MicroSegNet: A deep learning approach for prostate segmentation on micro-ultrasound images.

Micro-ultrasound (micro-US) is a novel 29-MHz ultrasound technique that provides 3-4 times higher resolution than traditional ultrasound, potentially enabling low-cost, accurate diagnosis of prostate cancer. Accurate prostate segmentation is crucial for prostate volume measurement, cancer diagnosis, prostate biopsy, and treatment planning. However, prostate segmentation on micro-US is challenging due to artifacts and indistinct borders between the prostate, bladder, and urethra in the midline. This paper presents MicroSegNet, a multi-scale annotation-guided transformer UNet model designed specifically to tackle these challenges. During the training process, MicroSegNet focuses more on regions that are hard to segment (hard regions), characterized by discrepancies between expert and non-expert annotations. We achieve this by proposing an annotation-guided binary cross entropy (AG-BCE) loss that assigns a larger weight to prediction errors in hard regions and a lower weight to prediction errors in easy regions. The AG-BCE loss was seamlessly integrated into the training process through the utilization of multi-scale deep supervision, enabling MicroSegNet to capture global contextual dependencies and local information at various scales. We trained our model using micro-US images from 55 patients, followed by evaluation on 20 patients. Our MicroSegNet model achieved a Dice coefficient of 0.939 and a Hausdorff distance of 2.02 mm, outperforming several state-of-the-art segmentation methods, as well as three human annotators with different experience levels. Our code is publicly available at https://github.com/mirthAI/MicroSegNet and our dataset is publicly available at https://zenodo.org/records/10475293.

Assessing cancer risk in the anterior part of the prostate using micro-ultrasound: validation of a novel distinct protocol.

PURPOSE: To develop and validate a micro-ultrasound risk score that predicts the likelihood of significant prostate cancer in the anterior zone. METHODS: Patients were enrolled from three expert institutions familiar with micro-ultrasound. The study was conducted in two phases. First, the PRI-MUS anterior score was developed by assessing selected prostate videos from patients who subsequently underwent radical prostatectomy. Second, seven urology readers with varying levels of experience in micro-ultrasound examination evaluated prostate loops according to the PRI-MUS anterior score. Each reader watched the videos and recorded the likelihood of the presence of significant cancer in the anterior part of the prostate in a three-point scale. The coherence among the readers was calculated using the Fleiss kappa and the Cronbach alpha. RESULTS: A total of 102 selected prostate scans were used to develop the risk assessment for anterior zone cancer in the prostate. The score comprised three categories: likely, equivocal, and unlikely. The median (IQR) sensitivity, specificity, positive predictive value, and negative predictive value for the seven readers were 72% (68-84), 68% (64-84), 75% (72-81), and 73% (71-80), respectively. The mean SD ROC AUC was 0.75 ± 2%, while the Fleiss kappa and the Cronbach alpha were 0.179 and 0.56, respectively. CONCLUSION: Micro-ultrasound can detect cancerous lesions in the anterior part of the prostate. When combined with the PRI-MUS protocol to assess the peripheral part, it enables an assessment of the entire prostate gland. Pending external validation, the PRI-MUS anterior score developed in this study might be implemented in clinical practice.

Reducing retraction forces with tactile feedback during robotic total mesorectal excision in a porcine model.

Excessive tissue-instrument interaction forces during robotic surgery have the potential for causing iatrogenic tissue damages. The current in vivo study seeks to assess whether tactile feedback could reduce intraoperative tissue-instrument interaction forces during robotic-assisted total mesorectal excision. Five subjects, including three experts and two novices, used the da Vinci robot to perform total mesorectum excision in four pigs. The grip force in the left arm, used for retraction, and the pushing force in the right arm, used for blunt pelvic dissection around the rectum, were recorded. Tissue-instrument interaction forces were compared between trials done with and without tactile feedback. The mean force exerted on the tissue was consistently higher in the retracting arm than the dissecting arm (3.72 ± 1.19 vs 0.32 ± 0.36 N, p < 0.01). Tactile feedback brought about significant reductions in average retraction forces (3.69 ± 1.08 N vs 4.16 ± 1.12 N, p = 0.02), but dissection forces appeared unaffected (0.43 ± 0.42 vs 0.37 ± 0.28 N, p = 0.71). No significant differences were found between retraction and dissection forces exerted by novice and expert robotic surgeons. This in vivo animal study demonstrated the efficacy of tactile feedback in reducing retraction forces during total mesorectal excision. Further research is required to quantify the clinical impact of such force reduction.

A System for Co-Registration of High-Resolution Ultrasound, Magnetic Resonance Imaging, and Whole-Mount Pathology for Prostate Cancer.

In order to evaluate the diagnostic accuracy of high-resolution ultrasound (HRUS) for detection of prostate cancer, it must be validated against whole-mount pathology. An ex-vivo HRUS scanning system was developed and tested in phantom and human tissue experiments to allow for in-plane computational co-registration of HRUS with magnetic resonance imaging (MRI) and whole-mount pathology. The system allowed for co-registration with an error of 1.9mm±1.4mm, while also demonstrating an ability to allow for lesion identification.Clinical Relevance- Using this system, a workflow can be established to co-register HRUS with MRI and pathology to allow for the diagnostic accuracy of HRUS to be determined with direct comparison to MRI.

Bio-Inspired Haptic Feedback for Artificial Palpation in Robotic Surgery.

Adding haptic feedback has been reported to improve the outcome of minimally invasive robotic surgery. In this study, we seek to determine whether an algorithm based on simulating responses of a cutaneous afferent population can be implemented to improve the performance of presenting haptic feedback for robot-assisted surgery. We propose a bio-inspired controlling model to present vibration and force feedback to help surgeons localize underlying structures in phantom tissue. A single pair of actuators was controlled by outputs of a model of a population of cutaneous afferents based on the pressure signal from a single sensor embedded in surgical forceps. We recruited 25 subjects including 10 expert surgeons to evaluate the performance of the bio-inspired controlling model in an artificial palpation task using the da Vinci surgical robot. Among the control methods tested, the bio-inspired system was unique in allowing both novices and experts to easily identify the locations of all classes of tumors and did so with reduced contact force and tumor contact time. This work demonstrates the utility of our bio-inspired multi-modal feedback system, which resulted in superior performance for both novice and professional users, in comparison to a traditional linear and the existing piecewise discrete algorithms of haptic feedback.

Multi-Modal Haptic Feedback for Grip Force Reduction in Robotic Surgery.

Minimally invasive robotic surgery allows for many advantages over traditional surgical procedures, but the loss of force feedback combined with a potential for strong grasping forces can result in excessive tissue damage. Single modality haptic feedback systems have been designed and tested in an attempt to diminish grasping forces, but the results still fall short of natural performance. A multi-modal pneumatic feedback system was designed to allow for tactile, kinesthetic, and vibrotactile feedback, with the aims of more closely imitating natural touch and further improving the effectiveness of HFS in robotic surgical applications and tasks such as tissue grasping and manipulation. Testing of the multi-modal system yielded very promising results with an average force reduction of nearly 50% between the no feedback and hybrid (tactile and kinesthetic) trials (p < 1.0E-16). The multi-modal system demonstrated an increased reduction over single modality feedback solutions and indicated that the system can help users achieve average grip forces closer to those normally possible with the human hand.

Artificial palpation in robotic surgery using haptic feedback.

BACKGROUND: The loss of tactile feedback in minimally invasive robotic surgery remains a major challenge to the expanding field. With visual cue compensation alone, tissue characterization via palpation proves to be immensely difficult. This work evaluates a bimodal vibrotactile system as a means of conveying applied forces to simulate haptic feedback in two sets of studies simulating an artificial palpation task using the da Vinci surgical robot. METHODS: Subjects in the first study were tasked with localizing an embedded vessel in a soft tissue phantom using a single-sensor unit. In the second study, subjects localized tumor-like structures using a three-sensor array. In both sets of studies, subjects completed the task under three trial conditions: no feedback, normal force tactile feedback, and hybrid vibrotactile feedback. Recordings of correct localization, incorrect localization, and time-to-completion were used to evaluate performance outcomes. RESULTS: With the addition of vibrotactile and pneumatic feedback, significant improvements in the percentage of correct localization attempts were detected (p = 0.0001 and p = 0.0459, respectively) during the first experiment with phantom vessels. Similarly, significant improvements in correct localization were found with the addition of vibrotactile (p = 2.57E-5) and pneumatic significance (p = 8.54E-5) were observed in the second experiment involving tumor phantoms. CONCLUSIONS: This work demonstrates not only the superior benefits of a multi-modal feedback over traditional single-modality feedback, but also the effectiveness of vibration in providing haptic feedback to artificial palpation systems.

Suture Breakage Warning System for Robotic Surgery.

As robotic surgery has increased in popularity, the lack of haptic feedback has become a growing issue due to the application of excessive forces that may lead to clinical problems such as intraoperative and postoperative suture breakage. Previous suture breakage warning systems have largely depended on visual and/or auditory feedback modalities, which have been shown to increase cognitive load and reduce operator performance. This work catalogues a new sensing technology and haptic feedback system (HFS) that can reduce instances of suture failure without negatively impacting performance outcomes including knot quality. Suture breakage is common in knot-tying as the pulling motion introduces prominent shear forces. A shear sensor mountable on the da Vinci robotic surgical system's Cadiere grasper detects forces that correlate to the suture's internal tension. HFS then provides vibration feedback to the operator as forces near a particular material's failure load. To validate the system, subjects tightened a total of four knots, two with the Haptic Feedback System (HFS) and two without feedback. The number of suture breakages were recorded and knot fidelity was evaluated by measuring knot slippage. Results showed that instances of suture failure were significantly reduced when HFS was enabled (p = 0.0078). Notably, knots tied with HFS also showed improved quality compared to those tied without feedback (p = 0.010). The results highlight the value of HFS in improving robotic procedure outcomes by reducing instances of suture failures, producing better knots, and reducing the need for corrective measures.