Abstract
Microcalcifications are among the first signs of breast cancer and early detection can completely change the course of treatment and recovery of a patient. However, they are extremely small, often only a few pixels, and very hard to detect against dense breast tissue, so they are easily overlooked during routine checks, and such human mistakes can delay diagnosis and treatment. Automatically localizing them is very difficult, given the severe tumor to background imbalance. We propose a unified multi-stage framework for automated microcalcification analysis that combines semantic segmentation, patch level false-positive suppression, detection based false-negative recovery, and image-level BI-RADS assessment within a single workflow. We further investigate the impact of different segmentation architectures, domain-specific foundation model initialization, and detection strategies on the overall pipeline performance. Our results on the INbreast dataset show that the proposed multi-stage design reduces both false positives and missed lesions, substantially increasing the proportion of lesions the complete pipeline recovers, while pixel-level segmentation remains on par with strong domain-pretrained baselines.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/Deep_Brea3th_038.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=I3lqfh5Sup
BibTex
@InProceedings{SimAna_AMultiStage_MICCAISAT2026,
author = { Simion, Ana-Maria AND Florea, Adina Magda AND Mocanu, Irina},
title = { { A Multi-Stage Deep Learning Framework for the Automated Assessment of Mammographic Microcalcifications } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17256},
month = {pending},
page = {pending}
}