基于深度学习的前路坐骨神经超声图像分割

Deep learning-based segmentation of anterior approach sciatic nerve ultrasound images

摘要

目的 构建一套超声影像数据集，建立基于深度学习的超声图像识别系统，用以探索前入路区域坐骨神经阻滞区域的识别方法。方法 通过超声采集前路坐骨神经图像，借助ITK-SNAP软件手动标记，建立数据集。采用PyTorch深度学习框架进行训练数据及感兴趣区域分割输出，以平均交并比、平均骰子相似系数作为评价指标评估模型的性能。结果 以获得的3 000张标记超声图像作为数据集，其中，训练集1 800张，验证集600张，测试集600张。测试集坐骨神经的平均交并比为0.675，平均骰子相似系数为0.775，模型平均 F1分数为0.718，中位数为0.720。经5折交叉验证，坐骨神经的平均交并比中位数为0.805。结论 基于深度学习模型，在自动识别前入路区域坐骨神经解剖结构时获得了良好效果，可实现局麻药在神经旁间隙的安全、精准注射，具有良好的临床应用价值。

Abstract

Objective To construct a set of ultrasound image datasets and a deep learning-based ultrasound image recognition system in order to explore the recognition method for the sciatic nerve block area in the anterior approach region. Methods Ultrasound images of their anterior approach sciatic nerve were collected and manually labeled using ITK-SNAP software, and a dataset was established. The PyTorch deep learning framework was used for training data processing and segmentation output of regions of interest. The model performance was evaluated using the mean intersection over union (mIoU) and mean dice similarity coefficient (mDice) as evaluation metrics. Results A total of 3 000 labeled ultrasound images were used as the dataset, including 1 800 images for the training set, 600 for the validation set, and 600 for the test set. For the sciatic nerve in the test set, the mIoU was 0.675, the mDice coefficient was 0.775, the model achieved a mean F1-score of 0.718 and a median of 0.720. A 5-fold cross-validation determined the median of mIoU for the sciatic nerve to be 0.805. Conclusion Based on the deep learning model, favorable results were achieved in the automatic identification of the anatomical structure of the sciatic nerve in the anterior approach region. This method enables a safe and precise injection of local anesthetics into the paraneural space, presenting promising clinical application value.

关键词

深度学习 / 前路坐骨神经 / 超声 / 图像识别 / 三元注意力网络 / 自动分割 / 神经阻滞

Key words

deep learning / anterior sciatic nerve / ultrasound / image recognition / triplet attention net / automatic segmentation / nerve block anesthesia

参考文献

[1] SIERŻANTOWICZ R, LEWKO J, BITIUCKA D, et al. Evaluation of pain management after surgery: an observational study[J]. Medicina, 2020, 56(2): 65.

[2] FALLON F, MOORTHY A, SKERRITT C, et al. Latest advances in regional anaesthesia[J]. Medicina, 2024, 60(5): 735.

[3] DESAI N, KIRKHAM K R, ALBRECHT E. Local anaesthetic adjuncts for peripheral regional anaesthesia: a narrative review[J]. Anaesthesia, 2021, 76: 100-109.

[4] ORI A, JINDAL A, NWAZOTA N, et al. Peripheral nerve injections[J]. Physical medicine and rehabilitation clinics of north america, 2022, 33(2): 489-517.

[5] FERNÁNDEZ MARTIN M T, ALVAREZ LOPEZ S, ALDECOA ALVAREZ—SANTULLANO C. Role of adjuvants in regional anesthesia: A systematic review[J]. Revista española de anestesiología y reanimación, 2023, 70(2): 97-107.

[6] NOBLE J A, NAVAB N, BECHER H. Ultrasonic image analysis and image—guided interventions[J]. Interface focus, 2011, 1(4): 673-685.

[7] MCLEOD G, REINA M A. Nerve block, nerve damage, and fluid injection pressure: overturning the myth[J]. British journal of anaesthesia, 2024, 132(5): 1022-1026.

[8] BAIS K, GUIRGUIS F, GUIRGUIS M. Nerve injury following regional nerve block: a literature review of its etiologies, risk factors, and prevention[J]. Current pain and headache reports, 2024, 28(9): 863-868.

[9] CATALANI B, JONES J J. Peripheral nerve block complications in children[J]. Orthopedic clinics of north america, 2022, 53(2): 179-186.

[10] SHAMS D, SACHSE K, STATZER N, et al. Regional anesthesia complications and contraindications[J]. Anesthesiology clinics, 2024, 42(2): 329-344.

[11] VAGHARI D, MOHANKUMAR G, TAN K, et al. AI—guided patient stratification improves outcomes and efficiency in the amaranth alzheimer’s disease clinical trial[J]. Nature communications, 2025, 16: 6244.

[12] LI T, BO W, HU C, et al. Applications of deep learning in fundus images: a review[J]. Medical image analysis, 2021, 69: 101971.

[13] VAN DER VELDEN B H M, KUIJF H J, GILHUIJS K G A, et al. Explainable artificial intelligence in deep learning—based medical image analysis[J]. Medical image analysis, 2022, 79: 102470.

[14] CHANDRASEKARAN J, GOPAL D, SURESHKUMAR L V, et al. Integrated ai and machine learning pipeline identifies novel kinase inhibitors for targeted cancer therapy[J]. Molecular diversity, 2025, 29(4): 3425-3448.

[15] HUANG C, ZHOU Y, TAN W, et al. Applying deep learning in recognizing the femoral nerve block region on ultrasound images[J]. Annals of translational medicine, 2019, 7(18): 453.

[16] HUANG A, JIANG L, ZHANG J, et al. Attention—VGG16—UNet: a novel deep learning approach for automatic segmentation of the median nerve in ultrasound images[J]. Quantitative imaging in medicine and surgery, 2022, 12(6): 3138-3150.

[17] COSMO M , CHIARA FIORENTINO M , VILLANI F P , et al. Learning—based median nerve segmentation from ultrasound images for carpal tunnel syndrome evaluation[C]// IEEE engineering in medicine and biology society. Piscataway: Institute of Electrical and Electronics Engineers (IEEE), 2021: 3025-3028.

[18] DING Y, YANG Q, WANG Y, et al. MallesNet: a multi—object assistance based network for brachial plexus segmentation in ultrasound images[J]. Medical image analysis, 2022, 80: 102511.

[19] KULL C, MARTIN R, ROSSEL J B, et al. Femoral vs sciatic nerve block to provide analgesia after medial open wedge high tibial osteotomy in the setting of multimodal analgesia: a randomized, controlled, single—blinded trial[J]. Journal of clinical anesthesia, 2024, 93: 111355.

[20] ZHAO Y, ZHENG S, CAI N, et al. Utility of artificial intelligence for real—time anatomical landmark identification in ultrasound—guided thoracic paravertebral block[J]. Journal of digital imaging, 2023, 36(5): 2051-2059.

[21] SMISTAD E, JOHANSEN K F, IVERSEN D H, et al. Highlighting nerves and blood vessels for ultrasound—guided axillary nerve block procedures using neural networks[J]. Journal of medical imaging, 2018, 5(4): 044004.

[22] ALKHATIB M, HAFIANE A, VIEYRES P, et al. Deep visual nerve tracking in ultrasound images[J]. Computerized medical imaging and graphics, 2019, 76: 101639.

[23] BOWNESS J, VARSOU O, TURBITT L, et al. Identifying anatomical structures on ultrasound: assistive artificial intelligence in ultrasound guided regional anesthesia[J]. Clinical anatomy, 2021, 34(5): 802-809.

[24] GUNGOR I, GUNAYDIN B, OKTAR S O, et al. A real—time anatomy identification via tool based on artificial intelligence for ultrasound—guided peripheral nerve block procedures: an accuracy study[J]. Journal of anesthesia, 2021, 35(4): 591-594.

基金资助

国家自然科学基金资助项目(82560177)

甘肃省自然科学基金资助项目(22JR5RA954)

兰州市科技计划资助项目(2024-3-37)

甘肃海智计划资助项目(KPZX-010533)