Domain-adaptive semi-supervised learning for efficient rare pathological lesion detection with minimal annotation

Artificial intelligence for rare pathological lesion detection faces dual challenges: expert annotation scarcity and domain shifts across institutions. Using multi-institutional kidney biopsies from 22 hospitals with 3 scanner types (NDPI, VSI, SVS), we demonstrate that model performance decreases dramatically across domains, with up to 70.3% reduction in detection precision for rare lesions such as crescents and segmental sclerosis (comprising only 2-3% of annotations). We present an approach integrating semi-supervised learning with residual CycleGAN-based domain adaptation, reducing mean Fréchet inception distance between institutions from 55.9 to 20.2 while preserving diagnostic morphology. We identified context-dependent optimal strategies: semi-supervised learning with 50% confidence threshold excelled in same-hospital scenarios (15.2-17.7% improvement for rare lesions), while our combined GAN-Semi-Supervised approach demonstrated superior performance in cross-scanner scenarios between NDPI and VSI formats (up to 63.4% improvement for crescents). This methodology enables robust performance across diverse healthcare settings with minimal expert annotation.

Authors

Institutions

Publication Details

Journal
npj Digital Medicine
Published
2025-11-23
DOI
https://doi.org/10.1038/s41746-025-02160-6
Citations
2
Primary Topic
AI in cancer detection
Type
article
Field-Weighted Citation Impact
4.96

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Domain-adaptive semi-supervised learning for efficient rare pathological lesion detection with minimal annotation

Yasuo Kusunoki, Akira Suzuki, Takao Masaki, Takahiro Kuragano et al.
2 citations
npj Digital Medicine
AI in cancer detection
4.96
article

Domain-adaptive semi-supervised learning for efficient rare pathological lesion detection with minimal annotation

Yasuo Kusunoki, Akira Suzuki, Takao Masaki, Takahiro Kuragano, Shoichi Maruyama, Imari Mimura, Kazunori Inoue, Hiroyuki Abé, Isao Matsui, Tetsuo Ushiku, Kengo Furuichi, Eiichiro Kanda, Yoshitaka Isaka, Takaaki Kosugi, Kazuhiko Tsuruya, Yoshiyasu Ueda, Yoshimasa Kawazoe, Eiichiro Uchino, Hirohiko Niioka, Masaki Ohya, Naohiko Fujii, Kei Fukami, Hajime Nagasu, Hiroki Okushima, Osamu Yamaguchi, Naotake Tsuboi, Masaomi Nangaku, Tamaki Harada, Masanobu Takeji, Ryohei Kaseda, Atsuhiro Imai, Yuka Kurokawa, Yuichiro Yano, Yasuyuki Nagasawa, Sawako Kato, Yukio Ando, Toshiaki Nakano, Keiji Fujimoto, Motoko Yanagita, Toshiyuki Imasawa, Akira Shimizu, Masahiro Koizumi, Hitoshi Yokoyama, Ryohei Yamamoto, Katsuyuki Nagatoya, Shin‐ichi Araki, Tadashi Sofue, Noriyuki Tamai, Tatsuo Tsukamoto, Hirotaka Komaba, Masatoshi Abe, Ken‐ichi Miyoshi, Ayumi Matsumoto, Yasuhiko Ito, Hiroshi Kinashi, Yuta Matsukuma, Kazumasa Oka, Toshiyuki Komiya, Masaaki Izumi, Suguru Yamamoto, Tetsuo Minamino
article en
2 citations

Abstract

Artificial intelligence for rare pathological lesion detection faces dual challenges: expert annotation scarcity and domain shifts across institutions. Using multi-institutional kidney biopsies from 22 hospitals with 3 scanner types (NDPI, VSI, SVS), we demonstrate that model performance decreases dramatically across domains, with up to 70.3% reduction in detection precision for rare lesions such as crescents and segmental sclerosis (comprising only 2-3% of annotations). We present an approach integrating semi-supervised learning with residual CycleGAN-based domain adaptation, reducing mean Fréchet inception distance between institutions from 55.9 to 20.2 while preserving diagnostic morphology. We identified context-dependent optimal strategies: semi-supervised learning with 50% confidence threshold excelled in same-hospital scenarios (15.2-17.7% improvement for rare lesions), while our combined GAN-Semi-Supervised approach demonstrated superior performance in cross-scanner scenarios between NDPI and VSI formats (up to 63.4% improvement for crescents). This methodology enables robust performance across diverse healthcare settings with minimal expert annotation.

npj Digital MedicineVol. 8(1)
Kurume University (JP), Osaka Gakuin University (JP), Tokai University (JP), Kyushu University (JP), Hyogo Medical University (JP), Fujita Health University (JP), Aichi Medical University (JP), Kanazawa Medical University (JP), Kawasaki Medical School (JP), Wakayama Medical University (JP), Kagawa University (JP), Kyoto University (JP), Osaka Prefectural Medical Center (JP), Tokyo Health Care University (JP), Juntendo University (JP), Toyonaka Municipal Hospital (JP), Kyushu University Hospital (JP), Osaka City General Hospital (JP), Higashiosaka College & Higashiosaka Junior College (JP), The Japanese Society of Gastroenterological Surgery (JP), Osaka Rosai Hospital (JP), Nara City Hospital (JP), Kansai Rosai Hospital (JP), Nara Medical University Hospital (JP), Higashi Osaka City General Hospital (JP), Osaka Minami Medical Center (JP), Osaka Hospital (JP), Hiroshima University Hospital (JP), Kanazawa Medical University Hospital (JP), Otsu Red Cross Hospital (JP), Chiba Hospital (JP), Japanese Society of Medical Oncology (JP), Hyogo Prefectural Nishinomiya Hospital (JP), Kitano Hospital (JP), Ehime University (JP), Nagoya University (JP), Nara Medical University (JP), Niigata University (JP), The University of Tokyo (JP), Nippon Medical School (JP), Hyogo University (JP), Kindai University (JP), The University of Osaka (JP)
Japan Society for the Promotion of Science
Openalex Percentile: Top 4%
AI in cancer detection
4.96
DOIOpen Access PDFOpenAlex
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.