PCA-YOLO: A Small Liver Tumor Detection Model with Patch-Contrastive Attention

Xueyang Li, Han Xiao, Zongpeng Weng, Xinrong Hu, Danny Chen, Yiyu Shi

Keywords: Liver tumor detection, Attention mechanism, CT scan dataset, YOLO model.

Abstract:

Liver tumors, as one of the most common malignant tumor types, represent a significant clinical challenge, with the detection of small tumors being particularly problematic.
Despite the rapid advances in deep learning (DL) offering significant support in reducing the workload of radiologists, current detection models still struggle with the detection of small tumors. This is particularly troubling as these are the cases where even experienced radiologists are more prone to errors, underscoring the critical need for improved accuracy of detection methods in this area. Addressing this critical gap, this article introduces patch-contrastive attention YOLO (PCA-YOLO), an innovative adaptation of the YOLO framework, incorporating a patch-based attention module to specifically target the detection of small liver tumors. Furthermore, we collected a specialized CT dataset focusing exclusively on small liver tumors, complemented with meticulously annotated bounding boxes, to facilitate this study. Our experimental findings demonstrate that our approach achieves a leading mean Average Precision (mAP) score of 77.2\% at a 50\% Intersection Over Union (IoU) threshold, surpassing all current leading detection methods tested against our specialized dataset.

Primary Subject Area: Detection and Diagnosis

Secondary Subject Area: Application: Radiology

Paper Type: Methodological Development

Registration Requirement: Yes

Visa & Travel: Yes

Midl Latex Submission Checklist: Ensure no LaTeX errors during compilation., Created a single midl25_NNN.zip file with midl25_NNN.tex, midl25_NNN.bib, all necessary figures and files., Includes \documentclass{midl}, \jmlryear{2025}, \jmlrworkshop, \jmlrvolume, \editors, and correct \bibliography command., Did not override options of the hyperref package, Did not use the times package., All authors and co-authors are correctly listed with proper spelling and avoid Unicode characters., Author and institution details are de-anonymized where needed. All author names, affiliations, and paper title are correctly spelled and capitalized in the biography section., References must use the .bib file. Did not override the bibliographystyle defined in midl.cls. Did not use \begin{thebibliography} directly to insert references., Tables and figures do not overflow margins; avoid using \scalebox; used \resizebox when needed., Included all necessary figures and removed *unused* files in the zip archive., Removed special formatting, visual annotations, and highlights used during rebuttal., All special characters in the paper and .bib file use LaTeX commands (e.g., \'e for é)., Appendices and supplementary material are included in the same PDF after references., Main paper does not exceed 9 pages; acknowledgements, references, and appendix start on page 10 or later.

Submission Number: 6