Search

Solving program induction problems requires searching through an enormous space of possibilities. DreamCoder is an inductive program synthesis system that, whilst solving problems, learns to simplify search in an iterative wake-sleep procedure. The cost of search is amortised by training a neural search policy, reducing search breadth and effectively "compiling" useful information to compose program solutions across tasks. Additionally, a library of program components is learnt to express discovered solutions in fewer components, reducing search depth. In DreamCoder, the neural search policy has only an indirect effect on the library learnt through the program solutions it helps discover. We present an approach for library learning that directly leverages the neural search policy, effectively "decompiling" its amortised knowledge to extract relevant program components. This provides stronger amortised inference: the amortised knowledge learnt to reduce search breadth is now also used to reduce search depth. We integrate our approach with DreamCoder and demonstrate faster domain proficiency with improved generalisation on a range of domains, particularly when fewer example solutions are available.

Multimodal VAEs seek to model the joint distribution over heterogeneous data (e.g.\ vision, language), whilst also capturing a shared representation across such modalities. Prior work has typically combined information from the modalities by reconciling idiosyncratic representations directly in the recognition model through explicit products, mixtures, or other such factorisations. Here we introduce a novel alternative, the MEME, that avoids such explicit combinations by repurposing semi-supervised VAEs to combine information between modalities implicitly through mutual supervision. This formulation naturally allows learning from partially-observed data where some modalities can be entirely missing---something that most existing approaches either cannot handle, or do so to a limited extent. We demonstrate that MEME outperforms baselines on standard metrics across both partial and complete observation schemes on the MNIST-SVHN (image--image) and CUB (image--text) datasets. We also contrast the quality of the representations learnt by mutual supervision against standard approaches and observe interesting trends in its ability to capture relatedness between data.

Background Xpert MTB/RIF assay is a rapid automated molecular test with excellent reported sensitivity, specificity for diagnosis of pulmonary and extrapulmonary Mycobacterium tuberculosis (MTB) infections. However, the clinical utility and accuracy in STB is not well established. A study was conducted to report on the sensitivity, specificity and clinical utility of the Xpert MTB/RIF assay in spinal tuberculosis (STB). Methods A retrospective review of medical records was performed for 136 patients that underwent spinal biopsy for suspected spondylodiscitis. Reports for acid fast bacilli (AFB) smear, gram stain, pyogenic culture, MTB culture, histopathology, Xpert MTB/RIF assay, and drug sensitivity testing were reviewed. ‘Reference standard for diagnosis of STB’ was based on positive histopathology and/or MTB culture evidence and was considered as MTB positive. Any samples returning a positive pyogenic or fungal culture were considered as MTB negative. The sensitivity, specificity for Xpert MTB/RIF was assessed against the reference standard. Results A total of 125 patients were considered for final analysis, 86 patients met the criteria for ‘Reference standard for diagnosis of TB spine’ (MTB positive). This includes nine patients that were MGIT culture only positive; 45 that were histopathology only positive and 32 were both culture and histopathology positive. There were 39 culture proven (pyogenic-37 and fungal-2) patients included in MTB negative group. The 86 MTB positive patients, included 53 (61.6%) tissue samples and 33 (38.4%) pus samples. The overall analysis showed a 65.1% sensitivity, 100% specificity, 100% PPV and 56.5% NPV for the Xpert MTB/RIF. Conclusions Gene Xpert MTB/RIF showed excellent specificity and was accurate in the identification of drug resistance. The sensitivity was 65% and sampling techniques using pus samples rather than tissue samples could be a possible reason for lower sensitivity.

Study Design: A multicenter, pilot study, for radiological assessment of thoracolumbar spine fractures was performed with the participation of 7 centers in Africa, Europe, Asia, and South America as a part of the AO Foundation network. Objectives: To determine the interobserver variability for computed tomography (CT) scan–based evaluation of posterior ligament complex (PLC) injury in thoracolumbar fractures. Methods: Forty-two observers including 1 principal investigator at each participating center performed variability assessment. Each center contributed toward a total of 91 patient images with A3 or A4 thoracolumbar burst fractures (T11-L2) with or without suspected PLC injury. Pathological fractures, multilevel injuries, obvious posterior bony element injury and translation/dislocation injuries were excluded. Ten patients were randomly selected and commonly reported CT parameters indicating PLC injury, including superior inferior endplate angle, vertebral body height loss, local kyphotic deformity, interspinous distance and interpedicular distance were assessed for variability. Observer values were compared with an experienced gold rater in spinal trauma. Analysis of variability was performed for all observers, between the principal investigators and also between observers participating in each center. Results: The studied parameters showed considerable variability in measurements among all observers and amongst all participating centers. The variability between the principal investigators was lower, but still substantial. The deviation of observer measurements from the gold rater were also significant for all CT parameters. Conclusions: CT-based radiological parameters previously reported to be suggestive of PLC injury showed considerable variability and magnetic resonance imaging verification of a PLC injury in all doubtful cases is suggested.

Multimodal learning for generative models often refers to the learning of abstract concepts from the commonality of information in multiple modalities, such as vision and language. While it has proven effective for learning generalisable representations, the training of such models often requires a large amount of related multimodal data that shares commonality, which can be expensive to come by. To mitigate this, we develop a novel contrastive framework for generative model learning, allowing us to train the model not just by the commonality between modalities, but by the distinction between "related" and "unrelated" multimodal data. We show in experiments that our method enables data-efficient multimodal learning on challenging datasets for various multimodal VAE models. We also show that under our proposed framework, the generative model can accurately identify related samples from unrelated ones, making it possible to make use of the plentiful unlabeled, unpaired multimodal data.

The COVID-19 pandemic has highlighted the importance of in-silico epidemiological modelling in predicting the dynamics of infectious diseases to inform health policy and decision makers about suitable prevention and containment strategies. Work in this setting involves solving challenging inference and control problems in individual-based models of ever increasing complexity. Here we discuss recent breakthroughs in machine learning, specifically in simulation-based inference, and explore its potential as a novel venue for model calibration to support the design and evaluation of public health interventions. To further stimulate research, we are developing software interfaces that turn two cornerstone COVID-19 and malaria epidemiology models COVID-sim, (https://github.com/mrc-ide/covid-sim/) and OpenMalaria (https://github.com/SwissTPH/openmalaria) into probabilistic programs, enabling efficient interpretable Bayesian inference within those simulators.

Diabetes and Obesity are major risk factors which confer vulnerability to Covid 19 . Diabetes has immune defects which makes the individual susceptible to infections and covid 19 is no exception . Also covid 19 can cause pancreatic damage as well as stress hyperglycaemia in hospitals which may need Insulin . Among diabetes male gender,elderly,hypertension ,heart disease and chronic renal disease are more vulbwdvale to covid 19 and need strict supervision . Diabetes management in hospitalised situation merits early diabetes specific nutrition with Insulin. Adherence to lifestyle with self monitoring of blood glucose and adequate supply of Insulin and Oral antidiabetic agents is encouraged.

Background Shoulder balance (SB) is an important cosmetic concern and is a determinant to assess outcomes following scoliosis surgery. Shoulder imbalance (SI) has been studied predominantly in idiopathic scoliosis with limited literature on other etiologies. Purpose To assess SB following scoliosis surgery and correlate SRS-30 self-image scores with radiological shoulder imbalance SI. Methods A retrospective analysis was performed for 43 patients undergoing scoliosis surgery with a minimum 2-year follow-up. Radiological SI was assessed by the coracoid height difference on whole spine standing radiographs at preoperative, 12-weeks postoperative and 2-year follow-up. The preoperative Cobb angle, postoperative Cobb angle and mean correction were recorded. The upper and lower end vertebrae (UEV/LEV), and instrumented vertebrae (UIV/LIV) were documented. The aetiology of scoliosis and functional outcomes based on the SRS-30 questionnaire were noted. Results There were 19 males and 24 females with a mean age of 14.5 ± 4.7 years. The 12-week postoperative radiographs showed SI in 15 patients and residual SI in 7 patients at 2-year follow-up. SI was seen in four congenital scoliosis with segmentation or mixed anomalies in the proximal thoracic spine. Two idiopathic scoliosis (Lenke type 1 and type 3) showed SI where the UIV was T3 and T2 respectively. The mean preoperative SRS-30 score for patients with SB (N = 9) was 2.5 ± 0.72 and for SI (N = 34) was 3 ± 0.42 without significant difference (p > 0.5). The final mean SRS-30 self-image score for patients with SB was 3.7 ± 0.54 and for patients with SI was 3.7 ± 0.53 without significant difference (p > 0.05). Conclusion Radiological SI assessed by coracoid height difference was not associated with a significant difference in SRS 30 scores preoperatively and at 2 year follow up. SI was seen with congenital scoliosis associated with segmentation and mixed anomalies of the proximal thoracic spine.

Study Design: Prospective case series. Purpose: This study aimed to investigate the impact of education, financial income, occupation, and patient counseling on the timing of enrolment in a spinal cord injury (SCI) rehabilitation program. Overview of Literature: A rehabilitation program following SCI is essential to improve functional outcomes. Socioeconomic factors can affect the timing of enrolment to a rehabilitation program. Literature on the effects of socioeconomic factors among patients with SCI in the Indian scenario is limited. Methods: A prospective, consecutive analysis of patients with SCI was performed with 1-year follow-up. Assessment of the timing of enrolment to a rehabilitation program was performed using the modified Kuppuswamy socioeconomic scores (MKSS). Patients admitted to the SCI unit (group A), underwent intensive individual, group, and family counseling sessions to encourage early enrolment into a rehabilitation program. Patients presenting directly for rehabilitation (group B) were analyzed for comparison. Results: A total of 153 patients were recruited. Group A was composed of 122 patients who started the rehabilitation program after a mean of 28 days, compared with a mean of 149 days for 31 patients in group B. In group A, 104 patients (85%; mean MKSS, 14.02) and 18 patients (15%; mean MKSS, 15.61) enrolled for rehabilitation 0.05). Conclusions: Early patient counseling in the acute care unit helps in the early enrolment of patients with poor socioeconomic demographic profile to a rehabilitation program.

Study Design Prospective cohort study. Purpose The study was aimed at evaluating clinicoradiological factors affecting recovery of neurological deficits in cases of lumbar disc herniation (LDH) treated by lumbar microdiscectomy. Overview of Literature The majority of the available literature on neurological recovery following neurodeficit is limited to retrospective series. The literature is currently limited regarding variables that can help predict the recovery of neurodeficits following LDH. Methods A prospective analysis was performed on 70 consecutive patients who underwent lumbar microdiscectomy (L1–2 to L5–S1) owing to neurological deficits due to LDH. Patients with motor power ≤3/5 in L2–S1 myotomes were considered for analysis. Follow-up was performed at 2, 6, and 12 months to note recovery of motor deficits. Clinicoradiological parameters were compared between the recovered and nonrecovered groups. Results A total of 65 patients were available at the final follow-up: 41 (63%) had completely recovered by 2 months; four showed delayed recovery at the 6-month follow-up; and 20 (30.7%) showed no recovery at 1 year. Clinicoradiological factors, including diabetes, complete initial deficit, areflexia, multilevel disc prolapse, longer duration since initial symptoms, and ≥2 previous symptomatic episodes were associated with a significant risk of poorer recovery (p0.05 for all). Diabetes mellitus (p=0.033) and complete initial motor deficit (p=0.028) were significantly associated with delayed recovery in the multivariate analysis. Conclusions The overall neurological recovery rate in our study was 69%. Diabetes mellitus (p=0.033) and complete initial motor deficit were associated with delayed motor recovery.

STUDY DESIGN Prospective observational study. PURPOSE To determine the incidence of postoperative urinary retention (POUR) in patients undergoing elective posterior lumbar spine surgery and identify the risk factors associated with the development of POUR. OVERVIEW OF LITERATURE POUR following surgery can lead to detrusor dysfunction, urinary tract infections, prolonged hospital stay, and a higher treatment cost; however, the risk factors for POUR in spine surgery remain unclear. METHODS A prospective, consecutive analysis was conducted on patients undergoing elective posterior lumbar surgery in the form of lumbar discectomy, lumbar decompression, and single-level lumbar fusions during a 6-month period. Patients with spine trauma, preoperative neurological deficit, previous urinary disturbance/symptoms, multiple-level fusion, and preoperative catheterization were excluded from the study. Potential patient- and surgery-dependent risk factors for the development of POUR were assessed. Univariate analysis and a multiple logistical regression analysis were performed. RESULTS A total of 687 patients underwent posterior lumbar spine surgery during the study period; among these, 370 patients were included in the final analysis. Sixty-one patients developed POUR, with an incidence of 16.48%. Significant risk factors for POUR were older age, higher body mass index (BMI), surgery duration, intraoperative fluid administration, lumbar fusion versus discectomy/ decompression, and higher postoperative pain scores (p

Background There is a lack of a classification system providing uniformity in description and guiding management decisions for kyphotic spinal deformities. We developed such a classification based on column deficiency, flexibility of disc spaces, curve magnitude, and correlation with the corrective osteotomy required. Methods A classification was developed based on analysis of 180 patients with thoracolumbar kyphosis requiring osteotomy. The deformity was classified as Type I if the anterior and posterior columns were intact (IA indicated mobile disc spaces and IB, ankylosed segments). Type II indicated deficiency of only 1 column (IIA = anterior column and IIB = posterior column). Type III indicated deficiency of both columns (IIIA = kyphosis of ≤60°, IIIB = kyphosis of >60°, and IIIC = buckling collapse). A prospective analysis of 76 patients was performed to determine interobserver variability and the ability of the classification to guide selection of osteotomies of increasing complexity, including the Ponte osteotomy, pedicle subtraction osteotomy, disc bone osteotomy, single vertebrectomy, multiple vertebrectomies, and anterior in situ strut fusion procedure. Results The mean age of the 76 patients was 21.2 years, the mean kyphosis was 69.9° (range, 26° to 120°), and the mean follow-up duration was 30 months. Six deformities were classified as IA, 5 as IB, 5 as IIA, 2 as IIB, 13 as IIIA, 35 as IIIB, and 10 as IIIC. Four surgeons classifying the deformities had a high agreement rate (kappa = 0.83), with the highest agreement for Types IA, IB, and IIIB. A correlation between the type of deformity and the osteotomy performed demonstrated that the classification could indicate the type of osteotomy required. All 18 patients with Type-I or II kyphosis were treated with Ponte, pedicle subtraction, or disc bone osteotomy. Forty-three (90%) of the 48 patients with Type IIIA or IIIB underwent vertebrectomy (single in 27 [56%] and multiple in 16 [33%]), and only 5 (10%) underwent disc bone osteotomy. Seven of the 10 patients with Type-IIIC kyphosis were treated with multiple vertebrectomies, with 5 of them needing preoperative halo gravity traction; the other 3 patients underwent an anterior in situ strut fusion procedure. Conclusions The proposed classification based on the morphology of column deficiency, flexibility, and curve magnitude demonstrated a high interobserver agreement and ability to guide selection of the appropriate osteotomy. Clinical relevance A novel classification system for kyphosis based on spinal column deficiency, flexibility of disc spaces, and curve magnitude would bring uniformity in management and help guide surgeons in the choice of the appropriate corrective osteotomy.

Study DesignRetrospective observational study.PurposeTo analyze the effect of low-density (LD) strategic pedicle screw fixation on the correction of coronal and sagittal parameters in adolescent idiopathic scoliosis (AIS) patients.Overview of LiteratureLD screw fixation achieves favorable coronal correction, but its effect on sagittal parameters is not well established. AIS is often associated with decreased thoracic kyphosis (TK), and the use of multi-level pedicle screws may result in further flattening of the sagittal profile.MethodsA retrospective analysis was performed on 92 patients with AIS to compare coronal and sagittal parameters preoperatively and at 2-year follow-up. All patients underwent posterior correction via LD strategic pedicle screw fixation. Radiographs were analyzed for primary Cobb angle (PCA), coronal imbalance, cervical sagittal angle (CSA), TK, lumbar lordosis (LL), pelvic incidence, pelvic tilt (PT), sacral slope (SS), C7 plumb line, spino-sacral angle, curve flexibility, and screw density.ResultsPCA changed significantly from 57.6°±13.9° to 19°±8.4° (p p =0.893), TK (from 24.4° to 22.8°, p =0.145), and LL (from 50.3° to 51.1°, p =0.415). However, subgroup analysis of the hypokyphosis group (40°) showed significant correction of TK (p p =0.0001) and PT (from 15° to 14°, p =0.025).ConclusionsLD strategic pedicle screw fixation provides favorable coronal correction and improves overall sagittal sacro-pelvic parameters. This technique does not cause significant flattening of TK and results in a favorable restoration of TK in patients with hypokyphosis or hyperkyphosis.

Study DesignRetrospective case series.PurposeTo correlate functional outcomes with spinopelvic parameters in patients with high-grade spondylolisthesis (HGS) treated with instrumented in-situ surgery or reduction and fusion.Overview of LiteratureSatisfactory functional outcomes are reported with reduction and in-situ fusion strategies in HGS. However, reasons for this are unclear. We hypothesize that following lumbosacral fusion, the L5 becomes part of the sacrum, which improves spinopelvic parameters, resulting in equivalent functional outcomes in both surgical methods.MethodsTwenty-six patients undergoing HGS (reduction group A, 13; in-situ group B, 13) were clinically evaluated using the Oswestry Disability Index (ODI), short form-12 (SF-12), and Visual Analogue Scale (VAS) scores. Spinopelvic parameters, including pelvic incidence, pelvic tilt (PT), sacral slope (SS), lumbar lordosis (LL), lumbosacral kyphosis (LSK) angle, and sacrofemoral distance (SFD) were measured preoperatively from S1 and postoperatively from L5 as the new sacrum at 1 year follow-up. Sagittal alignment was assessed using the sagittal vertical axis.ResultsBoth groups were comparable in terms of age, sex, severity of slip, and preoperative spinopelvic parameters (p>0.05). Postoperative VAS, SF-12, and ODI scores significantly improved in both groups (ppin-situ fusion group PT changed from 26.78° to 11.08°, SFD from 62.9 to 36.99 mm, and LSK from 67.23° to 113.38° (pp>0.05).ConclusionsAfter fusion, the L5 becomes the new sacrum and influences spinopelvic parameters to change favorably. This possibly explains why reduction and in-situ fusion achieve equivalent functional outcomes in HGS.

To develop a classification based on the technical complexity encountered during pedicle screw insertion and to evaluate the performance of AIRO® CT navigation system based on this classification, in the clinical scenario of complex spinal deformity. 31 complex spinal deformity correction surgeries were prospectively analyzed for performance of AIRO® mobile CT-based navigation system. Pedicles were classified according to complexity of insertion into five types. Analysis was performed to estimate the accuracy of screw placement and time for screw insertion. Breach greater than 2 mm was considered for analysis. 452 pedicle screws were inserted (T1–T6: 116; T7–T12: 171; L1–S1: 165). The average Cobb angle was 68.3° (range 60°–104°). We had 242 grade 2 pedicles, 133 grade 3, and 77 grade 4, and 44 pedicles were unfit for pedicle screw insertion. We noted 27 pedicle screw breach (medial: 10; lateral: 16; anterior: 1). Among lateral breach (n = 16), ten screws were planned for in–out–in pedicle screw insertion. Among lateral breach (n = 16), ten screws were planned for in–out–in pedicle screw insertion. Average screw insertion time was 1.76 ± 0.89 min. After accounting for planned breach, the effective breach rate was 3.8% resulting in 96.2% accuracy for pedicle screw placement. This classification helps compare the accuracy of screw insertion in range of conditions by considering the complexity of screw insertion. Considering the clinical scenario of complex pedicle anatomy in spinal deformity AIRO® navigation showed an excellent accuracy rate of 96.2%.

To document a rare complication of a delayed ‘chance fracture pattern’-type injury through the proximal end of a pedicle screw construct in the clinical scenario of skeletal fluorosis. A 72-year-old man with fluorosis presented following a fall which resulted in a T12–L1 fracture. Investigations revealed an unstable three-column injury, so the patient was treated with surgical stabilisation using pedicle screw fixation from T11 to L2. He presented 1 month following surgery with worsening back pain. Investigations revealed a fracture through T11 in a ‘chance fracture pattern’ along the pedicle screw tracts at the proximal end of the construct. An extension of fixation was performed proximally to T8 and he made an uneventful recovery showing fusion at 20-month follow-up. Complication of delayed pedicle fractures, in a ‘chance fracture pattern’ at the ends of a pedicle screw fixation constructs are a rarely reported in the literature. The occurrence of such a complication in a hyperostotic spine associated with fluorosis makes this a unique clinical scenario which is previously unreported to the best of our knowledge. This report highlights a very rare complication of chance fracture pattern injury in the clinical scenario of fluorosis. A hyperostotic stiff spine, poor quality of bone and extension of pedicle screw tracts to anterior cortex during primary surgery may have resulted in the occurrence of this rare complication.