+ HEALTH Rationale and objectives To propose a new imaging biomarker, the pulmonary artery-to-vein volume difference (…± HEALTH Rationale and objectives The international application of artificial intelligence (AI) for lesion detection b…± HEALTH Rationale and objectives To address the challenge of preoperative prediction of synchronous liver metastasis…± HEALTH Objectives To conduct a nationwide survey among professionals working in oncology departments in China to inv…+ HEALTH The utilization of deep convolutional neural networks for the purpose of diagnosing diseases in the skin area…± HEALTH BACKGROUND . Gadolinium-based contrast agents remain essential for MRI but carry risks. Deep learning (DL) me…+ SCIENCE Gas Chromatography is a versatile separation technique widely used in analytical chemistry for constituent de…+ HEALTH Background Intracardiac echocardiography (ICE) facilitates left atrial (LA) reconstruction during atrial fibr…
TruaceTracing the truth around AIThursday, August 27, 2026
Subscribe
Health·The Trace·Dual reading·Published 2026-08-27

AI lesion detection on digital breast tomosynthesis across Western and Eastern databases

Source article: International Application of Artificial Intelligence for Lesion Detection on Digital Breast Tomosynthesis: Comparing Western and Eastern Databases

Abstract: Rationale and objectives The international application of artificial intelligence (AI) for lesion detection based on digital breast tomosynthesis (DBT) is limited due to disease variations among populations. We hypothesized that lesion detection models trained on either the Western or Eastern DBT dataset would exhibit reduced performance on another dataset. We proposed transfer learning to enhance lesion detection across DBT databases. Materials and methods The Western database (94 patients) was obtained from th…

TRV-2026-0908Peer-reviewedPermanent record — cite & verify
Trace impact reading

Contested: both sides are scored from claims and sources, not community votes.

P 71The P score combines the specificity and measured human impact of the grounded problem claim with the strength of this Trace’s cited sources.G 71The G score combines the specificity and measured human impact of the grounded gain claim with the strength of this Trace’s cited sources.
International Application of Artificial Intelligence for Lesion Detection on Digital Breast Tomosynthesis: Comparing Western and Eastern Databases

Full-field digital (FFD)-galactography and Digital breast tomosynthesis (DBT)-galactography by Marco Moschetta, Vincenzo De Ruvo, Angelica Drago, Nicoletta Troiano, Simona Paolicelli, Giuseppe Rubini, Amato Antonio Stabile Ianora & Michele Telegrafo. CC BY-SA 4.0 · https://creativecommons.org/licenses/by-sa/4.0

The quick read

A study tested YOLO-based AI models for breast lesion detection on digital breast tomosynthesis using a 94-patient Western database from the Cancer Imaging Archive and a 157-patient Eastern database from a single medical center, with lesions grouped into six types.

The work matters because population differences in lesion types reduced cross-dataset performance, raising questions about international deployment of DBT AI, while transfer learning showed promise for adaptation; uncertainty remains about performance beyond these two limited datasets and six lesion categories.

Main points
  • Study compared Western database of 94 patients from Cancer Imaging Archive and Eastern database of 157 patients from one anonymous medical center.
  • All breast lesions were stratified into six types and evaluated with YOLO lesion detection models.
  • Performance was measured using Intersection over Union (IoU), sensitivity, and precision to assess generality across databases.
Gain

Transfer learning enhanced AI lesion detection performance when models were adapted from one regional DBT database to another.

Problem

AI lesion detection models trained on Western or Eastern DBT data showed reduced performance when applied to the other population due to differences in lesion types.

The rundown

Researchers trained YOLO lesion detection models on either the Western or Eastern DBT dataset, with and without transfer learning to the other dataset, stratifying lesions into six types.

Evaluation used IoU, sensitivity, and precision, finding significant differences in lesion types between databases that impacted model generality, while transfer learning improved performance in the target dataset.

What this doesn’t fix

Findings are based on small samples of 94 and 157 patients with Eastern data from a single anonymous center, limiting broad international generalizability.

Sources

Reader signal

How should this claim be treated?

The debate