Related Experiment Video
Updated: Jul 12, 2025
06:44
Modeling Brain Metastases Through Intracranial Injection and Magnetic Resonance Imaging
Published on: June 7, 2020
7.4K
Reducing false positives in deep learning-based brain metastasis detection by using both gradient-echo and spin-echo
Suyoung Yun1, Ji Eun Park2, NakYoung Kim3
1Department of Radiology, Busan Paik Hospital, Inje University College of Medicine, Busan, Republic of Korea.
European Radiology
|October 27, 2023
Summary
A new deep learning model using dual-enhanced MRI improves brain metastasis detection and reduces false positives compared to single-sequence models. This dual-enhanced deep learning approach achieves performance comparable to human experts.
Area of Science:
- Artificial Intelligence in Medical Imaging
- Deep Learning for Diagnostic Radiology
- Neuro-oncology Imaging Analysis
Background:
- Accurate detection of brain metastasis (BM) is crucial for treatment planning, particularly for stereotactic radiosurgery.
- Current deep learning (DL) models for BM detection often rely on single MRI sequences, potentially limiting their diagnostic performance.
- Integrating multiple contrast-enhanced MRI sequences may enhance the accuracy of DL-based BM detection.
Purpose of the Study:
- To develop and evaluate a novel deep learning (DL) model for brain metastasis (BM) detection that incorporates both gradient-echo (GRE) and turbo spin-echo (TSE) contrast-enhanced MRI (dual-enhanced DL).
- To compare the performance of the dual-enhanced DL model against a GRE-only DL model (GRE DL) and human readers in a clinical cohort.
- To assess the impact of dual-enhanced DL on detection sensitivity, positive predictive value (PPV), and overestimation of BM compared to existing methods.
Main Methods:
- A dual-enhanced DL model was trained on 200 patients with BM and validated on internal (62 patients) and external (48 patients) cohorts.
- The dual-enhanced DL model and a GRE DL model were evaluated for BM detection sensitivity and PPV.
- The relative differences (RDs) in BM counts between the DL models and two independent neuroradiologists were compared against a reference standard.
Main Results:
- Sensitivity was comparable between GRE DL (93%) and dual-enhanced DL (92%).
- The dual-enhanced DL model demonstrated a significantly higher positive predictive value (89%) compared to GRE DL (76%, p < .001).
- GRE DL significantly overestimated metastases (RD: 0.05), while dual-enhanced DL (RD: 0.00) showed no significant difference from neuroradiologists (RD: 0.00).
Conclusions:
- The dual-enhanced DL model, integrating both GRE and TSE contrast-enhanced MRI, significantly improves the positive predictive value for brain metastasis detection.
- This advanced DL approach effectively reduces the overestimation of metastases, achieving performance comparable to expert neuroradiologists.
- Dual-enhanced DL offers a promising tool to enhance the accuracy of BM detection, aiding in treatment guidance for stereotactic radiosurgery.
