Percutaneous nephrolithotomy, a minimally invasive procedure, is widely used for the management of kidney stones [7]. The key point of this procedure is to establish a safe percutaneous access, which could influence the subsequent operation.

Residual stones and renal hemorrhage still occur due to inaccurate puncture [8]. In particular, massive renal hemorrhage may require renal artery embolization or even nephrectomy, which seriously endangers patient safety [9]. Percutaneous puncture is the most critical step when performing percutaneous nephrolithotomy. Currently, fluoroscopy and ultrasound guidance are the preferred methods for percutaneous puncture. However, proficiency in these techniques involves a steep learning curve, often necessitating approximately 60 surgeries [4].

Currently, urologists are trying to improve puncture accuracy through the use of advanced guidance devices and puncture needles. Wu et al. introduced a visual navigation system to aid puncture, which could successfully assist the puncture of calices, thereby reducing the puncture time and radiation dose [10]. In a retrospective study, Francesco et al. reported that three-dimensional mixed reality hologram guidance for percutaneous puncture was safe, that the procedure was highly effective in clinical practice, and that it was associated with a low intraoperative radiation exposure [11]. New imaging techniques, such as three-dimensional printing, visual needles, and electromagnetic navigation system, aim to simplify percutaneous puncture and reduce complications [12,13,14]. In our study, the navigation system based on deep learning and mixed reality, which could achieve real-time navigation, was applied for the treatment of kidney stones during percutaneous nephrolithotomy.

Accurate puncture of ideal calices is crucial step in percutaneous nephrolithotomy. Through ideal calycles, it is possible to reach as many kidney collection systems as possible and improve the stone-free rate. M. Rassweiler et al. postulated that iPad-assisted navigation could ensure accurate positioning, and assisted surgeon in establishing kidney access [15]. In a controlled study, computer assisted percutaneous nephrolithotomy demonstrated obvious advantages in terms of puncture success rates compared with standard percutaneous nephrolithotomy [16]. The percutaneous nephrolithotomy three-dimensional model, as reported by Wang et al., demonstrated notable accuracy and provided valuable guidance for puncture [17]. In our study, while the stone-free rate remained unchanged, the puncture time was reduced in the surgical navigation group (4.8 ± 0.8 min vs. 4.4 ± 0.7 min, P = 0.003).

Traditionally, fluoroscopy or ultrasound is typically used for guiding percutaneous puncture in percutaneous nephrolithotomy. However, radiation exposure inevitably occurs when fluoroscopic guidance is used. In a single academic center study, the optimal target calyces were identified using a flexible ureteroscope and punctured with a needle tip sensor guided by a real-time navigation system, and radiation exposure was significantly different between the two methods (p < 0.01) [18]. To reduce radiation exposure, we employed a navigation system based on deep learning and mixed reality to assist ultrasound-guiding percutaneous puncture, which achieved radiation-free.

To date, ultrasound guidance has garnered increased attention and is used to perform percutaneous puncture because there is no radiation. However, there are limitations in poor observation of the needle, and it is crucial for surgeons to closely monitor the position of the needle in the kidney in real time. John et al. revealed that a computerized needle navigation training system could accurately identify needles in ultrasound images [19]. Chau et al. demonstrated that magnetic field-based navigation ultrasound could visualize the position of the puncture needle relative to the target calyces during puncture [20]. In the surgical navigation group, the navigation system accurately identified the puncture needle in CT images. The difference in the success rate of a single puncture between the two groups was found to be statistically significant.

Although the risk of trauma is significantly lower than that of traditional surgical lithotomy, percutaneous nephrolithotomy can still cause complications, and bleeding is a prevalent and serious complications. Yamaguchi et al. reported that the incidence of bleeding during percutaneous nephrolithotomy was as high as 9.4% [21]. While the majority of bleeding can be managed conservatively, severe bleeding may require selective renal artery embolization, which occurs in approximately 1% of patients [3]. Therefore, it is more important to prevent bleeding than to take remedial measures after bleeding. Meng et al. analyzed computed tomography angiography before surgery to select puncture sites away from large vessels to reduce the risk of bleeding. This approach ensured a greater safety during the procedure [22]. To reduce bleeding, we selected puncture sites with fewer vessels according to computed tomography angiography. The percutaneous puncture was guided by the navigation system in the surgical navigation group, and the surgeon accurately penetrated the target calyces, minimizing the risk of bleeding.

Despite advances in technology and equipment, percutaneous puncture remains the most crucial step in percutaneous nephrolithotomy and necessitates extensive surgical procedures to achieve proficiency. In the last decade, urology has adopted three-dimensional printing technology for training [23, 24]. Electromagnetic guided puncture has a high success rate and a short learning curve for beginners [14]. To simplify and enhance puncture success, especially for novices, we attempted to perform percutaneous puncture with a navigation system based on deep learning and mixed reality. Compared with the control group, the surgical navigation group demonstrated a shorter puncture time. The difference in the success rate of single punctures was also found to be statistically significant.

However, several limitations should be noted in the current study, which we plan to address in future studies. First, the small sample size of our study necessitates further research with larger samples across multiple centers. Second, the retrospective nature of our study introduces selection bias, and a prospective cohort study is warranted for more definitive conclusions. Third, we performed all procedures under ultrasound guidance, limiting the applicability of our findings to fluoroscopy guidance. Nevertheless, both fluoroscopy and ultrasound guidance are primary methods used in urology.