Automated Planning and Robotic Delivery of Needle Biopsies under CT Image Guidance

点击翻译按钮获取中文摘要

PROJECT SUMMARY / ABSTRACT Primary lung cancer is by far the leading cause of cancer death worldwide, with approximately 150,000 deaths yearly in the United States. When symptoms arise, the lung cancer survival rate at five years is a dismal 17%. Lung cancer screening with low-dose CT has been shown to reduce mortality from lung cancer among high- risk patients as the cancer is typically caught early, in stage IA. Definitive diagnosis requires tissue sampling, and despite risks of pneumothorax, sampling is often performed by percutaneous transthoracic lung biopsies under CT guidance. Since sampling can cause immediate perilesional hemorrhage and obscure views of the lesion, there is little room for error. However, current procedural challenges, involving translating the patient in- and-out of the bore repetitively for frequent freehand needle adjustments and advancements, introduce errors, take significant time, cost, and confers ionizing radiation and risk of complications to the patient. The purpose of this project is to develop an autonomous needle biopsy procedure performed under artificially intelligent robot guidance, optimizing for patient safety and targeting accuracy. The approach involves (i) a highly dexterous, force-sensitive redundant robot design with an active needle placer that operates inside the CT scanner, and can articulate and steer needles for any thoracic approach or patient position; (ii) an artificially intelligent planner that finds new, less traumatic, and safer approaches to biopsy lesions in a patient-specific manner, and (iii) closed-loop CT-image feedback control to precisely steer needles to suspect lesions. The work of this project is to be carried out via the following Specific Aims: (1) develop the force sensitive robot based on our previous robotic designs and validate on real human cases for reachability and safety analysis, (2) develop metrics and algorithms for planning needle biopsy approaches in a patient- specific way, and (3) develop high-fidelity breathing phantoms across biologically relevant variables and compare the automated robotic approach to freehanded needle placement a user study. The proposed approach offers a solution that could significantly broaden the approach direction and positioning of needles for biopsies. The semi-autonomy provides significant value by computing a variety of factors in planning the approach that optimizes accuracy and safety, including patient anatomy, safety from sensitive structures, and depth of insertions --- all of which can also have a significant effect on patient health during screening, who are already have compromised pulmonary status or have elevated risk of acute pneumothorax. Finally, the integration of machine learning, automation, and robotics reduces the variation between clinicians, leverages population data to make data-backed informed plans, and can reach super- human precision while reducing procedure time and ionizing imaging. Our long-term goal is to evaluate how the semi-autonomous approach can be advantageous to, more effective than, and/or affordable to traditional manual biopsy approaches. The outcome of this project will be a validated system ready for a clinical study.