Enhancing Anterior Quadratus Lumborum Block Accuracy with Artificial Intelligence: A Segmentation Approach Evaluated by Dice Score Metrics

Enhancing Anterior Quadratus Lumborum Block Accuracy with Artificial Intelligence: A Segmentation Approach Evaluated by Dice Score Metrics | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Enhancing Anterior Quadratus Lumborum Block Accuracy with Artificial Intelligence: A Segmentation Approach Evaluated by Dice Score Metrics View ORCID Profile Daniel Bidstrup , View ORCID Profile Anuj Pareek , View ORCID Profile Jens Børglum doi: https://doi.org/10.1101/2025.10.29.25339036 Daniel Bidstrup † Department of Anesthesiology, Zealand University Hospital , Roskilde, Denmark MD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Daniel Bidstrup For correspondence: daniel.bidstrup{at}gmail.com Anuj Pareek ‡ Department of Radiology, Copenhagen University Hospital , Copenhagen, Denmark MD, PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Anuj Pareek Jens Børglum † Department of Anesthesiology, Zealand University Hospital , Roskilde, Denmark MD, PhD Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jens Børglum Abstract Full Text Info/History Metrics Data/Code Preview PDF ABSTRACT Anterior quadratus lumborum (QL) block is a regional anesthesia technique shown to provide both somatic and visceral pain relief by targeting lower thoracic nerves and the thoracic sympathetic trunk. Despite its clinical benefits, success depends on accurate sonoanatomic identification, which can be challenging due to individual anatomical variations. In this study, we developed an artificial intelligence (AI) model to automatically segment key sonoanatomic landmarks for the anterior QL block. A total of 82 ultrasound videos from 42 healthy volunteers yielded 460 labeled images capturing the vertebral body (L3/L4), posterior renal fascia, transverse abdominal muscle, quadratus lumborum muscle, psoas major muscle, and the injection point. We trained a 2D U-Net–based model (nnU-Net) with five-fold cross-validation. Training data was split into an 80% training set (368 images) and 20% validation set (92 images). The performance of the AI model was tested on images obtained from 20 patients receiving the anterior QL block as a part of standard treatment. The model achieved a moderate-high Dice score of 0.62 across six classes, with especially high segmentation performance for vertebral bodies (Dice 0.90) and the psoas major muscle (Dice 0.85). Low-moderate performance was observed for the posterior renal fascia (Dice 0.35) and the injection point (Dice 0.38), likely reflecting their subtle sonographic appearance. In conclusion, this is the first AI model that can delineate the sonoanatomy of the anterior QL block region. Our findings underscore the potential of AI to improve the precision and consistency of ultrasound-guided anterior QL blocks. What is already known on this topic The anterior quadratus lumborum (QL) block effectively reduces postoperative opioid consumption and pain but relies on precise sonoanatomic identification. The efficacy of the block has been concluded in a recent systematic review with meta-analyses. What this study adds It presents the first AI model to segment key landmarks for the anterior QL block in a clinical setting. How this study might affect research, practice or policy AI-based segmentation may improve consistency, reduce operator dependence, and enhance beneficial patient outcomes. INTRODUCTION The ultrasound-guided quadratus lumborum (QL) block is a regional anesthesia technique that has recently gained increasing popularity for enhancing a beneficial postoperative pain management. There are different approaches to perform a QL block, including the anterior, lateral, and posterior approach. Each block technique has their unique application guidelines and possible benefits. Introduced in 2013, the ultrasound-guided anterior QL block was shown in a cadaveric study to enable injectate to spread from its lumbar site of injection cranially towards the diaphragmatic openings and further into the thoracic paravertebral space, and thereby reaching the thoracic sympathetic trunk and the lower thoracic nerves. 1 The anterior QL block also simultaneously anesthetizes the subcostal, ilioinguinal, and iliohypogastric nerves, suggesting its ability to provide both somatic and visceral pain relief. 1 Recent studies show that the anterior QL block reduces opioid usage post cesarean section, laparoscopic nephrectomy, and percutaneous nephrolithotomy. 2 – 4 While the anterior QL block has gained increasing popularity as an effective technique for postoperative pain management, its success largely depends on accurate identification of the sonoanatomy of the block. The block technique can be challenging for the ultrasound operator, and anatomical differences can lead to suboptimal analgesia and potential complications due to failed placement of blocks. However, when performed correctly the efficacy of the block has been confirmed in a recent systematic review with meta-analyses. 5 During the last decade, there has been growing interest in the use of Artificial Intelligence (AI) to improve the accuracy and consistency of regional anesthesia techniques. Several recent studies have explored the use of AI in ultrasound-guided regional anesthesia. 6 For example, Mwikirize et al. (2018) developed an AI algorithm to assist with needle detection during 2D ultrasound. 7 Similarly, Huang et al. (2019) successfully trained an AI algorithm for the segmentation of the sonoanatomy in the femoral nerve region. 8 In our current study, we aimed to develop an AI model for recognizing the sonoanatomy of the anterior QL block. To achieve this, we gathered a dataset of ultrasound images of the anterior QL block region from a cohort of healthy volunteers (42 medical students). Data was collected on a total anonymous basis. The images were annotated by a 5th year resident in anesthesiology (DB) to provide ground truth labels for the AI model. We then used this dataset to train a convolutional neural network 9 (CNN) to recognize predefined elements of the sonoanatomy associated with the anterior QL block. The achieved neural network was then evaluated on a test set (20 patients) using the Dice score. 10 – 12 Our results demonstrate that our AI model can accurately identify prespecified elements of the sonoanatomy in the anterior QL block. We believe that our approach has the potential to improve the accuracy and consistency of anterior QL block procedures and ultimately improve patient outcomes. METHODS Training data In our study, we utilized video recordings of simulated ultrasound-guided anterior QL block procedures at vertebral level L3-L4. The recordings were obtained from 42 healthy medical students who volunteered while on clinical training at our department (see Table 1 ). Of these, 40 volunteers were scanned bilaterally and 2 were scanned unilaterally, resulting in a total of 82 unique recordings. Each ultrasound recording had a 30-second duration, which was sufficient to capture dynamic anatomical changes across respiratory cycles. These dynamic changes guide the ultrasound operator for better anatomical recognition in daily clinical practice. Important anatomical structures like the para- and perinephric fat, as well as the posterior renal fascia, change position and appearance during respiratory cycles 2 . This approach allowed us to collect a broader range of sonoanatomical variation and thereby helped us obtain more useful training data from each volunteer. View this table: View inline View popup Download powerpoint Table 1: Descriptive Characteristics of the Healthy Volunteers Constituting the Training Data. Anthropometric data presented as median (interquartile range). Training Data Selection Each 30-second ultrasound video consisted of 360 image frames, and 5-15 representative image frames were carefully selected from each video, yielding a total dataset of 460 images with a resolution of 784 × 580 pixels. The key factors for choosing the images were: (1) Clear visibility of essential sonoanatomic landmarks, and (2) diversity in the appearance of anatomical structures to represent variations encountered in clinical practice. This method ensured a broad representation of the anterior QL block sonoanatomy. Annotation and Labeling A resident in anesthesiology (DB) annotated all images, providing detailed ground truth labels for 6 different classes. These labels included the following sonoanatomic classes: 1) the vertebral body of L3 or L4 and their respective transverse process, 2) posterior renal fascia, 3) transverse abdominal muscle, 4) quadratus lumborum muscle, 5) psoas major muscle, and 6) the location of the injection point for the anterior QL blockade in the fascial interspace between the QL and psoas major muscle posterior to the transversalis fascia. This step was crucial for the subsequent training of our AI model. The annotation and labelling were done using the Labelbox online tool. 13 AI Model Development We setup our AI model as the nnU-Net 14 self-configuring method for semantic segmentation of biomedical images in PyTorch 15 . For images the nnU-Net pipeline makes use of the 2D U-net architecture 9 which has consistently demonstrated robust performance on segmentation tasks. The nnU-Net pipeline automatically normalizes the input images and performs data augmentation using rotations, scaling, Gaussian noise etc. The nnU-Net was trained on an Nvidia L4 24Gb GPU using 5-fold cross-validation meaning the data was split into an 80% training set (368 images) and 20% validation set (92 images) for each fold. Each fold was trained for 1000 epochs of 250 minibatches with a minibatch size of 6. The model was optimized using stochastic gradient descent with Nesterov momentum and a decaying learning rate. The loss function combined the sum of Cross-Entropy loss and Dice loss to balance pixel-wise classification with overlap-based segmentation performance. Model performance was evaluated using the average Dice score across all six label classes, and following 5-fold cross-validation, the final model was selected as the best-performing ensemble from the trained folds. Performance Evaluation The effectiveness of our AI model was quantitatively assessed using the Dice score metric 10 . The Dice score, also known as the Dice Similarity Coefficient, is a statistical tool used to gauge the similarity between two samples 11 , 12 . In our study, it was used to evaluate the agreement between the AI-predicted segmentations and the manually annotated sonoanatomy in the ultrasound images. We categorized Dice similarity coefficients into five bins; low [0.00, 0.20), low-moderate [0.20, 0.40), moderate [0.40, 0.60), moderate-high [0.60, 0.80), and high [0.80, 1.00] consistent with established practice in biomedical image segmentation. 16 Test Data & Evaluation To evaluate the clinical applicability of our trained CNN model, we utilized a test set consisting of images from 20 patients who received an anterior QL block as part of their standard perioperative care. Of these patients, 18 received the block post-elective cesarean section, while two received it pre-surgery for laparoscopic nephrectomy (see Table 2 ). For each patient, a single ultrasound image capturing optimal sonoanatomy was selected. View this table: View inline View popup Download powerpoint Table 2: Descriptive Characteristics of the Patients Constituting the Test Data. Anthropometric data presented as median (interquartile range). All 20 test images were labeled and annotated by DB and subsequently reviewed by JB (anesthesia consultant) to provide ground truth labels. The final trained nnU-Net model ensemble was used to predict segmentation labels for each of the 20 images. The model performance on these unseen images was then quantitatively assessed for each sonoanatomic class using the Dice score by comparing the predicted labels with our ground truth labels. This evaluation allowed us to determine the accuracy of our model in a clinical setting, thereby validating its potential utility in assisting with the placement of anterior QL blocks. Ethical Approval and Data Protection The study was reviewed by the Regional Committee on Health Research Ethics for Region Zealand (EMN-2025-03485) who concluded that ethical approval was not required. All participating patients provided written informed consent prior to inclusion. Data were processed in accordance with the EU General Data Protection Regulation (GDPR) and Danish Data Protection legislation. RESULTS We achieved an average Dice score of 0.62 across all anatomical classes in the test set (see Table 3 ). This value reflects the overall accuracy of the model in identifying and segmenting the relevant sonoanatomical landmarks for the anterior QL block. Dice scores for the individual label classes are presented in Table 3 . View this table: View inline View popup Download powerpoint Table 3: Performance Evaluation of Our AI Model on Test Data Using Dice Score. Figure 1 illustrates our AI model’s segmentation performance in five selected patients by comparing ground truth labels with model predicted labels. In Patient 1, the predicted labels closely matched the reference annotations across all label classes. A comparable level of performance was observed in 11 out of the 20 patients (data not shown). Patient 2 exemplifies cases in which the model showed reduced consistency in identifying the posterior renal fascia and the transverse abdominal muscle. Patient 3 shows that the anterior quadratus QL block injection point was generally detected, although the predicted region was smaller than the reference annotation. Patients 4 and 5 demonstrate, that despite achieving a Dice score of 0.69 for the quadratus lumborum muscle, the predicted label is spatially displaced in some images. Patient 5 also represents an instance of out-of-distribution anatomy not present in our training data; this patient had renal cancer, which altered the local anatomy and was associated with reduced segmentation accuracy. Download figure Open in new tab Figure 1: Comparison of Ground Truth and Predicted Labels for AI Model Performance on Five Selected Patients. Patient 1 shows near-perfect agreement. Patient 2 highlights challenges in identifying the posterior renal fascia and transverse abdominal muscle. Patient 3 demonstrates underestimation of the anterior QL blockade injection point. Patients 4 and 5 show occasional misplacement of the quadratus lumborum muscle label despite a Dice score of 0.69. Patient 5 further illustrates poor performance on out-of-distribution sonoanatomy due to renal cancer-induced anatomical distortion. DISCUSSION To the best of our knowledge this is the first AI model to segment the sonoanatomy of the anterior QL block. Despite using only, a very limited size of labeled training data, we were able to achieve low-moderate to high segmentation performance. For the anterior QL blockade injection point, the Dice score was 0.38, reflecting a low-moderate accuracy in identifying the precise location. This may be due to the subtle anatomic nature of the injection point. In contrast, the model performed very well in recognizing the vertebral body of L3 and L4, achieving a high Dice score of 0.90. This shows that the vertebrae, being more distinct and well-defined on ultrasound, were consistently and accurately segmented by our model. The quadratus lumborum muscle was identified with moderate-high success, achieving a Dice score of 0.69, which indicates that our model generally recognized the muscle but faced difficulties with some anatomical variations. The segmentation of the psoas major muscle showed high accuracy, with a Dice score of 0.85, suggesting that this structure was consistently identified by the model. For the transverse abdominal muscle, our model achieved a moderate Dice score of 0.51. The lower score may be due to the variable visualization of this muscle, particularly in patients with higher body mass index (BMI), where subcutaneous fat tissue deteriorates the image quality on ultrasound. The volunteers in our training dataset had a median BMI of 23.7 (see Table 1 ), whereas the patients in the test dataset had a median BMI of 32.5 (see Table 2 ). This difference can largely be attributed to the high proportion of pregnant women in the test cohort. However, the elevated BMI in the test data may have challenged our model’s ability to delineate the transverse abdominal muscle. The posterior renal fascia was the most challenging structure for our model to segment, with a low-moderate Dice score of 0.35. This may be due to its less distinct appearance and the variable quality of the ultrasound images in capturing this structure, particularly in patients with increased adiposity. Overall, our AI model demonstrated moderate-high performance in segmenting sonoanatomical landmarks. Our results exemplify the feasibility of using AI to assist physicians and ultrasound operators in delineating sonoanatomy for accurate ultrasound guided interventions including the anterior QL block. The efficacy of the anterior QL block as a technique for postoperative pain management has been well established, but its effectiveness is largely dependent on accurate identification of the sonoanatomy associated with the block. 2 , 5 The achieved accuracy by our AI model is indicative of its potential as a reliable tool in assisting medical professionals in identifying the sonoanatomy for anterior QL block procedures, thereby enhancing the precision of this regional anesthesia technique. In the context of the anterior QL block, our study indicates that AI can play a crucial role in recognizing the sonoanatomy, thereby enhancing the success of the block. One of the advantages of our AI model is that it has the potential to provide real-time feedback to the ultrasound operator during the anterior QL block procedure. This can help to reduce the time required to perform the block and potentially reduce the risk of complications. Our results suggest that the use of AI in the anterior QL block procedures can improve patient outcomes by reducing the risk of suboptimal analgesia and complications. Moreover, the successful development of our AI model demonstrates the potential of machine learning in revolutionizing ultrasound-guided regional anesthesia. By enhancing the visibility and understanding of sonoanatomy, these tools can potentially alleviate the learning curve associated with acquiring ultrasound skills. Applying AI as an educational tool has been presented in the anesthesia literature. 17 The integration of AI in regional anesthesia techniques, such as the anterior QL block, represents an area of cutting-edge research. While our current study suggests that AI can significantly enhance the recognition of sonoanatomy the full scope of these technologies in clinical practice is still unfolding. Limitations The success of any AI algorithm greatly relies on the quality and diversity of the data on which it is trained. In our study, the ultrasound images were obtained from a limited cohort of medical students which are considered as healthy, young volunteers. This might not accurately represent the broad demographic and anatomical diversity found in real-world patients. Factors such as age, body habitus, and underlying medical conditions can significantly alter the sonoanatomy associated with the anterior QL block; i.e. potentially leading to discrepancies in the algorithm’s performance across different patient populations. Future research should aim to incorporate a more diverse range of ultrasound images to improve the generalizability of our AI model. Our study did not compare the performance of our AI model to that of experienced human operators. While our algorithm performed well on our dataset, it is highly possible that experienced operators may still be able to identify the sonoanatomy of the anterior QL block more accurately. Another limitation relates to the method used to validate our AI model’s performance. The ground truth labels for the training dataset were provided by a single experienced 5th year resident in anesthesiology. While this is a common practice, it relies on the assumption that the annotator’s assessment is entirely accurate, which may not always be the case. An assessment error or bias from the annotator can propagate through the AI training and affect its performance. Ideally, future studies should incorporate multiple expert labels to cross-verify the annotations and reduce the risk of bias or error. Lastly, the transparency or ‘explainability’ of AI models can often be a challenge, where it can be difficult to interpret exactly why the model has made a particular decision. The clinical application of our AI model also needs to be approached with caution. While the AI system was successful in identifying sonoanatomy on ultrasound images, it’s important to remember that the decision to perform the block, and the actual execution of it, remains a highly skilled task that requires a thorough understanding of anatomy, patient history, and the potential complications. Future Perspectives Despite these potential limitations and the need for further testing, our study underscores the potential of AI in improving regional anesthesia techniques. By applying AI to ultrasound-guided anterior QL block, we hope to make the procedure more precise, efficient, and safer, contributing to better postoperative pain management and improving patient outcomes. To further validate the effectiveness of our AI model, it would be beneficial to conduct studies in real-time clinical settings. These studies could investigate whether our AI model can enhance the accuracy and efficiency of the anterior QL block procedures when used in tandem with an anesthesiologist, including how it impacts patient outcomes. CONCLUSION In conclusion, we created the first AI model that can delineate the sonoanatomy of the QL block region at the paraspinal level L3-L4. Our model demonstrated high performance in segmenting clear anatomical landmarks but struggled with less defined or more variable structures. While there’s still much to learn about the applications of AI in anesthesia and pain management, the results of our study provide a glimpse into a future where machine learning tools can assist anesthesiologists in carrying out complex procedures like the anterior QL block with improved accuracy and consistency. Our research forms the foundation for further exploration into this field, and we anticipate further advancements that will build on our work. Data Availability All data produced in the present study are available from the corresponding author upon reasonable request and in compliance with Danish data protection regulations. ACKNOWLEDGMENTS We acknowledge the use of AI technology in the preparation of this manuscript. Specifically, ChatGPT version 5 (OpenAI) was used to assist in refining the language, improving clarity, and proofreading the text. ChatGPT was not involved in data analysis, interpretation, or decision-making related to the study findings. All final edits and content were reviewed and approved by the authors. Footnotes Conflicts of Interest: The authors declare no conflicts of interest. Funding: The authors have no sources of funding to declare for this manuscript. Prior presentation: No prior presentation. REFERENCES 1. ↵ Dam M , Moriggl B , Hansen CK , Hoermann R , Bendtsen TF , Børglum J . The pathway of injectate spread with the transmuscular quadratus lumborum block: A cadaver study . Anesth Analg . 2017 ; 125 ( 1 ): 303 – 312 . OpenUrl PubMed 2. ↵ Dam M , Hansen C , Poulsen TDi , et al. Transmuscular quadratus lumborum block reduces opioid consumption and prolongs time to first opioid demand after laparoscopic nephrectomy . Reg Anesth Pain Med . 2021 ; 46 ( 1 ): 18 – 24 . OpenUrl Abstract / FREE Full Text 3. Dam M , Hansen CK , Poulsen TD , et al. Transmuscular quadratus lumborum block for percutaneous nephrolithotomy reduces opioid consumption and speeds ambulation and discharge from hospital: a single centre randomised controlled trial . Br J Anaesth . 2019 ; 123 ( 2 ): e350 – e358 . OpenUrl PubMed 4. ↵ Hansen CK , Dam M , Steingrimsdottir GE , et al. Ultrasound-guided transmuscular quadratus lumborum block for elective cesarean section significantly reduces postoperative opioid consumption and prolongs time to first opioid request: A double-blind randomized trial . Reg Anesth Pain Med . 2019 ; 44 ( 9 ): 896 – 900 . OpenUrl Abstract / FREE Full Text 5. ↵ Tanggaard K , Gronlund C , Nielsen M V. , et al. Anterior quadratus lumborum blocks for postoperative pain treatment following intra-abdominal surgery: A systematic review with meta-analyses and trial sequential analyses . Acta Anaesthesiol Scand. John Wiley and Sons Inc . 2025 ; 69 ( 1 ). 6. ↵ Viderman D , Dossov M , Seitenov S , Lee MH . Artificial intelligence in ultrasound-guided regional anesthesia: A scoping review . Front Med (Lausanne) . 2022 ; 9 : 994805 . 7. ↵ Mwikirize C , Nosher JL , Hacihaliloglu I . Convolution neural networks for real-time needle detection and localization in 2D ultrasound . Int J Comput Assist Radiol Surg . 2018 ; 13 ( 5 ): 647 – 657 . OpenUrl CrossRef PubMed 8. ↵ Huang C , Zhou Y , Tan W , et al. Applying deep learning in recognizing the femoral nerve block region on ultrasound images . Ann Transl Med . 2019 ; 7 ( 18 ): 453 – 453 . OpenUrl PubMed 9. ↵ Ronneberger O , Fischer P , Brox T . U-Net: Convolutional Networks for Biomedical Image Segmentation . Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference . 2015 : 234 - 241 . 10. ↵ Zou KH , Warfield SK , Bharatha A , et al. Statistical Validation of Image Segmentation Quality Based on a Spatial Overlap Index . Acad Radiol . 2004 ; 11 ( 2 ): 178 – 189 . OpenUrl CrossRef PubMed Web of Science 11. ↵ Dice LR . Measures of the Amount of Ecologic Association Between Species . Ecology . 1945 ; 26 ( 3 ): 297 – 302 . OpenUrl CrossRef Web of Science 12. ↵ Sørensen T . A method of establishing groups of equal amplitude in plant sociology based on similarity of species and its application to analyses of the vegetation on Danish commons . Kongelige Danske Videnskabernes Selskab - Biologiske Skrifter . 1948 ; 5 ( 4 ): 1 – 34 . OpenUrl 13. ↵ Labelbox , "Labelbox," Online , 2025 . [Online]. Available: https://labelbox.com . 14. ↵ Isensee F , Jaeger PF , Kohl SAA , Petersen J , Maier-Hein KH . nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation . Nature Methods . 2020 ; 18 ( 2 ): 203 – 211 . OpenUrl PubMed 15. ↵ Paszke A , Gross S , Massa F , et al. PyTorch: An Imperative Style , High-Performance Deep Learning Library. Neural Information Processing Systems . 2019 : 8024 – 8035 . 16. ↵ Wilson SM , Bautista A , Yen M , Lauderdale S , Eriksson DK . Validity and reliability of four language mapping paradigms . Neuroimage Clin . 2017 ; 16 : 399 – 408 . OpenUrl PubMed 17. ↵ Jacobs E , Wainman B , Bowness J . Applying artificial intelligence to the use of ultrasound as an educational tool: a focus on ultrasound guided regional anesthesia . Anat Sci Educ . 2024 ; 17 ( 5 ): 919 – 925 OpenUrl PubMed View the discussion thread. Back to top Previous Next Posted November 02, 2025. Download PDF Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Enhancing Anterior Quadratus Lumborum Block Accuracy with Artificial Intelligence: A Segmentation Approach Evaluated by Dice Score Metrics Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Enhancing Anterior Quadratus Lumborum Block Accuracy with Artificial Intelligence: A Segmentation Approach Evaluated by Dice Score Metrics Daniel Bidstrup , Anuj Pareek , Jens Børglum medRxiv 2025.10.29.25339036; doi: https://doi.org/10.1101/2025.10.29.25339036 Share This Article: Copy Citation Tools Enhancing Anterior Quadratus Lumborum Block Accuracy with Artificial Intelligence: A Segmentation Approach Evaluated by Dice Score Metrics Daniel Bidstrup , Anuj Pareek , Jens Børglum medRxiv 2025.10.29.25339036; doi: https://doi.org/10.1101/2025.10.29.25339036 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Anesthesia Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (300) Cardiovascular Medicine (4440) Dentistry and Oral Medicine (444) Dermatology (383) Emergency Medicine (608) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1510) Epidemiology (15229) Forensic Medicine (30) Gastroenterology (1126) Genetic and Genomic Medicine (6605) Geriatric Medicine (668) Health Economics (998) Health Informatics (4541) Health Policy (1369) Health Systems and Quality Improvement (1613) Hematology (543) HIV/AIDS (1265) Infectious Diseases (except HIV/AIDS) (15923) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (147) Nephrology (668) Neurology (6604) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1145) Occupational and Environmental Health (957) Oncology (3334) Ophthalmology (974) Orthopedics (369) Otolaryngology (420) Pain Medicine (436) Palliative Medicine (130) Pathology (663) Pediatrics (1693) Pharmacology and Therapeutics (692) Primary Care Research (711) Psychiatry and Clinical Psychology (5448) Public and Global Health (9235) Radiology and Imaging (2199) Rehabilitation Medicine and Physical Therapy (1370) Respiratory Medicine (1196) Rheumatology (594) Sexual and Reproductive Health (712) Sports Medicine (530) Surgery (712) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a017690abbadb3d6',t:'MTc3OTc0MjIwNA=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();