We describe the DS@GT submissions to the ImageCLEFmedical Caption 2026 challenge, which continues a long-running benchmark on the ROCOv2 dataset with two tracks: Concept Detection (Task 1), assigning UMLS Concept Unique Identifiers (CUIs) to radiology images, and Caption Prediction (Task 2), generating natural-language captions. For Task 1, our primary submission was a three-way late-fusion ensemble of ConvNeXt-V2, BiomedCLIP ViT-B/16, and DenseNet-169 with a regularized''Honest Threshold Tuning''procedure designed to avoid validation overfitting on rare concepts; this submission ranked first on the official submission with a primary $F_1$ of $0.5790$ and a secondary $F_1$ of $0.9657$. In parallel, we submitted a training-free KNN retrieval pipeline over frozen BiomedCLIP embeddings, which reached a primary $F_1$ of $0.5780$ and a secondary $F_1$ of $0.9599$-essentially matching the fine-tuned ensemble on the primary track at a fraction of the cost. For Task 2, our submissions included a fine-tuned Gemma-3 27B model (overall $0.3571$, ranking third in the official submission), a fully fine-tuned BLIP pipeline with custom Vizwins merging ($0.3564$), and a zero-shot MedGemma-4B run with a PubMed-style prompt ($0.3186$), spanning a wide range of model scales and training costs. Code: https://github.com/dsgt-arc/imageclef-caption-2026.
PURPOSE
Hepatocellular carcinoma (HCC) is a leading cause of cancer-related death worldwide, with current therapies often limited by significant drug resistance. Owing to the Warburg effect, targeting cancer-specific metabolic vulnerabilities is a promising therapeutic strategy. This study aims to investigate the role of RNF114 in HCC progression and its regulatory mechanism, as well as its clinical translational potential as a therapeutic target.
METHODS
We evaluated the clinical significance of RNF114 using tissue microarrays and database analysis. RNF114 function in promoting HCC progression by regulating glucose uptake was investigated using knockdown experiments in cell lines and subcutaneous xenograft models. Furthermore, a therapeutic xenograft model was employed to assess the potential of RNF114 knockdown in overcoming Sorafenib resistance.
RESULTS
RNF114 was highly expressed in HCC and correlated with poor prognosis. Knockdown of RNF114 significantly suppressed HCC cell proliferation, migration, invasion, and glycolysis. Co-immunoprecipitation identified PACSIN3 as a key substrate of RNF114. RNF114 interacted with the SH3 domain of PACSIN3, promoting its ubiquitination and proteasomal degradation. Subcellular fractionation revealed that the F-BAR domain of PACSIN3 facilitated GLUT1 vesicular trafficking. Consequently, RNF114 impaired this process, leading to increased plasma membrane retention of GLUT1 and enhanced glycolytic flux. Consistently, in both HCC cells and subcutaneous xenograft models, RNF114 knockdown sensitized tumors to Sorafenib treatment.
CONCLUSIONS
Collectively, our findings reveal that the RNF114-PACSIN3-GLUT1 axis regulates glucose uptake and metabolic reprogramming in HCC, thereby promoting tumor progression and contributing to therapy resistance. Targeting this signaling axis provides a novel insight into metabolic therapy for HCC.
Yue Song, Yi-lu Lu, Qixiang Liu et al.· Cellular Oncology· 0 citations
