دورية أكاديمية
أعرض في EDS

Bone Metastases Lesion Segmentation on Breast Cancer Bone Scan Images with Negative Sample Training.

التفاصيل البيبلوغرافية
العنوان: Bone Metastases Lesion Segmentation on Breast Cancer Bone Scan Images with Negative Sample Training.
المؤلفون: Chen YY; Department of Biomedical Imaging and Radiological Science, China Medical University, Taichung 404, Taiwan., Yu PN; Department of Biomedical Imaging and Radiological Science, China Medical University, Taichung 404, Taiwan., Lai YC; Department of Nuclear Medicine, Feng Yuan Hospital, Ministry of Health and Welfare, Taichung 420, Taiwan., Hsieh TC; Department of Biomedical Imaging and Radiological Science, China Medical University, Taichung 404, Taiwan.; Department of Nuclear Medicine and PET Center, China Medical University Hospital, Taichung 404, Taiwan., Cheng DC; Department of Biomedical Imaging and Radiological Science, China Medical University, Taichung 404, Taiwan.
المصدر: Diagnostics (Basel, Switzerland) [Diagnostics (Basel)] 2023 Sep 25; Vol. 13 (19). Date of Electronic Publication: 2023 Sep 25.
نوع المنشور: Journal Article
اللغة: English
بيانات الدورية: Publisher: MDPI AG Country of Publication: Switzerland NLM ID: 101658402 Publication Model: Electronic Cited Medium: Print ISSN: 2075-4418 (Print) Linking ISSN: 20754418 NLM ISO Abbreviation: Diagnostics (Basel) Subsets: PubMed not MEDLINE
أسماء مطبوعة: Original Publication: Basel, Switzerland : MDPI AG, [2011]-
مستخلص: The use of deep learning methods for the automatic detection and quantification of bone metastases in bone scan images holds significant clinical value. A fast and accurate automated system for segmenting bone metastatic lesions can assist clinical physicians in diagnosis. In this study, a small internal dataset comprising 100 breast cancer patients (90 cases of bone metastasis and 10 cases of non-metastasis) and 100 prostate cancer patients (50 cases of bone metastasis and 50 cases of non-metastasis) was used for model training. Initially, all image labels were binary. We used the Otsu thresholding method or negative mining to generate a non-metastasis mask, thereby transforming the image labels into three classes. We adopted the Double U-Net as the baseline model and made modifications to its output activation function. We changed the activation function to SoftMax to accommodate multi-class segmentation. Several methods were used to enhance model performance, including background pre-processing to remove background information, adding negative samples to improve model precision, and using transfer learning to leverage shared features between two datasets, which enhances the model's performance. The performance was investigated via 10-fold cross-validation and computed on a pixel-level scale. The best model we achieved had a precision of 69.96%, a sensitivity of 63.55%, and an F1-score of 66.60%. Compared to the baseline model, this represents an 8.40% improvement in precision, a 0.56% improvement in sensitivity, and a 4.33% improvement in the F1-score. The developed system has the potential to provide pre-diagnostic reports for physicians in final decisions and the calculation of the bone scan index (BSI) with the combination with bone skeleton segmentation.
References: Semin Nucl Med. 2009 Nov;39(6):357-68. (PMID: 19801216)
Diagnostics (Basel). 2023 Jul 06;13(13):. (PMID: 37443695)
Diagnostics (Basel). 2023 Apr 02;13(7):. (PMID: 37046538)
World J Radiol. 2015 Aug 28;7(8):202-11. (PMID: 26339464)
Nucl Med Rev Cent East Eur. 2013;16(2):66-9. (PMID: 24068635)
Nucl Med Commun. 2006 May;27(5):417-23. (PMID: 16609352)
Eur Urol. 2012 Jul;62(1):78-84. (PMID: 22306323)
J Nucl Med. 2006 Feb;47(2):287-97. (PMID: 16455635)
Ann Nucl Med. 2019 Nov;33(11):866-874. (PMID: 31493203)
Int J Comput Assist Radiol Surg. 2020 Mar;15(3):389-400. (PMID: 31836956)
Diagnostics (Basel). 2021 Mar 15;11(3):. (PMID: 33803921)
J Clin Oncol. 2004 Jul 15;22(14):2942-53. (PMID: 15254062)
Phys Med. 2021 Apr;84:141-148. (PMID: 33894584)
Lancet Oncol. 2009 Jun;10(6):606-14. (PMID: 19482249)
Med Image Anal. 2020 Oct;65:101784. (PMID: 32763793)
Clin Cancer Res. 1998 Jul;4(7):1765-72. (PMID: 9676853)
J Nucl Med. 1997 Sep;38(9):1401-6. (PMID: 9293797)
Cancers (Basel). 2023 Jul 13;15(14):. (PMID: 37509267)
PLoS One. 2020 Dec 3;15(12):e0243253. (PMID: 33270746)
J Cancer Res Ther. 2016 Apr-Jun;12(2):787-92. (PMID: 27461652)
Sci Rep. 2021 Feb 19;11(1):4223. (PMID: 33608560)
Phys Med Biol. 2022 Nov 09;67(22):. (PMID: 36137545)
Br J Cancer. 1987 Jan;55(1):61-6. (PMID: 3814476)
J Imaging. 2023 Aug 16;9(8):. (PMID: 37623695)
Phys Med Biol. 2022 Jan 17;67(1):. (PMID: 34933282)
J Nucl Med. 2008 Dec;49(12):1958-65. (PMID: 18997038)
معلومات مُعتمدة: MOST 111-2314-B-039-040 National Science and Technology Council
فهرسة مساهمة: Keywords: Double U-Net; bone metastasis segmentation; deep learning; negative mining; pre-train; transfer learning
تواريخ الأحداث: Date Created: 20231014 Latest Revision: 20231031
رمز التحديث: 20231215
مُعرف محوري في PubMed: PMC10572884
DOI: 10.3390/diagnostics13193042
PMID: 37835785
قاعدة البيانات: MEDLINE
الوصف
تدمد:2075-4418
DOI:10.3390/diagnostics13193042