Sudha Garimella

Data Driven Development of Haptic Models for Needle Biopsy Phantoms

Needle biopsy is an important and common procedure for lesion detection or tissue extraction within the human body. Physicians conducting such procedures rely primarily on the sense of “touch” (kinesthetic feedback from needle) to estimate the current needle position and organs within its vicinity. This skill takes time to acquire and mature, often by biopsies on live patients. Medical residents and fellow trainees thus have limited opportunities both in terms of real life scenarios as well as testing platforms to develop and validate their skills. This paper focuses on building a biopsy simulator for training on virtual phantoms (using both visual and force feedback) and cross validation using a real physical phantom. In order to develop a virtual-haptic model of biopsy phantom, material testing experiments were conducted to obtain motion-force profiles from an instrumented 6-DOF robot platform serving as a needle driver. The measured force-displacement data was then used to develop three types of haptic models for the phantom to calculate the force feedback for the haptic device. Neural network based models provided a more accurate force-reflection model compared to the other two methods from the literature and will form the basis of the virtual phantoms within our framework.Copyright

The Current State of the Diagnosis and Management of Acute Kidney Injury by Pediatric Critical Care Physicians.

Objectives: Increasingly prevalent in pediatric intensive care, acute kidney injury imparts significant short- and long-term consequences. Despite advances in acute kidney injury research, clinical outcomes are worsening. We surveyed pediatric critical care physicians to describe the current state of acute kidney injury diagnosis and management in critically ill children. Design: Anonymous electronic questionnaire. Participants: Pediatric critical care physicians from academic centers, the Pediatric Acute Lung Injury and Sepsis Investigators network, and/or the pediatric branch of Society of Critical Care Medicine. Interventions: None. Measurements and Main Results: Of 201 surveys initiated, 170 surveys were more than 50% completed and included in our results. The majority of physicians (74%) diagnosed acute kidney injury using serum creatinine and urine output. Acute kidney injury guidelines or criteria were used routinely by 54% of physicians; Risk, Injury, Failure, Loss, and End stage criteria were the most commonly used. Awareness of any acute kidney injury guideline or definition was associated with five-fold higher odds of using any guideline (odds ratio, 5.22; 95% CI, 1.84–14.83) and four-fold higher odds of being dissatisfied with available acute kidney injury biomarkers (odds ratio, 4.88; 95% CI, 1.58–15.05). Less than half of respondents recognized the limitations of serum creatinine. Physicians unaware of the limitations of serum creatinine had two-fold higher odds of being unaware of newer biomarker availability (odds ratio, 2.34; 95% CI, 1.14–4.79). Novel biomarkers were available to 37.6% of physicians for routine use. Physicians with access to novel biomarkers more often practiced in larger (odds ratio, 3.09; 95% CI, 1.18–8.12) and Midwestern (odds ratio, 3.38; 95% CI, 1.47–7.78) institutions. More physicians with access to a novel biomarker reported satisfaction with current acute kidney injury diagnostics (66%) than physicians without access (48%); this finding approached significance (p = 0.07). Conclusions: Half of PICU attending physicians surveyed are not using recent acute kidney injury guidelines or diagnostic criteria in their practice. There is a positive association between awareness and clinical use of acute kidney injury guidelines. Serum creatinine and urine output are still the primary diagnostics; novel biomarkers are frequently unavailable.

VIDEO-ANALYTICS FOR ENHANCING SAFETY & DECISION-SUPPORT IN SURGICAL WORKFLOWS

The present state-of-the-art surgical robotic systems though have a host of safety and failure systems still rely primarily on the surgeons-in-the-loop for safety amidst real-world uncertainties. Recent case studies involving surgical accidents and injuries have further increased the attention towards: (i) incorporating multiple levels of hard- and soft- safety layers within the testing, validation and deployment phases; and (ii) objective based surgical skill assessment. We therefore would like to propose a method based on the rich information content with automated-assists for safer and smarter robotic surgeries. BACKGROUND Surgical tool detection and tracking, as it has been considered so far, is found to be useful only for low level feedback (Krupa, Gangloff et al. (2003)). Besides, prior tracking efforts based on motion detection using external sensing— marker-based methods that had to be bio-compatible (Groeger, Arbter et al. (2008)) and flexible joint sensing resulting in uncertainties; and image-analysis— assume clear conditions and heavily rely on tool geometric information (Reiter, Allen et al. (2012)), were limited. In this work, our aim is to realize: (i) automated tracking and semantic labeling of tools; and (ii) surgical expertise evaluation and assessment based on micro-motion analysis; using only the recorded surgical videos. Though these aspects have been studied already, we believe our efforts will be a critical enabler towards increasing the level-of-autonomy. EXPERIMENTAL SETUP Recorded videos of hysterectomy procedures and simulator tasks using the da Vinci Surgical System-Si served as our inputs. The uncalibrated videos were cropped into segments (150-300 frames per slice) without making assumptions on tools and attributes (allowing varied tool configurations and types; occlusion by tissues, blood stains and other tools; imagespecularities due to varied light intensities and smoke etc.) to yield robust classifiers. The proposed cascaded framework (Kumar, Narayanan et al. (2013)) is shown in Fig.1.

Radial Basis Function Network (RBFN) Approximation of Finite Element Models for Real-Time Simulation

Nonlinearities inherent in soft-tissue interactions create roadblocks to realization of high-fidelity real-time haptics-based medical simulations. While finite element (FE) formulations offer greater accuracy over conventional spring-mass-network models, computational-complexity limits achievable simulation-update rates. Direct interaction with sensorized physical surrogates, in offline or online modes, allows a temporary sidestepping of computational issues but hinders parametric analysis and true exploitation of a simulation-based testing paradigm. Hence, in this paper, we develop Radial-Basis Neural-Network approximations, to FE-model data within a Modified Resource Allocating Network (MRAN) framework. Real-time simulation of the reduced order neural-network approximations at high temporal resolution provided the haptic-feedback. Validation studies are being conducted to evaluate the kinesthetic realism of these models with medical experts.Copyright

985: USE OF NONINVASIVE METHODS TO DETECT SEPSIS-INDUCED ACUTE KIDNEY INJURY IN CRITICALLY ILL CHILDREN

Learning Objectives: Sepsis is the most common cause of acute kidney injury (AKI) in critically ill pediatric patients in the United States. Current clinical practice relies on serum creatinine (SCr), creatinine clearance, and urine output to detect AKI; despite these being late and insensitive markers of AKI. The resistive index (RI) in the renal artery, as determined by Doppler methods, has proven to be a useful technique to quantify alterations in renal blood flow. Renal near infrared regional spectroscopy (RNIRS) is a novel technology that has shown to correlate with elevated SCr values 48hr later when placed on infants after cardiac surgery. Our study was performed to test the hypothesis that RNIRS and or RI could detect clinical changes consistent with sepsis-associated AKI before the changes in SCr occur. Methods: Prospective observational study of critically ill pediatric patients aged 2 weeks to 20 yr old with suspected or confirmed sepsis. RNIRS probes were in place until 48 hr after sepsis resolution. Using bedside Sonosite ultrasound machine, renal arterial RI was evaluated every 12–24 hr. The primary outcome of AKI was determined by an increase in adjusted SCr by 25% from baseline using the definition of “Risk” “Injury” “Failure” “Loss” and “End-stage” (pRIFLE) AKI staging criteria. SCr values were adjusted for net fluid balance. Results: Of the 18 patients enrolled in the study, four patients (22.2%) developed “Risk” according to pRIFLE criteria. Lower RNIRS values were associated with SCr increases 1 and 2 days later in septic AKI (rs=-0.775; p=0.041 and rs=-0.900; p=0.039 respectively). There were no meaningful associations between RI and changes in SCr. 8 patients had technical problems with the application of RNIRS including displacement of probes with movement, sweating. Conclusions: Lower RNIRS on sepsis day 2 and 3 were associated with changes in renal function on sepsis day 3 to 4 and days 5 to 6 in critically ill children. RNIRS can be technically challenging in children with obesity, scoliosis, excessive sweating. Bedside Doppler RI was not reliable in predicting AKI in this study.

965: PEDIATRIC CRITICAL CARE PHYSICIANS’ PERSPECTIVE ON “FLUID OVERLOAD”

Learning Objectives: Recent evidence has shown that fluid overload (FO) is an independent diagnosis with increased risk of morbidity and mortality in critically ill children with all types of conditions. This study was performed to describe the current perspective of pediatric critical care physicians on the diagnosis and management of FO. Methods: An electronic questionnaire sent to an international sample of attending physicians in pediatric critical care intended to measure the current state of AKI practice, with one third of the content specific to FO. Results: Of the 201 surveys begun, 170 (85%) were completed. Most respondents (94%) reported fluid balance as a daily “vital” sign and prescribed daily weights in those at risk for FO (76%). Almost all physicians (92%) were using FO to detect AKI and the majority (62%) was aware that FO impacts serum creatinine. When asked to define FO, the most prevalent answer was that FO is not an independent diagnosis (37%). The most frequent number threshold used for FO was 10% as chosen by 25% of respondents, the remainder used “gestalt.” Only 4% chose the intended answer of 15% or net positive more than 150mL per kilogram. The majority of the physicians (73%) denied that avoiding FO can prevent AKI. In patients at risk for FO or AKI, respondents were split over the use of fluid restriction: 56% fluid restrict and 44% do not. Physicians who reported that avoiding FO can prevent AKI had 4.8 fold higher odds of fluid restricting at risk patients (OR 4.76, 95%CI 2.02–11.21, p<0.0001). Although 86% of physicians report FO was an independent indication for renal replacement therapy (RRT), only 36% used strict numeric thresholds to initiate RRT. More than 50% of respondents reported using urine output as a decision point in initiating RRT for FO. Conclusions: Pediatric critical care physicians recognize the importance of daily fluid balance as a “vital” sign but are not routinely diagnosing FO as an independent medical problem. Strict numeric thresholds are not being used to diagnose FO or to initiate RRT.