Ujjainkar Priyanka Anup, Raut Shital A
Computer Science and Engineering, Visvesvaraya National Institute of Technology, Nagpur, Maharashtra, 440010, India.
Surg Endosc. 2025 Jun 30. doi: 10.1007/s00464-025-11932-w.
Accurate tracking and enumeration of surgical instruments are critical for patient safety and operational efficiency in laparoscopic procedures. Advanced tracking systems enhance object detection by maintaining instrument identity despite rapid movements, overlaps, or occlusions, ensuring real-time precision and preventing misplacement.
This research aims to improve real-time detection and tracking of surgical instruments in minimally invasive surgery (MIS) by integrating a sophisticated deep learning (DL) framework. Utilizing cutting-edge computer vision (CV) techniques, the proposed approach enhances surgical workflows, minimizes errors, and improves patient safety with increased precision and efficiency.
The framework incorporates YOLOv9n, the latest iteration released in 2024, along with advanced tracking algorithms such as ByteTrack and BoT-SORT to enable real-time detection, tracking, and enumeration of surgical instruments. The YOLOv9n model underwent evaluation against YOLOv8n, YOLOv5n, YOLOv11n, and Faster R-CNN to ensure an optimal balance of speed and accuracy. The study employed the m2cai16-tool-locations detection dataset, utilizing advanced preprocessing and data augmentation techniques to enhance model performance and mitigate dataset imbalances.
Comparative assessments demonstrated that YOLOv9n, when combined with ByteTrack and BoT-SORT, achieved outstanding performance with a Mean Average Precision (mAP50) of 98.4% and an inference speed of 0.3 milliseconds. This system ensures robust tracking even under rapid movements and partial occlusions, significantly improving precision and operational efficiency in MIS.
This automated deep learning solution reduces the cognitive burden on surgical teams, enhances patient safety, and optimizes procedural efficiency. Its potential integration with IoT, mobile applications, and intelligent operating rooms underscores its transformative role in modern surgery, setting new standards for AI-driven laparoscopic procedures and surgical innovation.
在腹腔镜手术中,准确跟踪和清点手术器械对于患者安全和手术效率至关重要。先进的跟踪系统通过在器械快速移动、重叠或遮挡的情况下保持器械标识,增强目标检测能力,确保实时精度并防止误放。
本研究旨在通过集成先进的深度学习(DL)框架,改进微创手术(MIS)中手术器械的实时检测和跟踪。利用前沿的计算机视觉(CV)技术,所提出的方法优化手术流程,减少误差,并通过提高精度和效率来提升患者安全。
该框架整合了2024年发布的最新版本YOLOv9n,以及诸如ByteTrack和BoT-SORT等先进的跟踪算法,以实现手术器械的实时检测、跟踪和清点。对YOLOv9n模型与YOLOv8n、YOLOv5n、YOLOv11n和Faster R-CNN进行了评估,以确保速度和准确性的最佳平衡。该研究采用了m2cai16-tool-locations检测数据集,利用先进的预处理和数据增强技术来提升模型性能并缓解数据集不平衡问题。
对比评估表明,YOLOv9n与ByteTrack和BoT-SORT相结合时,取得了出色的性能,平均精度均值(mAP50)为98.4%,推理速度为0.3毫秒。该系统即使在快速移动和部分遮挡的情况下也能确保稳健跟踪,显著提高了MIS中的精度和手术效率。
这种自动化的深度学习解决方案减轻了手术团队的认知负担,增强了患者安全,并优化了手术效率。它与物联网、移动应用和智能手术室的潜在整合凸显了其在现代手术中的变革性作用,为人工智能驱动的腹腔镜手术和手术创新树立了新的标准。