A multimodal ConvNeXt-Tiny deep learning model for simultaneous prediction of IDH mutation and Ki-67 expression in gliomas
Juan Du, Linsha Yang, Duo Zhang, Jinguo Wang, Shuo Wu, Defeng Liu, Huiling Yi, Xin Liang · 2026
Read A multimodal ConvNeXt-Tiny deep learning model for simultaneous prediction of IDH mutation and Ki-67 expression in gliomas by Juan Du, Linsha Yang, Duo Zhang, Jinguo Wang, Shuo Wu, Defeng Liu, Huiling Yi, Xin Liang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0351757.txt.
Description
Juan Du and seven colleagues teach a neural network to read brain tumour scans in this 2026 study. Two traits guide glioma treatment: whether the tumour carries an IDH mutation, and how fast its cells divide, marked by Ki-67. Both normally need tissue from surgery. The team built one model that predicts both at once. It feeds on four kinds of MRI, taken within a month before surgery at two hospitals. One hospital trained the model, the other tested it independently.
Hand-crafted image measurements and clinical details were fused with the network's features. Rival models served as comparisons. Accuracy was judged with standard curves, calibration checks, and decision analysis. Explanation tools highlighted which image regions drove each prediction. Non-invasive prediction matters because some patients cannot be biopsied. A validated scan-based estimate could shape treatment planning days earlier, while sparing patients an operation whose only purpose is information.
Catalog record
| CREATOR | Juan Du, Linsha Yang, Duo Zhang, Jinguo Wang, Shuo Wu, Defeng Liu, Huiling Yi, Xin Liang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 47365 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |