Keloids are chronic fibrotic inflammatory lesions located in the reticular dermis that are characterized by abnormal fibroblast proliferation and excessive
Transjugular intrahepatic portosystemic shunt (TIPS) is an effective treatment for portal hypertension, but post-TIPS hepatic encephalopathy (HE) remains a major complication. Conventional clinical scores use limited clinical and biochemical variables and cannot fully characterize preoperative imaging phenotypes. Although contrast-enhanced abdominal computed tomography(CT) contains morphological, textural, and spatial cues, available datasets are small, CT sequences are redundant, and local details must be modeled together with broader context. We therefore propose Medical Global Zoom Mamba (MGZ Mamba). A medical visual invariance constraint stratifies CT slices by feature density using the nonzero-pixel ratio and gray-level entropy, allowing multiple training instances to be constructed from each patient sequence. A double-nested parallel aggregate attention module combines multi-receptive-field convolution with global channel recalibration, while a feature rearrangement 2D scan module introduces dynamic windows and spatial-neighborhood rearrangement alongside the original two-dimensional selective scan branch. In patient-level five-fold cross-validation on the Post-TIPS Hepatic Encephalopathy Dataset of Shanxi Medical University (HE-SMU), MGZ Mamba achieved 81.65% accuracy, 82.76% recall, 80.53% F1-score, and an ROC-AUC of 0.829. Relative to the strongest baseline for each metric, accuracy, recall, and F1-score increased by 0.41, 1.07, and 0.56 percentage points, respectively, and ROC-AUC increased by 0.007. These findings support MGZ Mamba as an imaging-based approach for preoperative risk stratification and may provide complementary evidence for clinical assessment and treatment planning.
Lung imaging enables direct visualization of lesions and is vital for pneumonia diagnosis. However, current deep learning models show unsatisfactory classification accuracy and generalization for multi-class lung disease recognition, limiting their clinical decision-support value. To tackle this issue, we present LPC-Transformer, a framework optimized via improved training strategies with Swin Transformer as the backbone; performance gains are realized without deepening the network. We propose an LMF hybrid loss integrating Focal Loss and LDAMLoss to mitigate class imbalance. Furthermore, the SCION optimizer is adopted to stabilize training, and PolyReLU is used in MLP modules to model complex nonlinear features. Evaluated on CL-COVIDset containing 11 classes of lung CT images, our method is compared with mainstream benchmark networks. Experiments reveal that LPC-Transformer surpasses the baseline SwinT model substantially, with accuracy, recall, precision and F1-score improved by 4.2%, 4.4%, 2.0% and 3.8%, respectively. These results validate the effectiveness of the proposed training strategy optimization.