Artificial Intelligence-Guided Segmentation and Path Planning Software for Transthoracic Lung Biopsy.

PURPOSE: To validate the sensitivity and specificity of a 3-dimensional (3D) convolutional neural network (CNN) artificial intelligence (AI) software for lung lesion detection and to establish concordance between AI-generated needle paths and those used in actual biopsy procedures. MATERIALS AND METHODS: This was a retrospective study using computed tomography (CT) scans from 3 hospitals. Inclusion criteria were scans with 1-5 nodules of diameter ≥5 mm; exclusion criteria were poor-quality scans or those with nodules measuring <5mm in diameter. In the lesion detection phase, 2,147 nodules from 219 scans were used to develop and train the deep learning 3D-CNN to detect lesions. The 3D-CNN was validated with 235 scans (354 lesions) for sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) analysis. In the path planning phase, Bayesian optimization was used to propose possible needle trajectories for lesion biopsy while avoiding vital structures. Software-proposed needle trajectories were compared with actual biopsy path trajectories from intraprocedural CT scans in 150 patients, with a match defined as an angular deviation of <5° between the 2 trajectories. RESULTS: The model achieved an overall AUC of 97.4% (95% CI, 96.3%-98.2%) for lesion detection, with mean sensitivity of 93.5% and mean specificity of 93.2%. Among the software-proposed needle trajectories, 85.3% were feasible, with 82% matching actual paths and similar performance between supine and prone/oblique patient orientations (P = .311). The mean angular deviation between matching trajectories was 2.30° (SD ± 1.22); the mean path deviation was 2.94 mm (SD ± 1.60). CONCLUSIONS: Segmentation, lesion detection, and path planning for CT-guided lung biopsy using an AI-guided software showed promising results. Future integration with automated robotic systems may pave the way toward fully automated biopsy procedures.

Phantom and Animal Study of a Robot-Assisted, CT-Guided Targeting System using Image-Only Navigation for Stereotactic Needle Insertion without Positional Sensors.

PURPOSE: To evaluate the feasibility and accuracy of a robotic system to integrate and map computed tomography (CT) and robotic coordinates, followed by automatic trajectory execution by a robotic arm. The system was hypothesized to achieve a targeting error of <5 mm without significant influence from variations in angulation or depth. MATERIALS AND METHODS: An experimental study was conducted using a robotic system (Automated Needle Targeting device for CT [ANT-C]) for needle insertions into a phantom model on both moving patient table and moving gantry CT scanners. Eight spherical markers were registered as targets for 90 insertions at different trajectories. After a single ANT-C registration, the closed-loop software targeted multiple markers via the insertion of robotically aligned 18-gauge needles. Accuracy (distance from the needle tip to the target) was assessed by postinsertion CT scans. Similar procedures were repeated to guide 10 needle insertions into a porcine lung. A regression analysis was performed to test the effect of needle angulation and insertion depth on the accuracy of insertion. RESULTS: In the phantom model, all needle insertions (median trajectory depth, 64.8 mm; range, 46.1-153 mm) were successfully performed in single attempts. The overall accuracy was 1.36 mm ± 0.53, which did not differ between the 2 types of CT scanners (1.39 mm ± 0.54 [moving patient table CT] vs 1.33 mm ± 0.52 [moving gantry CT]; P = .54) and was not significantly affected by the needle angulation and insertion depth. The accuracy for the porcine model was 9.09 mm ± 4.21. CONCLUSIONS: Robot-assisted needle insertion using the ANT-C robotic device was feasible and accurate for targeting multiple markers in a phantom model.

Fabrication and characterization of bit-patterned media beyond 1.5 Tbit/in2.

We fabricated bit-patterned media (BPM) at densities as high as 3.3 Tbit/in(2) using a process consisting of high-resolution electron-beam lithography followed directly by magnetic film deposition. By avoiding pattern transfer processes such as etching and liftoff that inherently reduce pattern fidelity, the resolution of the final pattern was kept close to that of the lithographic step. Magnetic force microscopy (MFM) showed magnetic isolation of the patterned bits at 1.9 Tbit/in(2), which was close to the resolution limit of the MFM. The method presented will enable studies on magnetic bits packed at ultra-high densities, and can be combined with other scalable patterning methods such as templated self-assembly and nanoimprint lithography for high-volume manufacturing.