross-sample correlations among a group of samples, enabling the transfer of label prior knowledge to unlabeled data. The OCC module constructs omni-correlations between the unlabeled and labeled datasets and regularizes dual models by constraining the omni-correlation matrix of each sub-model to be consistent. Experiments on the Atrial Segmentation Challenge dataset demonstrate that our proposed approach outperforms state-of-the-art methods, highlighting the effectiveness of our framework in medical image segmentation tasks. The codes, pre-trained weights, and data are publicly available.

Xuanting Hao, Shengbo Gao, Lijie Sheng et al.

Consistency training has proven to be an advanced semi-supervised framework and achieved promising results in medical image segmentation tasks through enforcing an invariance of the predictions over different views of the inputs. However, with the iterative updating of model parameters, the models would tend to reach a coupled state and eventually lose the ability to exploit unlabeled data. To address the issue, we present a novel semi-supervised segmentation model based on parameter decoupling strategy to encourage consistent predictions from diverse views. Specifically, we first adopt a two-branch network to simultaneously produce predictions for each image. During the training process, we decouple the two prediction branch parameters by quadratic cosine distance to construct different views in latent space. Based on this, the feature extractor is constrained to encourage the consistency of probability maps generated by classifiers under diversified features. In the overall training process, the parameters of feature extractor and classifiers are updated alternately by consistency regularization operation and decoupling operation to gradually improve the generalization performance of the model. Our method has achieved a competitive result over the state-of-the-art semi-supervised methods on the Atrial Segmentation Challenge dataset, demonstrating the effectiveness of our framework. Code is available at https://github.com/BX0903/PDC.

Tianming Du, Peijie Yu, Sihan Shang et al.

The most plausible near-term role of medical LLMs is to assist rather than replace physicians, yet current evaluations often test isolated capabilities: clinical knowledge, EHR system interaction, or patient communication. Physician assistance instead requires coordinating these capabilities within the same interaction, where physicians issue underspecified requests, patients describe symptoms ambiguously, and EHR systems demand precise tool use. We introduce PhysAssistBench, a benchmark for interactive doctor-patient-EHR assistance. Built from real MIMIC-IV cases, PhysAssistBench uses a scalable pipeline to construct agentic patients: interactive, record-grounded agents that turn static EHR records into multi-turn clinical scenarios while preserving clinical factuality. PhysAssistBench provides a curated bilingual evaluation set of 1,296 manually reviewed and physician-validated turns. Experiments with leading LLMs show that current models remain unreliable in this setting, which exposes a key bottleneck for clinical LLMs: reliable assistance requires coordination across knowledge, communication, and systems, not isolated gains in any of them.