In spinal surgery, ensuring surgical precision and safety is paramount. Traditionally, surgeons have relied on their experience to determine when to cease milling as the cutter approaches the spinal cord; However, improper technique during this process can lead to complications, such as vertebral plate fractures and spinal cord injuries. This paper investigates the development of a robot capable of high-precision recognition of the milling state. Initially, we identify vibration signals as the basis for state recognition, establishing their feasibility through theoretical analysis, which provides a foundation for the creation of datasets for subsequent milling experiments. We then conducted milling experiments using pig scapulae and designed neural networks for state identification. Vibration signals corresponding to varying milling depth and the proportion of cortical and cancellous bone layers were collected. A BiLSTM-based neural network was developed to identify the milling depth and the proportion of the bone layers, achieving the desired outcomes within an acceptable error range.The results demonstrate that the proposed system achieves high accuracy in state recognition, with errors falling within an acceptable range. This research highlights the potential of integrating advanced neural networks and vibration analysis into robotic systems to enhance precision and safety in spinal surgery.

Bone milling is widely used in orthopaedic surgery. But the high-speed rotating ball-end milling tool (BMT) may cause damage during surgery. This article presents a method for sensing and controlling the milling state (angle and depth) of surgical robots based on acoustic signals. First, a theoretical model of acoustic signal and milling state is proposed. Second, a dual band-pass filter and feature extraction method are designed according to signal frequency domain characteristics. Then, a robotic platform is built to complete calibration experiments and obtain model function parameters. Finally, comparative experiments are conducted to validate the method. Experimental results show that adding angle control can improve milling depth accuracy and reduce bone compression force compared with only depth control. The method can effectively enhance surgical results.

This article aims to optimize robotic actions for bone milling during laminectomy. The layer-by-layer spinal lamina milling operation during robot-assisted laminectomy is analyzed in detail to summarize related control tasks. A milling sound signal processing method based on band filters and the two-order Prony algorithm is proposed to overcome the frequency disturbance of the surgical power device's rotation motion caused by bone milling to accurately extract harmonic amplitudes. A depth estimation model based on the total value of major harmonics is fitted to monitor the actual depth when the ball end is cutting into an easily displaced bone. A milling feed speed optimization method based on the sound signal feedback and a fuzzy controller is designed to balance the efficiency and safety of bone milling. Milling experiments are carried out on artificial bone blocks and fresh swine cervical vertebrae. The results show that the depth monitoring accuracy is about 0.08 mm and the feed speed can be optimized effectively to match the actual bone mineral density and axial depth of cut. The proposed method can improve the depth accuracy and output a safe and fastest feed speed.

During bent bone grinding, the robot often needs to adjust its grinding depth and angle in real time to ensure surgical quality. Optical- or vision-based navigation methods are no longer applicable due to the interference caused by splashing cooling water in the operating field. Inspired by orthopedic surgeons, in this article, we present a tactile perception-based method to solve this problem. The effect of grinding parameters on the cutter's current and vibration is analyzed, and the current and vibration signals are further processed to predict depth and angle. Depth and angle estimation models were calibrated on homogeneous artificial bone. Grinding control experiments were carried out on femur models, and the statistical mean and standard deviation results of the measured depth show that the accuracy of depth control can be further improved by adding angle estimation and proper adjustment when depth is indirectly estimated based on sensing signals. The proposed approach relies solely on the estimation of the relative position and posture between the tool and the bone surface and is expected to improve the safety of automatic grinding with orthopedic robots.

Various deep learning methods have been proposed to segment breast lesions from ultrasound images. However, similar intensity distributions, variable tumor morphologies and blurred boundaries present challenges for breast lesions segmentation, especially for malignant tumors with irregular shapes. Considering the complexity of ultrasound images, we develop an adaptive attention U-net (AAU-net) to segment breast lesions automatically and stably from ultrasound images. Specifically, we introduce a hybrid adaptive attention module (HAAM), which mainly consists of a channel self-attention block and a spatial self-attention block, to replace the traditional convolution operation. Compared with the conventional convolution operation, the design of the hybrid adaptive attention module can help us capture more features under different receptive fields. Different from existing attention mechanisms, the HAAM module can guide the network to adaptively select more robust representation in channel and space dimensions to cope with more complex breast lesions segmentation. Extensive experiments with several state-of-the-art deep learning segmentation methods on three public breast ultrasound datasets show that our method has better performance on breast lesions segmentation. Furthermore, robustness analysis and external experiments demonstrate that our proposed AAU-net has better generalization performance in the breast lesion segmentation. Moreover, the HAAM module can be flexibly applied to existing network frameworks. The source code is available on https://github.com/CGPxy/AAU-net.

To optimize the cutting depth in robotic-assisted laminectomy, we present a real-time method to adjust the preoperatively planned feed rate in the depth direction of the robot cutting trajectory. Not only the linearity between the harmonic amplitude of the milling acoustic signal and the cutting depth is discussed by analyzing the milling dynamic model, but its influencing variables are analyzed. The amplitude of the harmonic components whose frequency are integer multiples of the spindle frequency of the surgical power tool is extracted by FFT (Fast Fourier transform). A digital PD (Proportional-Differential) controller with a dead zone generates the speed compensation amount according to the deviation of the harmonic amplitude from the expected value. In artificial bone sensitivity test experiments, the cutting depth can be estimated with a resolution of 0.15mm within the cutting depth range of 0-1.2mm by the harmonic amplitude signal. Furthermore, the safety of the proposed method under different bone deformations is verified by cutting depth control experiments.

Surgical navigation is challenging on complicated multi-branch structures such as intrarenal collecting systems or bronchi. The objective of this work is to help surgeons quickly establish the corresponding relationship between intraoperative endoscopic images and preoperative CT data. An endoscopic navigation method is proposed based on three-dimensional structure registration. It mainly includes three sections. First, a reconstruction method is presented to obtain three-dimensional information of porous structures from endoscopic images. It combines image enhancement, structure-from-motion and template matching. Second, a hole search strategy based on slicing is given for detecting and extracting three-dimensional porous structures from CT data. Third, a similarity measurement algorithm is developed for registering endoscopic images to CT data. The performance of this work is evaluated on the data from the ureteroscopic holmium laser lithotripsy and the results show its accuracy, robustness and time cost.

Bone milling is widely used in many hard tissue surgical procedures, and the main concern is the mechanical damage to some important tissues induced by the high-speed rotating tool. In this study, the behavior of the human auditory system is analyzed. Following this, a commercially available microphone is mounted on the robot arm and then measures the sound generated from bone milling. Inspired by the bandpass filtering in the cochlea, the recorded sound pressure signal is decomposed into a set of subband signals by wavelet packet transform. Inspired by encoding and perception mechanisms in human auditory system, the average amplitude of the wavelet coefficients in each subband is calculated and inputted to a self-organizing feature map so as to classify different milling states. In order to increase the robustness to noise, the sum of the Manhattan distances between all winning neurons is calculated to select the optimum dimension of the map. The experimental results in milling in vitro porcine spines prove that the proposed method can determine which type of tissue is being cut when the suction noise exists, and the success classification rate is no less than 85%. Therefore, the safety of the robot-assisted milling surgery is improved.

One of the major challenge in spine surgery is successfully cutting the desired bony structure while avoiding injury to the vital anatomy such as the spinal cord. This paper presents a vibration signal processing method to discriminate different types of tissue in robot-assisted spine surgery. During bone milling process, the tissue vibration signal measured by a laser displacement sensor is decomposed into some frequency sub-bands through the wavelet packet transform, and the harmonic component whose frequency is an integer times of the spindle frequency is obtained. The wavelet energy of the 1st, 2ndand 3rdharmonics is then used as input vector to an artificial neural network for discriminating tissue. To verify the effectiveness of the proposed method, extensive milling experiments are carried out on porcine spines. The experimental results indicate that the method produces up to 100% discrimination accuracy for the vertebra being cut and the spinal cord, and yields the success discrimination rate of more than 80% for the adjacent bony structure and the muscle, so the safety of robot-assisted spine surgery is improved.