Anatomical recognition of dissection layers, nerves, vas deferens, and microvessels using artificial intelligence during transabdominal preperitoneal inguinal hernia repair
- Open Access
- 01.12.2025
- How-I-Do-It
Erschienen in:
Abstract
Introduction
During transabdominal preperitoneal inguinal hernia repair (TAPP), recognition of the loose connective tissue that can be dissected is essential for proper peritoneal dissection, including the myopectineal orifice (MPO) [1]. Furthermore, intraoperative nerve recognition has been shown to be effective in preventing chronic postoperative inguinal pain (CPIP) [1]. Recognizing the vas deferens is important for parietalization. Intraoperative bleeding during TAPP is often caused by damage to microvessels, which must be monitored during dissection.
Recently, artificial intelligence (AI) has been developed to improve the quality of medical care, including diagnostic imaging support, drug development, and genomic medicine [2‐4].
Anzeige
Recently, the use of AI in laparoscopic inguinal hernia surgery has been reported, and Takeuchi et al. reported on the use of AI in recognizing the MPO in TAPP [5]. Furthermore, Zygomalas et al. reported on the recognition of anatomical landmarks and surgical instruments during TAPP surgery using AI [6], and Ortenzi et al. reported on the recognition of surgical flow using AI in laparoscopic totally extraperitoneal inguinal hernia repair (TEP) [7].
Kumazu et al. examined a surgical support system that uses AI to recognize and display anatomical landmarks [8]. This surgical support system, named EUREKA, is currently in practical use mainly for educational and research purposes.
EUREKA is a deep learning model that automatically predicts anatomical structures. The developers of this technology used video data from robot-assisted gastrectomy to create a prototype AI model using annotation and deep learning. They then verified the accuracy of EUREKA in recognizing loose connective tissue, which was defined as the dissection layers. The mean Intersection over Union (IoU) score was 0.606, and the mean F1/Dice score was 0.549. This prototype model was refined, leading to the development of EUREKA, which is now currently primarily for education and research. Recognizing and displaying not only dissection layers but also nerves, ureters, the vas deferens, and microvessels is now possible.
In this study, we validated the accuracy of EUREKA in recognizing important anatomical structures during TAPP, such as dissection layers, nerves, vas deferens, and microvessels.
Anzeige
Materials and methods
Three thousand still images were created using TAPP, laparoscopic gastrectomy and laparoscopic colectomy videos of 150 cases performed by twelve qualified surgeons from the endoscopic surgical skill qualification system of the Japanese Society of Endoscopic Surgery (JSES), including the author [9]. These images were used for annotation and deep learning. EUREKA was used to analyze 10 TAPP surgical videos.
IoU and F1/Dice scores were calculated to quantitatively evaluate AI predictive images [10, 11]. These are the most commonly used performance measures in machine learning to evaluate sensitivity and similarity, respectively.
The formulas for calculating the F1/Dice and IoU scores are as follows (the higher, the better):
$$\frac{F1}{Dice}=\frac{TP}{TP+\frac{1}{2}(FP+FN)},\,\,\,\,\,IoU=\frac{TP}{TP+FP+FN}$$
TP: true positive, FP: false positive, FN: false negative.
We used EUREKA to recognize and display the dissection layers, nerves, vas deferens, and microvessels in several surgical videos of TAPP. We compared the original surgical videos with the visualized videos and examined whether visualization was performed properly using EUREKA.
How to connect and use EUREKA
This software can be used in a configuration with an installed workstation and a display monitor that displays the output image. The workstation can be used by connecting it to the video signal output interface of medical equipment, such as endoscope systems and surgical operation support systems, using a video signal cable to capture video signals.
Results
Quantitative evaluation of AI predictive images
The mean IoU and F1/Dice scores were 0.33 and 0.50 for connective tissue, 0.24 and 0.38 for nerves, 0.50 and 0.66 for the vas deferens, and 0.30 and 0.45 for microvessels, respectively.
Dissection layers
Figure 1 presents a comparison of the original and AI-visualized images of the dissection layers. The AI recognized and displayed the dissection layer at the level of fine connective tissue fibers. Loose connective tissue that emerges by applying appropriate tension to the peritoneum was considered the dissection layer (Fig. 1A and B). As shown in Fig. 1C and D, the loose connective tissue observed between the adipose tissue that adhered to the peritoneum and was difficult to detach and the adipose tissue that remained on the abdominal wall was also recognized and displayed as a dissection layer.
Anzeige
Fig. 1
EUREKA’s visualization of loose connective tissue during transabdominal preperitoneal patch plasty (TAPP). A Original image during lateral dissection of right inguinal hernia. B Recognition display of loose connective tissue by EUREKA for the image in A. C Original image of left inguinal hernia during lateral dissection. D Recognition display of loose connective tissue intervening between adipose tissue that adhered to the peritoneum and adipose tissue to be detached in the image in C
Nerves
Figure 2 presents a comparison of the original and AI-visualized image images of the nerves. AI accurately recognized and displayed the lateral femoral cutaneous nerve, which was visible between the fatty tissues of the abdominal wall (Fig. 2A and B). It also recognized and displayed the femoral branch of the genitofemoral nerve, which runs near the psoas major muscle (Fig. 2A and B). The AI recognized and displayed many nerve fibers (paravasal nerves) distributed around the vas deferens, which were clearly visible as nerve fibers running from the peritoneal lining to the vas deferens (Fig. 2C and D).
Fig. 2
Neural recognition displays by EUREKA. A Original image of the abdominal wall after peritoneal dissection. B Nerve recognition display by EUREKA for the image in A. Arrows indicate the femoral branch of the genitofemoral nerve, and arrowheads indicate the lateral femoral cutaneous nerves. C Original image of the vas deferens and testicular vessels after peritoneal dissection. D Recognition display of nerve plexus by EUREKA for the image in C
Vas deferens
Figure 3 presents a comparison of the original and AI-visualized images of the vas deferens. Only the vas deferens was clearly recognized and displayed, although loose connective tissue and nerves surrounding the vas deferens were also visible, (Fig. 3A and B).
Anzeige
Fig. 3
EUREKA’s visualization of the vas deferens and microvessels. A Original image of the vas deferens and surrounding tissue during peritoneal dissection. B Recognition display of the vas deferens by EUREKA for the image in A. C Original image of microvessels distributed in the abdominal wall during peritoneal dissection. D Recognition display of microvessels by EUREKA for the image in C
Microvessels
Figure 3 presents a comparison of the original and AI-visualized images of microvessels. The electronic supplementary material shows a video showing EUREKA’s recognition of the dissection layer, nerves, vas deferens, and microvessels. As shown in the electronic supplementary material, with respect to microvessels, the AI-visualized image was in a constant rhythm of intensity. The AI recognized and displayed microvessels distributed in the abdominal wall and peritoneum (Fig. 3C and D).
Discussion
Efforts to improve the quality of medical care through AI have been made in radiology imaging, endoscopy, and pathology; however, they are still in the development stage with respect to surgical support [2‐4]. Many attempts at intraoperative guidance using AI have been reported for cholecystectomy, colon resection, robot-assisted total prostatectomy, thyroidectomy, and sleeve gastrectomy [12‐18]. For surgery to proceed smoothly, the accurate recognition of intraoperative anatomical landmarks is necessary, which is sometimes difficult for trainees. AI was used to provide enhanced visualization of these anatomical landmarks so that all surgeons could visually share the information. The newly developed EUREKA, an AI-based surgical support system, has been commercialized for research and can be used for surgical education [19]. We used EUREKA for the first time to validate the accuracy of AI-analyzed visualizations of anatomical landmarks for TAPP.
Takeuchi et al. and Zygomalas et al. reported that anatomical landmarks can be recognized and displayed using a bounding box, with high F1/Dice and recall or sensitivity scores [5, 6]. In contrast, EUREKA differs from these previous reports in that it recognizes and displays anatomical landmarks. In other words, the ability to recognize and display landmarks is critical. Therefore, we believe that the tendency toward lower IoU and F1/Dice scores than those reported in previous studies is satisfactory.
Anzeige
Recognizing the dissection layer is the most important factor for facilitating peritoneal dissection. Compared with the original image, the AI could recognize and display the dissection layer at the fiber level of the loose connective tissue. It was also surprising that the AI could recognize and display loose connective tissue between the adipose tissue adhering to the peritoneum and that dissected from the abdominal wall. This is because recognizing loose connective tissue between adipose tissues is not easy, even for a skilled surgeon.
Trainees may be preoccupied with dissection maneuvers that they neglect to check nerve travel. Nerve cauterization or tacking may cause CPIP. Regarding the recognition and display of nerves by AI, nerves that are the main cause of CPIP, such as the lateral femoral cutaneous and genitofemoral nerves, were appropriately recognized and displayed. We believe that confirmation of these nerve runs using AI can prevent CPIP. The AI also recognized and displayed the nerve fibers of the paravasal nerve, which can be expected to prevent postoperative orchialgia, a related complication.
Recognition of the run of the vas deferens is essential for proper parietalization. In this study, EUREKA showed highly accurate AI visualization of the vas deferens, rather than including the tissue it. In some cases, determining the vas deferens is easy; however, in cases with excessive fatty tissue or thickened peritoneum, confirming the route of the vas deferens is sometimes difficult. In these difficult cases, confirming and learning to recognize the vas deferens using EUREKA may help prevent damage to the vas deferens.
Intermittent highlighting of microvessels using AI may help trainees become aware of the importance of preventing intraoperative bleeding. Most bleeding during laparoscopic inguinal hernia repair is due to microvascular injury. We believe that confirming and learning the route of microvessels around the dissected area using AI visualization can prevent hemorrhage.
In this comparative study, we believe that AI visualization of anatomical landmarks during peritoneal dissection in TAPP was highly accurate and provided a sufficiently reliable surgical support system.
EUREKA’s visualization of anatomical landmarks allows trainees and skilled surgeons to share their visual understanding of anatomical landmarks. Thus, they can share and learn to recognize anatomical structures, determine surgical procedures based on this recognition, and verify the accuracy of actual surgical procedures.
However, the use of EUREKA is still in its infancy and many challenges remain: how many complications can be reduced or how the learning curve can be shortened with EUREKA is still under study. Furthermore, whether it can be used in patients with adhesions or fibrosis, such as those with a history of laparotomy or radiotherapy, remains unclear. Regarding hernia surgery, indications for TEP and laparoscopic abdominal wall hernia surgery have not yet progressed. We hope that our results will be appreciated by hernia surgeons around the world and that more studies on EUREKA will be conducted.
Declarations
Conflict of interest
The authors report no conflicts of interest.
Ethical approval
The study was approved by the Institutional Ethical Board of Tsudanuma Central General Hospital.
Human and animal rights
All procedures performed in studies involving human participants were performed according to the ethical standards of the Institutional and/or National Research Committee and the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. This article does not contain any studies with animals performed by any of the authors.
Informed consent
Informed consent was obtained from all individual participants included in the study.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
