logo
CanyamAI summaries for academic research
language
MobileMobile
  • Phone FeatureWeChat Mini Program
EnglishEnglish
noticeNotice
BlogBlog
delete
arrow
Return

MNiST: A deep learning framework for multi-scale spatial feature modeling and cellular landscape decoding in spatial

delete2025-08-06
delete0
PRE
AI
Z
Zhenghui Wang
R
Ruoyan Dai
K
Kaitai Han
M
Mengqiu Wang
L
Lixin Lei
张志威 cover
张志威 (Zhiwei Zhang)
Z
Zhenxing Li
L
Liu, XY
J
Jirui Zhang
H
Han Yan
郭
郭前进 (Qianjin Guo) *
DOI：10.1016/j.knosys.2025.114233delete
deleteOriginal
deleteOriginal request for help
deleteShare
deleteSave
Abstract

Abstract

En 中文
• Gmamba’s dual-branch encoder captures local details and long-range dependencies in spatial transcriptomics, enhancing structure modeling. • A Chebyshev-based multi-order frequency module captures multi-scale hierarchies and boundaries, enhancing spatial structure identification. • MNiST unifies spatial recognition and cell-type deconvolution via unsupervised embedding alignment, enhancing inference accuracy. • MNiST is validated on diverse human and model organism datasets, showing stable, broad applicability in spatial domain and cell-type tasks.
Keywords:
Gmamba
multi-order frequency module
spatial transcriptomics
cell-type deconvolution
unsupervised embedding alignment

Journal

K
Knowledge-Based Systems
IF:
7.6
Papers:
1.2W
Citations:
4.5W

Organization

No organization information available