Deep learning for malignancy and tumor origin prediction using cytology or histopathology whole slide images - Takeaways - MDSpire

Deep learning for malignancy and tumor origin prediction using cytology or histopathology whole slide images

  • By

  • Ching-Wei Wang

  • Tzu-Chiao Chu

  • Tzu-Kang Wu

  • Yu-Pang Chung

  • Sin-Si Lin

  • Tai-Kuang Chao

  • January 24, 2026

  • 0 min

Share

  • 1

    MAMILE-UNI is a deep learning framework that detects malignancy in pleural and ascitic effusions from whole slide images.

  • 2

    The model achieved high AUROC, MeanSS, and accuracy in evaluating 1250 whole slide images for malignancy detection.

  • 3

    MAMILE-UNI also demonstrated high accuracy in identifying cancer origin from cytology smears and histopathological slide images.

  • 4

    The model's predictions were validated using Fisher’s exact test, showing statistical significance (p < 0.001).

  • 5

    The implementation of MAMILE-UNI utilized Python and PyTorch, with the code publicly available for academic use.

Original Source(s)

Deep learning for malignancy and tumor origin prediction using cytology or histopathology whole slide images

  • by Ching-Wei Wang, Tzu-Chiao Chu, Tzu-Kang Wu, Yu-Pang Chung, Sin-Si Lin, Tai-Kuang Chao

  • January 24, 2026

Related Content

This Week’s Biopharma News: KRAS Breakthrough and ADC Momentum

Revolution’s KRAS breakthrough, strong ADC signals from GSK, and shifting regulatory and dealmaking dynamics define this week in biopharma

  • April 21, 2026

  • 4 min
A literature review: progress in the study of plastic bronchitis

A Framework for Independent Scientific Exploration in Cancer Pathology Using AI