Abstract
Cervical cancer is the fourth most common malignancy among women worldwide, yet Africa bears incidence and mortality seven to eight times those of high-income regions. Its staging is inherently multi-organ under FIGO, yet automated reporting remains unexplored in low-resource settings. To address this gap, we introduce the first African benchmark for cervical cancer CT diagnosis and report generation. To support this, we curated 100 patients and 902 DICOM series with paired reports from a tertiary Zambian cancer hospital, spanning abdominopelvic and thoracic acquisitions. Under a unified protocol, we benchmarked four 3D CT foundation models with patient-level bootstrap intervals. Experiments show Pillar-0 attains the strongest performance, achieving an average AUC of 71.4% across 10 organs, yet no model meaningfully outperforms a trivial always-positive baseline on the cervical mass itself, revealing a domain gap. By releasing this benchmark publicly, we aim to catalyze research into accessible cervical cancer imaging. Code is available at: https://github.com/xmed-lab/TriALS-Report.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/AFRICAI_042.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=fpyn7ytdQj
BibTex
@InProceedings{MukKan_Benchmarking_MICCAISAT2026,
author = { Mukuka, Kangwa E. AND Lambart, Lena AND Mwape, Festus AND Phiri, Lighton AND Spampinato, Concetto AND Lekadir, Karim AND Elbatel, Marawan},
title = { { Benchmarking Foundation Models for Cervical Cancer CT Reporting in Zambia } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17264},
month = {pending},
page = {pending}
}