Intelligent Lesion Selection: A Novel Method for Longitudinal Assessment of Breast Cancer Lung Metastases

Melika Qahqaie, Veronika A Zimmer, Eduardo Castaneda, Katariina Peltonen, Joonas Laaksolilja, Juho Lähteenmaa, Tobias Heimann, Andreas Maier, Dominik Neumann

Keywords: Longitudinal disease assessment, Lung metastasis in breast cancer, Computed Tomography (CT), Deep learning, Lesion detection and selection

TL;DR: A novel pipeline with temporal lesion pair classification improves tumor size progression assessment by 42% and treatment response classification accuracy by 22% in CT-based metastatic lung lesion monitoring.

Abstract: Breast cancer, the second most common cancer globally, often metastasizes to the lungs, requiring frequent computed tomography (CT) scans to monitor disease progression. Manual analysis by radiologists is time-consuming and prone to variability, underscoring the need for automated systems to enhance accuracy and efficiency. The goal of such systems is to optimize processes like RECIST score calculation for tumor response assessment. This study presents a pipeline for the automated temporal analysis of breast cancer lung metastases. Existing lung nodule detection and segmentation models were adapted for detecting and segmenting breast cancer metastases. Registration-based lesion tracking was incorporated, and a novel Temporal Lesion Pair Classifier was developed to identify significant lesions and estimate tumor load evolution by summing their diameters, following an adaptation of the RECIST guidelines. Evaluated on a unique dataset of breast cancer patients, each with multiple annotated CT scans at different disease stages, the proposed pipeline demonstrated a 42\% reduction in median tumor size progression discrepancy for consecutive study pairs and improved tumor response classification accuracy by 22\% at the patient level.

Primary Subject Area: Detection and Diagnosis

Secondary Subject Area: Application: Radiology

Paper Type: Both

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: 193