TY - JOUR

T1 - Transfer Learning Approaches for Brain Metastases Screenings

AU - Luu, Minh Sao Khue

AU - Tuchinov, Bair N.

AU - Suvorov, Victor

AU - Kenzhin, Roman M.

AU - Amelina, Evgeniya V.

AU - Letyagin, Andrey Yu

N1 - This work was supported by a grant for research centers, provided by the Analytical Center for the Government of the Russian Federation in accordance with the subsidy agreement and the agreement with the Novosibirsk State University dated 27 December 2023 No. 70-2023-001318.

PY - 2024/11

Y1 - 2024/11

N2 - Background: In this study, we examined the effectiveness of transfer learning in improving automatic segmentation of brain metastases on magnetic resonance imaging scans, with potential applications in preventive exams and remote diagnostics. Methods: We trained three deep learning models on a public dataset from the ASNR-MICCAI Brain Metastasis Challenge 2024, fine-tuned them on a small private dataset, and compared their performance to models trained from scratch. Results: Results showed that models using transfer learning performed better than scratch-trained models, though the improvement was not statistically substantial. The custom Tversky and Binary Cross-Entropy loss function helped manage class imbalance and reduce false negatives, limiting missed tumor regions. Medical experts noted that, while fine-tuned models worked well with larger, well-defined tumors, they struggled with tiny, scattered tumors in complex cases. Conclusions: This study highlights the potential of transfer learning and tailored loss functions in medical imaging, while also pointing out the models’ limitations in detecting very small tumors in challenging cases.

AB - Background: In this study, we examined the effectiveness of transfer learning in improving automatic segmentation of brain metastases on magnetic resonance imaging scans, with potential applications in preventive exams and remote diagnostics. Methods: We trained three deep learning models on a public dataset from the ASNR-MICCAI Brain Metastasis Challenge 2024, fine-tuned them on a small private dataset, and compared their performance to models trained from scratch. Results: Results showed that models using transfer learning performed better than scratch-trained models, though the improvement was not statistically substantial. The custom Tversky and Binary Cross-Entropy loss function helped manage class imbalance and reduce false negatives, limiting missed tumor regions. Medical experts noted that, while fine-tuned models worked well with larger, well-defined tumors, they struggled with tiny, scattered tumors in complex cases. Conclusions: This study highlights the potential of transfer learning and tailored loss functions in medical imaging, while also pointing out the models’ limitations in detecting very small tumors in challenging cases.

KW - brain metastases

KW - segmentation

KW - transfer learning

UR - https://www.scopus.com/record/display.uri?eid=2-s2.0-85210429535&origin=inward&txGid=d808a91b1e229b27de70b80f47812a97

UR - https://www.mendeley.com/catalogue/735be98b-684e-3a95-865c-aa21f969223d/

U2 - 10.3390/biomedicines12112561

DO - 10.3390/biomedicines12112561

M3 - Article

C2 - 39595126

VL - 12

JO - Biomedicines

JF - Biomedicines

SN - 2227-9059

IS - 11

M1 - 2561

ER -