Machine Learning Approach to Differentiation of Peripheral Schwannomas and Neurofibromas: A Multi-Center Study

BACKGROUND

Non-invasive differentiation between schwannomas and neurofibromas is important for appropriate management, preoperative counseling, and surgical planning, but has proven difficult using conventional imaging. The objective of this study was to develop and evaluate machine learning approaches for differentiating peripheral schwannomas from neurofibromas.

METHODS

We assembled a cohort of schwannomas and neurofibromas from 3 independent institutions and extracted high-dimensional radiomic features from gadolinium-enhanced, T1-weighted MRI using the PyRadiomics package on Quantitative Imaging Feature Pipeline. Age, sex, neurogenetic syndrome, spontaneous pain, and motor deficit were recorded. We evaluated the performance of 6 radiomics-based classifier models with and without clinical features and compared model performance against human expert evaluators.

RESULTS

107 schwannomas and 59 neurofibroma were included. The primary models included both clinical and imaging data. The accuracy of the human evaluators (0.765) did not significantly exceed the no-information rate (NIR), whereas the Support Vector Machine (0.929), Logistic Regression (0.929), and Random Forest (0.905) classifiers exceeded the NIR. Using the method of DeLong, the AUC for the Logistic Regression (AUC=0.923) and K Nearest Neighbor (AUC=0.923) classifiers was significantly greater than the human evaluators (AUC=0.766; p = 0.041).

CONCLUSIONS

The radiomics-based classifiers developed here proved to be more accurate and had a higher AUC on the ROC curve than expert human evaluators. This demonstrates that radiomics using routine MRI sequences and clinical features can aid in differentiation of peripheral schwannomas and neurofibromas.

Histosurgical mapping of endoscopic endonasal surgery of sinonasal tumours to improve radiotherapy guidance

PURPOSE

Endoscopic endonasal surgery (EES) is becoming a standard for most malignant sinonasal tumours. Margin analysis after piecemeal resection is complex and optimally relies on accurate histosurgical mapping. Postoperative radiotherapy may be adapted based on margin assessment mapping to reduce the dose to some sinonasal subvolumes. We assessed the use of histosurgical mapping by radiation oncologists (RO).

MATERIAL AND METHODS

A French practice survey was performed across 29 ENT expert RO (2 did not answer) regarding integration of information on EES, as well as quality of operative and pathology reportsto refine radiotherapy planning after EES. This was assessed through an electronic questionnaire.

RESULTS

EES was ubiquitously performed in France. Operative and pathology reports yielded accurate description of EES samples according to 66.7% of interviewed RO. Accuracy of margin assessment was however insufficient according to more than 40.0% of RO. Additional margins/biopsies of the operative bed were available in 55.2% (16/29) of the centres. In the absence of additional margins, quality of resection after EES was considered as microscopically incomplete in 48.3% or dubious in 48.3% of RO. As performed, histosurgical mapping allowed radiotherapy dose and volumes adaptation according to 26.3% of RO only.

CONCLUSIONS

Standardized histosurgical mapping with margin and additional margin analysis could be more systematic. Advantages of accurate EES reporting could be dose painting radiotherapy to further decrease morbidity in sinonasal tumours.

Primary cutaneous anaplastic large-cell lymphoma: a review of the SEER database from 2005 to 2016

INTRODUCTION

Primary cutaneous anaplastic large-cell lymphoma (PC-ALCL) is a rare T-cell lymphoma. A prior analysis of the Surveillance, Epidemiology, and End Results (SEER) database reported only 157 cases of localized primary cutaneous CD30+ T-cell lymphoproliferative disorders (PC-ALCL and lymphomatoid papulosis) from 1973 to 2004. Our analysis of the SEER database since 2004 is the largest to date and our results improve our understanding of this disease and their potential prognostic factors.

METHODS

We used the SEER database to retrospectively identify patients. Survival was analysed using the Kaplan-Meier method, and log-rank tests were used to compare survival distributions.

RESULTS

There were 501 cases of PC-ALCL recorded from 2005 to 2016. Overall survival rates at 5 and 10 years were found to be 80.6% (95% CI 76.3%-84.3%) and 61.5% (95% CI 54.1%-68.1%) respectively. Age ≥ 60 years [hazard ratio (HR) = 1.09, P = 0.001 and use of chemotherapy (HR = 1.86, P = 0.01)] were associated with lower overall survival. In contrast to the 1973-2004 cohort, the head and neck site was not significantly associated with prognosis on multivariate analysis.

CONCLUSION

PC-ALCL has been increasingly recognized over the past decade. Age > 60 years and use of chemotherapy are associated with a worse outcome. Contrary to prior studies, location was not associated with poor survival.

622 Potential Application of Electrocardiographs (ECG) to Diagnose Breech Presentations of Fetuses

Breech presentation, i.e. the abnormal lie of a foetus longitudinally with the buttocks closest to the mother’s cervix, is a common issue affecting up to 5% of women during delivery resulting in poor outcomes such as traumatic labour, infection and increased neonatal mortality.

Currently, abdominal palpation is the screening method of choice for breech presentation; however, 15,000 breech presentations remain undiagnosed across England annually. While routine ultrasonography has been suggested, its cost-effectiveness and scalability remain concerning, especially in low/middle-income countries.

Various algorithms have been applied to maternal trans-abdominal ECGs to obtain foetal ECGs, but current applications only allow heart rates, congenital heart defects, hypoxia, and foetal distress to be identified. The use of ECG is proposed to determine foetal position as breech foetuses tend to display QRS complexes in a similar pattern to the maternal trace, whereas cephalic foetuses would demonstrate inverted QRS complexes.

ECGs are cost-effective, safe and do not require highly skilled technicians to operate, which makes it an ideal starting platform for the development of a device with high sensitivity and reasonable diagnostic speeds to identify breech foetuses. Future considerations include the use of artificial intelligence to increase diagnostic accuracy and development of a patient-facing mobile application.

A20 DIAGNOSTIC ACCURACY OF A LOW COST, WIDELY AVAILABLE TEST STRIP FOR PREDICTING A POSITIVE FECAL CALPROTECTIN TEST

Background

Fecal calprotectin is a non-invasive test of colonic inflammation used for monitoring inflammatory bowel disease activity and for risk stratifying non-specific colonic symptoms. Calprotectin is a leukocyte specific enzyme. A similar test, leukocyte esterase is used to detect leukocytes in urine and is widely available as a low-cost point-of-care test strip. We hypothesize that an unmodified version of the urine test strip would be highly accurate in predicting a positive fecal calprotectin test in a real world sample of patients.

Aims

To explore a low cost, rapid alternative to the fecal calprotectin test

Methods

All inpatient and outpatient stool samples tested for calprotectin by the Vancouver General Hospital laboratory from February 2020 to November 2020 were included prospectively. Samples were simultaneously tested for fecal leukocyte esterase using an unmodified Roche Cobas Chemstrip urinalysis test strip by central lab personnel. An identical aliquot was sent to LifeLabs for calprotectin as per standard protocol. All samples were suspended in buffer using established laboratory protocols prior to testing. Fecal leukocyte esterase results were reported as 0–4+ based on visual interpretation, calprotectin results were reported as mcg/g of stool. REB review and approval was obtained prior to data collection. Sensitivity, Specificity and AUROC were calculated using Microsoft Excel and JROCFIT.

Results

26 samples were collected. Using a fecal calprotectin greater than 120 mcg/g as a gold standard an AUROC of 0.89 (SE= .06) was calculated. A leukocyte esterase reading of 2+ or greater had the best test characteristics based on ROC curve analysis. Using this cutoff, 21/26 samples were concordant, giving an accuracy of 80.8%, sensitivity of 90.9% and specificity of 73.3%. Positive likelihood ratio was 8.07 and negative likelihood ratio was 0.29. Assuming an AUROC of 0.8, the sample size N=26 is 90% powered (β=0.9) to predict the true AUROC within 0.1 with a type I error rate of .05 (α<.05).

Conclusions

This study suggests application of a prepared stool sample to a urinalysis test strip gives a result highly predictive of a positive fecal calprotectin test. Further results are being collected prospectively to improve the robustness of these preliminary data. Secondary outcomes including comparison to endoscopy and biopsy results where available are planned if an adequate sample size can be accrued. Future studies justifying independent clinical use of leukocyte esterase would require a common gold standard comparator such as endoscopy. Fecal calprotectin testing is not universally insured and is not available as a rapid test strip. Use of fecal leukocyte esterase may reduce costs and shorten time to results if proven to be independently reliable.

Funding Agencies

None

Neighborhood Social Cohesion Associates with Loneliness Differently among Older People According to Subjective Social Status

OBJECTIVES

To examine whether neighborhood social cohesion can alleviate the negative impact of low subjective social status on feelings of loneliness.

DESIGN

Cross-sectional study.

SETTING

Community, Hong Kong.

PARTICIPANTS

Older people who participated in a cohort study on osteoporosis and general health in Hong Kong (MrOs study).

METHODS

Data were sourced from the 14-year follow-up data of the MrOs study. Loneliness was measured using the 6-item De Jong Gierveld Loneliness Scale. Neighborhood social cohesion was measured by the Hong Kong version of Neighborhood Cohesion Instrument. Linear regression models were used to examine the associations between neighborhood social cohesion and loneliness, controlled for age, sex, marital status, educational level, lifestyle, number of diseases, and maximum lifetime income. The analyses were stratified by subjective social status as measured by a 10-rung self-anchoring scale.

RESULTS

1,037 participants with a mean age of 83 years were included in the study, of whom 72%, 83%, and 64% were classified as at risk of overall loneliness, emotional loneliness, and social loneliness, respectively. Those who were classified as at risk of overall loneliness reported lower subjective social status and had lower levels of neighborhood social cohesion. Linear regression models showed that higher levels of neighborhood social cohesion were associated with lower levels of overall and social loneliness. Stratified analyses showed that the associations between neighborhood social cohesion and loneliness vary across subjective social status groups. Among those with low/middle social status ranking, higher levels of neighborhood social cohesion were associated with lower overall (low-ranking B=-0.111, p=0.001; middle-ranking B=-0.057, p=0.026) and social (low-ranking B=-0.093, p<0.001; middle-ranking B=-0.073, p<0.001) loneliness scores. Among those with high ranking, higher levels of neighborhood social cohesion were associated with lower overall (B=-0.099, p=0.041) and emotional (B=-0.056, p=0.017) loneliness scores, but the associations became insignificant when controlling for maximum lifetime income.

CONCLUSIONS AND IMPLICATIONS

Neighborhood social cohesion may operate differently in different social ranking groups. Interventions to alleviate feelings of loneliness should be subjective social status specific.