Incorporating task-level autonomy into teleoperation systems for robot assisted laparoscopic surgery requires operators to effectively manage transitions from task-level autonomy to teleoperation control, particularly during system failures. This study investigated the impact of the level of guidance during take-over situations on participants’ performance and their trust in the teleoperation system. N = 34 students, unfamiliar with the system and task, performed a ring transfer task under two conditions: visual guidance, where they monitored the system visually, and visuo-haptic guidance, where they also received haptic feedback by holding the input device as the system operated in task-level autonomy. During some trials, system failures occurred, requiring participants to take over control. Results showed that participants noticed system failures faster with visuo-haptic guidance compared to visual guidance. Although visuo-haptic guidance did not improve task completion times or accuracy during task completion, it led to significantly higher trust in the teleoperation system and a stronger sense of perceived control. However, participants also reported higher physical demand and effort under visuo-haptic guidance, indicating an increased workload in this condition. In conclusion, these findings reveal important trade-offs of the influence of haptic guidance during human oversight of teleoperated autonomous systems.
We investigated whether projection restoration before reconstruction, followed by image-domain refinement, improves sparse-view computed tomography (CT) quality under aggressive undersampling. A compact dual-domain Swin framework restores sparse sinograms before geometry-aware differentiable filtered back projection (FBP) and refines the resulting images. The projection branch uses gated Swin processing and angular upsampling with a bicubic bypass. Abdominal AAPM sinograms contained 2304 full views, retaining 576, 288, and 144 views at 4x, 8x, and 16x. Comparators were FBP, bilinear and bicubic interpolation, ProCT, and geometry-adapted DuDoTrans; ablations used ten epochs. DuDoSwin achieved PSNR/SSIM/LPIPS of 45.64 dB/0.981/0.011 at 4x, 44.11 dB/0.975/0.020 at 8x, and 42.02 dB/0.963/0.035 at 16x. Mean PSNR and SSIM exceeded ProCT and DuDoTrans; DuDoTrans had lower LPIPS at 4x and 8x. DuDoSwin used 1.36–1.80 million parameters but was slower than both learned comparators. Comparisons against interpolation and DuDoTrans were significant in Holm-corrected paired tests. Bypass removal improved ablation PSNR by 0.35 dB at 8x. Evaluation included the validation patient; exclusion changed mean PSNR by at most 0.1 dB and altered no ranking. Slice-wise tests ignore within-patient dependence. Explicit projection restoration followed by image-domain refinement yielded favorable PSNR and SSIM, with perceptual metric and computational trade-offs. Neither residual shortcut improved the 8x ablation results. Independent patient evaluation remains necessary.
Statistical shape models (SSMs) from three-dimensional (3D) medical imaging describe morphology variation in normal and pathologic anatomy. Within the foot and ankle, weight-bearing computed tomography (WBCT) is used to characterize skeletal deformity but poorly visualizes soft tissue. Magnetic resonance imaging (MRI) provides improved evaluation of soft tissue structures but less accurate bone visualization. This methodologic analysis describes and tests reliability of a novel process to generate SSMs from coregistered MRI and WBCT structures. This method is illustrated via two example models: a combined musculoskeletal model and a distal tibia growth plate model. Bilateral WBCTs and T1-weighted DIXON MRIs of the lower legs were collected from 14 individuals aged 8 to 18 with no prior lower limb surgery in a control group and hereditary neuropathy group. Bones were segmented from WBCT and manually aligned in MRI space. Alignment reliability was evaluated with intraclass correlation coefficients (ICCs). Six muscles were segmented semi-automatically and the distal tibia growth plate was manually segmented from MRI. Following cropping, size normalization, and alignment multi-domain SSMs were generated. SSMs were evaluated qualitatively through comparison with expected anatomy and quantitively by calculation of compactness, generalization, and specificity. The alignment method demonstrated excellent reliability with an intrarater ICC of 1 and interrater ICC of 0.999. The musculoskeletal model demonstrated expected atrophy in neuromuscular disease and the preliminary growth plate model showed variation in metaphyseal thickness. Both models demonstrated compactness, specificity, and generalization comparable to published SSMs, requiring 11 and 12 modes to account for 95
Granular jamming grippers (GJGs) offer high geometric adaptability and intrinsic safety for robot-assisted surgical instrument handling but struggle with thin, planar instruments. This study investigates whether external pinching through a passive and partly compliant force-leveraging linkage can enhance grasping performance while maintaining a compact and simple design. The SCIP-GRIP, a novel GJG with an integrated Scott–Russell linkage to convert vertical down force into horizontal pinching, was developed. The performances of three gripper versions with varying pinching heights were compared to that of a non-pinching GJG across three instrument types: scalpel (flat), forceps (vertical surfaces), and scissors (undercuts). Both, holding forces and pickup success rates were evaluated using a Universal Testing Machine. Holding forces were measured 12 times per case, while success rates were assessed over 50 trials each. Pinching improves performance across all instruments, with the largest gains observed for scissors, achieving holding forces above 10 N. For forceps and scalpels, only the lowest pinching height yields significant improvements. Success rates for scissors and forceps remain at 100
Segmenting vascular plaques in the carotid arteries, aorta, and iliac arteries on CT is important because plaque burden relates to cardiovascular and stroke risk and can guide treatment. However, manual labeling is often infeasible and time-consuming across large vascular regions. Semi-supervised learning (SSL) can leverage unlabeled CT data to enable accurate plaque segmentation with only limited labels. We applied a diffusion-based semi-supervised learning (SSL) method for vascular plaque segmentation in non-contrast CT by formulating the task as multi-class segmentation of plaques and 26 adjacent anatomical structures. The method employs a teacher–student architecture with a shared diffusion encoder to learn domain-invariant feature representations and separate teacher and student decoders to decouple labeled and unlabeled data, reducing overfitting and improving robustness. To address the severe class imbalance caused by the small size of plaques, we developed a plaque-aware oversampling strategy and a labeled-data-guided training for the shared encoder and student decoder. We used 142 NIH non-contrast CT scans, divided into 119 training scans and 23 validation scans. We externally evaluated the method on 95 scans from the University of Wisconsin and 50 scans from Medipol University Mega Hospital and compared its performance with supervised nnU-Net and a strong SSL baseline. The improved SSL method significantly outperformed the original SSL method and supervised training with the same architecture when using 14 labeled training scans ( p<0.0001 ). With these 14 labeled training scans, it achieved DSC values of 58.8
