The robotic venipuncture is proved to obtain higher accuracy than manual methods in obtaining venous access for all groups of people including people with Difficult Venous Access (DVA). However, the limitations such as improper identification of vein, unstable movement of robotic arm and unexpected displacement of vein resulted in failure of venipuncture. In this paper, the authors proposed an effective model named Intelligent Controller for Stable Needle Insertion (INCSTANEIN) to achieve successful venipuncture. Initially, the ultrasound inducer is used to obtain the input images of the vein which are then processed by implementing Bi-fold filtering in which the pre-filtering is carried out using guided filter and later the filtered image is obtained by executing Speckle Reducing Bilateral Filtering (SRBF). The Filtered image is then fed into Attention-Based Convolutional Neural Network (ABCNN) in which the image is divided into several blocks and in each block, various features are extracted to achieve precise localization of vein center. Once the vein is located, the robotic arm which consists of needle is initialized with initial parameters and the future position of needle is predicted using Extended Kalman Filter (EKF) and the displacement of vein is estimated using finite-element method and the displacement error is computed to adjust the needle parameters to achieve successful venipuncture. Finally, the control of robotic arm to perform stable needle insertion is implemented using Prioritized-Deep Q Network (P-DQN) which utilizes image feedback and force feedback to accomplish successful venipuncture. The proposed INCSTANEIN model is simulated in Matlab R2020a and validated in terms of performance metrics such as accuracy, force, Y-displacement and following error to prove its efficiency.

事实证明，机器人静脉穿刺在为包括静脉通路困难 (DVA) 在内的所有人群获取静脉通路方面比手动方法获得更高的准确性。然而，静脉识别不当、机械臂运动不稳定、静脉意外移位等局限性导致静脉穿刺失败。在本文中，作者提出了一种有效的模型，称为稳定针插入智能控制器 (INCSTANEIN)，以实现成功的静脉穿刺。最初，超声诱导器用于获得静脉的输入图像，然后通过实施双倍滤波进行处理，其中使用引导滤波器进行预滤波，然后通过执行散斑减少双边滤波获得滤波后的图像(SRBF)。然后将过滤后的图像输入到基于注意力的卷积神经网络 (ABCNN) 中，其中将图像分成几个块，在每个块中提取各种特征以实现静脉中心的精确定位。一旦静脉定位，由针组成的机械臂用初始参数初始化，并使用扩展卡尔曼滤波器（EKF）预测针的未来位置，并使用有限元方法估计静脉的位移，位移误差为计算调整针参数以实现成功的静脉穿刺。最后，使用优先深度 Q 网络 (P-DQN) 控制机械臂执行稳定的针插入，该网络利用图像反馈和力反馈来完成成功的静脉穿刺。