A Data-Centric Approach to improve performance of deep learning models.

The Artificial Intelligence has evolved and is now associated with Deep Learning, driven by availability of vast amount of data and computing power. Traditionally, researchers have adopted a Model-Centric Approach, focusing on developing new algorithms and models to enhance performance without altering the underlying data. However, Andrew Ng, a prominent figure in the AI community, has recently emphasized on better (quality) data rather than better models, which has given birth to Data Centric Approach, also known as Data Oriented technique. The transition from model oriented to data oriented approach has rapidly gained momentum within the realm of deep learning. Despite its promise, the Data-Centric Approach faces several challenges, including (a) generating high-quality data, (b) ensuring data privacy, and (c) addressing biases to achieve fairness in datasets. Currently, there has been limited effort in preparing quality data. Our work aims to address this gap by focusing on the generation of high-quality data through methods such as data augmentation, multi-stage hashing to eliminate duplicate instances, to detect and correct noisy labels, using confident learning. The experiments on popular datasets, namely MNIST, Fashion MNIST, and CIFAR-10 were performed by utilizing ResNet-18 as the common framework followed by both Model Centric and Data Centric Approach. Comparative performance analysis revealed that the Data Centric Approach consistently outperformed the Model Centric Approach by a relative margin of at least 3%. This finding highlights the potential for further exploration and adoption of the Data-Centric Approach in various domains such as healthcare, finance, education, and entertainment, where the quality of data could significantly enhance the performance.

Network meta analysis to predict the efficacy of an approved treatment in a new indication.

Drug repurposing refers to the process of discovering new therapeutic uses for existing medicines. Compared to traditional drug discovery, drug repurposing is attractive for its speed, cost, and reduced risk of failure. However, existing approaches for drug repurposing involve complex, computationally-intensive analytical methods that are not widely used in practice. Instead, repurposing decisions are often based on subjective judgments from limited empirical evidence. In this article, we develop a novel Bayesian network meta-analysis (NMA) framework that can predict the efficacy of an approved treatment in a new indication and thereby identify candidate treatments for repurposing. We obtain predictions using two main steps: first, we use standard NMA modeling to estimate average relative effects from a network comprised of treatments studied in both indications in addition to one treatment studied in only one indication. Then, we model the correlation between relative effects using various strategies that differ in how they model treatments across indications and within the same drug class. We evaluate the predictive performance of each model using a simulation study and find that the model minimizing root mean squared error of the posterior median for the candidate treatment depends on the amount of available data, the level of correlation between indications, and whether treatment effects differ, on average, by drug class. We conclude by discussing an illustrative example in psoriasis and psoriatic arthritis and find that the candidate treatment has a high probability of success in a future trial.

Relation of mobilization after hip fractures on day of surgery to length of stay.

This study examined whether evaluation by physical therapy on the day of surgery impacts length of stay in patients with hip fractures. A total of 528 adult patients with hip fracture requiring surgery were prospectively included in the study over a 22-month period. The median length of stay of mobilized patients was 4.0 days, vs 5.0 days in nonmobilized patients (P = 0.0158). Of the 259 mobilized patients, 64 were discharged home (24.71%) vs 47 (17.84%) in the nonmobilized cohort (P = 0.0434), with no increase in readmission rates. In conclusion, mobilization on the day of surgery reduced length of stay by 1 day, with a higher percentage of patients being discharged home.

Factors that promote successful endoscopic management of laparoscopic sleeve gastrectomy leaks.

INTRODUCTION: Staple line leaks following laparoscopic sleeve gastrectomy (LSG) are associated with significant morbidity and mortality. Endoluminal techniques, including stent placement and endoluminal vacuum therapy (EVAC), have become viable options to treat these patients without the need for additional surgery. The purpose of this study was to define the conditions where certain endoscopic therapies are most likely to succeed compared to surgery. METHODS: An IRB approved prospectively maintained database was retrospectively reviewed for all patients treated for gastrointestinal leaks from July 2013 to March 2019. All patients who were treated for gastrointestinal leaks following LSG were included. Endpoints include success of leak closure and hospital-related morbidity for the patients treated solely by endoscopic only methods (EP) compared to the additional surgery group (SP). RESULTS: There were 39 patients (33 females; 6 males) with a median age of 45.9 years. The EP group included 23 patients (59%), whereas SP included 16 patients (31%). On average, the SP had longer days from sentinel surgery to our hospital admission (70 vs 41), a higher percentage of previous bariatric surgery prior to sentinel LSG (50% vs 17%), and a higher readmission rates following discharge (50% vs 39%). Total length of stay was also higher in the SP compared to the EP (45.4 vs 11). Using this data, a treatment algorithm was developed to optimally treat future patients who suffer from gastrointestinal leaks following LSG. CONCLUSIONS: Endoscopic therapies, such as EVAC, stent placement, internal drainage, and over-the-scope clips, have a higher chance of success if performed earlier to their sentinel surgery and if patients have had no prior bariatric surgeries. Patients who require additional surgery tend to have longer hospital stays and readmission rates. Using the treatment algorithm provided can help determine when endoscopic therapies are likely to succeed.

The role of preoperative workup in predicting dysphagia, dilation, or explantation after magnetic sphincter augmentation.

BACKGROUND: Magnetic sphincter augmentation (MSA) is a surgical treatment for gastroesophageal reflux disease using a ring of titanium beads to improve the function of the lower esophageal sphincter. Prior to implantation, a comprehensive preoperative esophageal workup is required to determine patient candidacy in an effort to reduce the dysphagia, dilation, and explantation rate of the device. This study was designed to assess the best predictors for these endpoints. METHODS: A prospectively maintained IRB-approved database was retrospectively reviewed for patients undergoing MSA implantation. Patients were divided into 3 groups, those that needed no intervention, those that needed medical intervention with oral steroids for reported dysphagia, and surgical intervention, which included endoscopic dilation and/or surgical explantation. Primary endpoints included preoperative objective and subjective testing from a comprehensive esophageal workup including intraoperative notation of number of beads on the device. RESULTS: There were 99 patients eligible for the study with a mean age of 52 and mean follow-up of 10.2 months. Mean BMI was 27 and 59% were female. The no-intervention group had 59 patients, medical intervention group had 25 patients, and surgical intervention group had 15 patients. Preoperative esophageal manometry findings, pH testing off medications, endoscopic and radiologic evaluation showed no difference between the 3 groups. No differences were seen in preoperative subjective evaluations based on GERD-HRQL or RSI scores. There was no difference in average number of beads on the device between the 3 groups. CONCLUSION: A comprehensive esophageal workup is important to confirm the presence of gastroesophageal reflux disease and rule out other esophageal pathology. However, this study shows that a preoperative comprehensive esophageal workup does not predict which patients will develop dysphagia or require either medical or surgical interventions following MSA implantation.

Impact of a Community-Based Healthy Lifestyle Program on Individuals With Traumatic Brain Injury.

OBJECTIVES: To examine adherence with and effect of an evidence-based healthy lifestyle intervention modified for individuals with traumatic brain injury (TBI). DESIGN: Pre-/postintervention without control. SETTING: Community. PARTICIPANTS: Eighteen individuals with TBI: primarily male (61%), white (67%), with private insurance (50%). Mean age was 45.6 ± 12.3 years, weight 210 ± 42.6 lb, and body mass index 31.8 ± 4.6 (obese category) at baseline. INTERVENTIONS: The primary goal of the Diabetes Prevention Program Group Lifestyle Balance program is 5% to 7% weight loss through increased physical activity and improved dietary behaviors. MAIN OUTCOME MEASURE(S): Adherence (ie, session attendance and self-monitoring of dietary behaviors), physiologic changes (ie, weight loss, blood pressure; waist and arm circumference; and lipid profile), and quality of life (ie, self-reported health, quality of life, and step count). RESULTS: Average participant attendance (85% over 12 months) and self-monitoring (90% over 6 months) were high. Significant decreases were observed in diastolic blood pressure and waist and arm circumference from baseline through 12 months and from baseline to 3 months only for weight and total cholesterol. No significant changes were observed in self-reported health, quality of life, or step count. CONCLUSIONS: Participants demonstrated high adherence with the program, suggesting that individuals with TBI are able to successfully engage in the program and achieve significant weight loss and changes in key physiologic outcomes.

Sacral examination in spinal cord injury: Is it really needed?

OBJECTIVE: To determine if a self-report measure of S4-5 motor and sensory function in patients with chronic SCI accurately predicts sacral examination results. DESIGN: Prospective, single-blinded self-report survey compared with sacral exam. SETTING: Outpatient SCI clinic. PARTICIPANTS: 116 patients aged 18+ with chronic SCI > 6 months who have undergone sacral exam. INTERVENTIONS: The survey included demographic/clinical and sacral function information such as light tough (LT), pinprick sensation (PP), deep anal pressure (DAP) and voluntary anal contraction (VAC). Survey results and sacral exam were compared and stratified by the patient's American Spinal Cord Injury Association Impairment Scale (AIS) category. OUTCOME MEASURES: Sacral self-report survey, AIS examination. RESULTS: Mean age was 41.3 ± 14.4 years with majority male (69%) and Caucasian (71.6%). Overall, Positive Predictive Value (PPV) ranged between 48% (VAC) to 73% (DAP) and Negative Predictive Value (NPV) between 92% (VAC) to 100% (LT). AIS-A had NPV of 100% across all categories, and AIS-D had PPV of 100% across all categories. CONCLUSION: Patient report of sacral sparing can predict negative sensation in patients with AIS-A and predict positive sensation in persons with AIS-D. Overall, the self-report of sacral sparing of motor and sensory function is not predictive enough to rely on for accurate classification.

Surgical training in robotic surgery: surgical experience of robotic-assisted transabdominal preperitoneal inguinal herniorrhaphy with and without resident participation.

Robotic-assisted surgery is becoming more popular in general surgery. Implementation of a robotic curriculum is necessary and will influence surgical training. The aim of this study is to compare surgical experience and outcomes with and without resident participation in robotic inguinal herniorrhaphy. A retrospective review of patients who underwent either unilateral or bilateral robotic-assisted transabdominal preperitoneal (TAPP) inguinal herniorrhaphy, with and without resident participation as console surgeons from January through December 2015, was performed. Patient demographics, procedure-related data, postoperative variables, and follow-up data were analyzed. A total of 104 patients were included. Patients were significantly older in the Resident group (57.5 ± 14.1 vs 50.6 ± 13.5 years, p = 0.01). Gender, BMI, and ASA classification were similar between groups. There were similar mean operative times for unilateral (89.9 ± 19.5 vs 84.8 ± 22.2 min, p = 0.42) and bilateral (128.4 ± 21.9 vs 129.8 ± 50.9 min, p = 0.90) inguinal herniorrhaphy as well as mean robot console times for unilateral (73.2 ± 18.4 vs 67.3 ± 29.9 min, p = 0.44) and bilateral (115.5 ± 24.6 vs 109.3 ± 55.4 min, p = 0.67) inguinal herniorrhaphy with and without resident participation, respectively. Postoperative complications included urinary retention (11.1 vs 2.0%, p = 0.11), conversion to open repair (0 vs 2%, p = 0.48), and delayed reoperation (0 vs 4%, p = 0.22) with and without resident participation, respectively. Patients' symptoms/signs at follow-up were similar among groups. Robotic-assisted TAPP inguinal herniorrhaphy with resident participation as console surgeons did not affect the hospital operative experience or patient outcomes. This procedure can be implemented as part of the resident robotic curriculum with rates of morbidity equivalent to those of published studies.Level of evidence 2b.

Postoperative continuous catheter-infused local anesthetic reduces pain scores and narcotic use after lower extremity revascularization.

Objective Postoperative pain following lower extremity revascularization procedures is traditionally controlled with narcotic administration. However, this may not adequately control the pain and puts the patient at risk for complications from opiate use. Here we report an alternative strategy for pain management using a continuous catheter-infused local anesthetic into the operative limb. Design Retrospective case-control study. Methods Patients undergoing lower extremity revascularization procedures using continuous catheter-infused local anesthetic were compared to similar patients undergoing similar procedures during the same time period who did not receive continuous catheter-infused local anesthetic. Records were reviewed for pain scores, narcotics consumption, length of stay, need for postoperative chest X-ray, supplemental oxygen use, wound complications, and 30-day readmission. Results There were 153 patients (mean age 69.5 years) from September 2011 to December 2014 who underwent common femoral artery procedures, femoral-popliteal bypass, femoral-tibial bypass, popliteal aneurysm repair, popliteal to pedal bypass, popliteal artery thrombo-embolectomy, sapheno-popliteal venous bypass, or ilio-femoral bypass. There were no significant differences between the continuous catheter-infused local anesthetic ( n=57) and control ( n=96) groups regarding age, body mass index, cardiac history, diabetes, hypertension, and procedures performed. The continuous catheter-infused local anesthetic group showed better cumulative average pain scores, better high pain scores on postoperative days 1-3, and better average pain scores on postoperative days 2-3 ( P<0.03). The continuous catheter-infused local anesthetic group had lower median narcotics consumption on postoperative days 1-2 ( P=0.02). No differences were found in postoperative length of stay, urinary catheter use, number of postoperative chest X-rays, oxygen use, mobilization, or fever. Wound complications occurred in 8.8% of the continuous catheter-infused local anesthetic group and in 11.5% of controls (P=0.79). Readmission rates were 23% (continuous catheter-infused local anesthetic) and 21% (controls; P=0.84). Conclusion Postoperative continuous catheter-infused local anesthetic reduces pain scores and pain medication use compared to standard opiate therapy in these patients, without increasing wound complication or readmission rates. Continuous catheter-infused local anesthetic appeared to have no effect on the incidence of pulmonary complications, mobilization, or fever.

Application of virtual environments in a multi-disciplinary day neurorehabilitation program to improve executive functioning using the Stroop task.

BACKGROUND: Virtual reality (VR) technology has demonstrated usefulness in diagnosis, education, and training. Studies supporting use of VR as a therapeutic treatment in medical rehabilitation settings remain limited. This study examines the use of VR in a treatment capacity, and whether it can be effectively integrated into neurorehabilitation. OBJECTIVE: To determine whether immersive VR treatment interventions improve executive dysfunction in patients with brain injury and whether performance is stronger on a VR version of the Stroop than traditional Stroop formats. METHODS: 15 patients with brain injury admitted to day neurorehabilitation. OUTCOME MEASURES: reaction time, inhibition, and accuracy indices on VR Stroop; Automated Neuropsychological Assessment Metrics (ANAM) Stroop, Delis-Kaplan Executive Function System Stroop, Golden Stroop, and Woodcock-Johnson, 3rd Edition (WJ-III): Pair Cancellation. RESULTS: Participants demonstrated significantly reduced response time on the word-reading condition of VR Stroop and non-significantly reduced response time on the interference condition. Non-significant improvements in accuracy and inhibition were demonstrated on the color-naming condition of VR Stroop. Significantly improved accuracy under time pressure was found for the ANAM, after VR intervention. CONCLUSION: Implementation of immersive VR interventions during neurorehabilitation is effective in improving specific executive functions and information processing speed in brain-injured patients during the subacute period.

A comparison of lipid minimization strategies in children with intestinal failure.

PURPOSE: The purpose of this study was to compare outcomes of lipid minimization with either Intralipid (IL) or Omegaven® in children with intestinal failure (IF) who developed intestinal failure-associated liver disease (IFALD) while receiving parenteral nutrition (PN). METHODS: A retrospective review of children with IF requiring PN who developed IFALD (direct bilirubin >2 mg/dL) while receiving IL (2009-2016) was performed. Clinical characteristics, nutritional, and laboratory values were compared between children treated with reduced IL or Omegaven®. RESULTS: 16 children were reviewed (8 treated with IL and 8 treated with Omegaven® at a median dose of 1g/kg/d). Both groups had similar demographics, small bowel length, and parenteral nutritional intake during the study (82.9±27.1 kcal/kg/d vs. 75.9±16.5 kcal/kg/d, p=0.54). The mean direct bilirubin (DBili) prior to initiating treatment was 7.8±4.3 mg/dL and 7.5±3.5 mg/dL (p=0.87) in the IL and Omegaven® groups, respectively. The IL group took a median of 113 days to achieve a DBili <0.5 mg/dL compared to 124 days in the Omegaven® group (p=0.49). There were no differences in markers of liver function or growth trajectories among groups. CONCLUSIONS: Lipid minimization with either IL or Omegaven® has similar success in achieving a normal DBili in children with IF and IFALD without major differences in nutritional status or growth. TYPE OF STUDY: Treatment Study LEVEL OF EVIDENCE: III.

The association between posttraumatic stress symptoms, depression, and length of hospital stay following traumatic injury.

OBJECTIVE: The present study examined the relationship between posttraumatic stress symptoms (PTSS) and depression symptoms with hospital outcome measures to explore how psychiatric factors relate to hospital length of stay (LOS). METHOD: Participants were adults admitted to a large Level I Trauma Center for longer than 24h. Depression was assessed at hospitalization using the Patient Health Questionnaire (PHQ-8), and PTSS was measured by the Primary Care PTSD Screen (PC-PTSD). Hospital outcome information was collected from the hospital's trauma registry. Pearson correlations were performed. RESULTS: 460 participants (mean age=44years, SD=16.8; 65.4% male) completed the study. Baseline PTSS and depression were significantly correlated with longer hospital LOS while controlling for demographics and injury severity (p=0.026; p=0.023). Both PTSS-positive and depression-positive groups had an average increased hospital LOS of two days. CONCLUSIONS: A significant proportion of individuals who are admitted to the hospital following trauma may be at risk for depression and PTSS, which may then increase hospital LOS. As national attention turns to reducing healthcare costs, early screenings and interventions may aid in minimizing psychiatric symptoms in trauma patients, in turn reducing the cost and outcomes associated with total hospital LOS.

Traumatic colon injury in damage control laparotomy-A multicenter trial: Is it safe to do a delayed anastomosis?

BACKGROUND: Delayed colonic anastomosis after damage control laparotomy (DCL) is an alternative to colostomies during a single laparotomy (SL) in high-risk patients. However, literature suggests increased colonic leak rates up to 27% with DCL, and various reported risk factors. We evaluated our regional experience to determine if delayed colonic anastomosis was associated with worse outcomes. METHODS: A multicenter retrospective cohort study was performed across three Level I trauma centers encompassing traumatic colon injuries from January 2006 through June 2014. Patients with rectal injuries or mortality within 24 hours were excluded. Patient and injury characteristics, complications, and interventions were compared between SL and DCL groups. Regional readmission data were utilized to capture complications within 6 months of index trauma. RESULTS: Of 267 patients, 69% had penetrating injuries, 21% underwent DCL, and the mortality rate was 4.9%. Overall, 176 received primary repair (26 in DCL), 90 had resection and anastomosis (28 in DCL), and 26 had a stoma created (10 end colostomies and 2 loop ileostomies in DCL). Thirty-five of 56 DCL patients had definitive colonic repair subsequent to their index operation. DCL patients were more likely to be hypotensive; require more resuscitation; and suffer acute kidney injury, pneumonia, adult respiratory distress syndrome, and death. Five enteric leaks (1.9%) and three enterocutaneous fistulas (ECF, 1.1%) were identified, proportionately distributed between DCL and SL (p = 1.00, p = 0.51). No difference was seen in intraperitoneal abscesses (p = 0.13) or surgical site infections (SSI, p = 0.70) between cohorts. Among SL patients, pancreas injuries portended an increased risk of intraperitoneal abscesses (p = 0.0002), as did liver injuries in DCL patients (p = 0.06). CONCLUSIONS: DCL was not associated with increased enteric leaks, ECF, SSI, or intraperitoneal abscesses despite nearly two-thirds having delayed repair. Despite this being a multicenter study, it is underpowered, and a prospective trial would better demonstrate risks of DCL in colon trauma. LEVEL OF EVIDENCE: Therapeutic study, level IV.

Predicting time to full enteral nutrition in children after significant bowel resection.

PURPOSE: Parenteral nutrition (PN) contributes to considerable morbidity in children after significant bowel resection. This study evaluates the utility of clinical variables in predicting time to independence from PN. METHODS: After IRB approval, a retrospective review (1999-2012) of 71 children who were on PN for >6weeks after intestinal resection and successfully weaned was performed. Clinical characteristics were evaluated to determine the relationship to time to full enteral nutrition. P-values<0.05 were significant. RESULTS: Most children had necrotizing enterocolitis (56%), intestinal atresia (20%), or gastroschisis (11%) with a median small bowel length of 55cm (IQR, 35-92cm). The duration of PN was independent of the etiology of intestinal loss, presence of the ileocecal valve or colon, or location of anastomosis, but was strongly associated with small bowel length (<0.01) and percent of expected small bowel based on gestational age (GA) (median 50%, <0.01). In general, children who had 25-50% of their small bowel were dependent on PN for at least 2years compared to approximately 1year for those with 51-75%. CONCLUSION: The duration of PN dependence in children after major bowel resection is best predicted by remaining small bowel length and can be estimated using a linear regression model. LEVEL OF EVIDENCE: 2b.