What question did this study set out to answer?

To determine whether deep learning applied to ex vivo fluorescence confocal microscopy (FCM) images enables automated, accurate, and spatially interpretable tumor identification in pediatric tissue specimens.

August 14, 2026European Journal of Pediatric Surgery

Interpretable Deep Learning Applied to Fluorescence Confocal Microscopy for Intraoperative Tissue Assessment in Pediatric Surgical Oncology

View Full Paper

Ask AI

Bookmark

Authors

DFDonatella Di FabrizioOspedali Riuniti di Ancona ASAgnese SbrolliniElectrophysiology EBEdoardo BindiMarche Polytechnic University

Discussion

Member takes

Implication

Diagnostic study demonstrates accurate automated tumor detection using deep learning with confocal microscopy in pediatric oncology specimens, highlighting potential for intraoperative guidance.

Key Points

To determine whether deep learning applied to ex vivo fluorescence confocal microscopy (FCM) images enables automated, accurate, and spatially interpretable tumor identification in pediatric tissue specimens.
Prospectively acquired FCM mosaics from 42 pediatric patients (141 surgical and biopsy specimens) were retrospectively collected, expert-annotated as malignant, benign, or healthy, and decomposed into 243,659 tiles (256 × 256 pixels at 0.5 µm/pixel).
Trained and evaluated a convolutional neural network on stratified datasets and generated tile-level activation heatmaps to interpret cellular features driving model predictions.
In an independent testing set of 84,751 tiles, the model identified malignant tissue with 91.35% accuracy, 95.88% sensitivity, 86.28% specificity, and an F1 score of 92.13%.
The system classified healthy tissue with 94.41% accuracy (95.58% specificity) and benign tissue with 90.14% accuracy (91.78% specificity).
Tile-level activation heatmaps consistently localized cellular-level architectural distortions in malignant tissue.

Cite This Study

Fabrizio et al. (2026) studied this question.

synapsesocial.com/papers/6a7ec707b70b84ec8b9132aa https://doi.org/10.1055/a-2936-0049

View Full Paper

Ask AI

Bookmark

Implication

Diagnostic study demonstrates accurate automated tumor detection using deep learning with confocal microscopy in pediatric oncology specimens, highlighting potential for intraoperative guidance.

Interpretable Deep Learning Applied to Fluorescence Confocal Microscopy for Intraoperative Tissue Assessment in Pediatric Surgical Oncology

Discussion

Member takes

Implication

Key Points

Cite This Study

Interpretable Deep Learning Applied to Fluorescence Confocal Microscopy for Intraoperative Tissue Assessment in Pediatric Surgical Oncology

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider