Development and validation of a deep learning system for detection of small bowel pathologies in capsule endoscopy: a pilot study in a Singapore institution.

INTRODUCTION: Deep learning models can assess the quality of images and discriminate among abnormalities in small bowel capsule endoscopy (CE), reducing fatigue and the time needed for diagnosis. They serve as a decision support system, partially automating the diagnosis process by providing probability predictions for abnormalities. METHODS: We demonstrated the use of deep learning models in CE image analysis, specifically by piloting a bowel preparation model (BPM) and an abnormality detection model (ADM) to determine frame-level view quality and the presence of abnormal findings, respectively. We used convolutional neural network-based models pretrained on large-scale open-domain data to extract spatial features of CE images that were then used in a dense feed-forward neural network classifier. We then combined the open-source Kvasir-Capsule dataset (n = 43) and locally collected CE data (n = 29). RESULTS: Model performance was compared using averaged five-fold and two-fold cross-validation for BPMs and ADMs, respectively. The best BPM model based on a pre-trained ResNet50 architecture had an area under the receiver operating characteristic and precision-recall curves of 0.969±0.008 and 0.843±0.041, respectively. The best ADM model, also based on ResNet50, had top-1 and top-2 accuracies of 84.03±0.051 and 94.78±0.028, respectively. The models could process approximately 200-250 images per second and showed good discrimination on time-critical abnormalities such as bleeding. CONCLUSION: Our pilot models showed the potential to improve time to diagnosis in CE workflows. To our knowledge, our approach is unique to the Singapore context. The value of our work can be further evaluated in a pragmatic manner that is sensitive to existing clinician workflow and resource constraints.

Colonic stenting in acute malignant large bowel obstruction: audit of efficacy and safety in a Singapore tertiary referral centre.

Introduction: Acute malignant large bowel obstruction (MBO) occurs in 8%-15% of colorectal cancer patients. Self-expandable metal stents (SEMS) have progressed from a palliative modality to use as bridge to surgery (BTS). We aimed to assess the safety and efficacy of SEMS for MBO in our institution. Methods: The data of patients undergoing SEMS insertion for MBO were reviewed. Technical success was defined as successful SEMS deployment across tumour without complications. Clinical success was defined as colonic decompression without requiring further surgical intervention. Rates of complications, median time to surgery, types of surgery and rates of recurrence were studied. Results: Seventy-nine patients underwent emergent SEMS placement from September 2013 to February 2020. Their mean age was 68.8 ± 13.8 years and 43 (54%) patients were male. Mean tumour length was 4.2 cm ± 2.2 cm; 89.9% of malignant strictures were located distal to the splenic flexure. Technical and clinical success was 94.9% and 98.7%, respectively. Perforation occurred in 5.1% of patients, with none having stent migration or bleeding. Fifty (63.3%) patients underwent SEMS insertion as BTS. Median time to surgery was 20 (range 6-57) days. Most (82%) patients underwent minimally invasive surgery. Primary anastomosis rate was 98%. Thirty-nine patients had follow-up beyond 1-year posttreatment (median 34 months). Local recurrence and distant metastasis were observed in 4 (10.3%) and 5 (12.8%) patients, respectively. Conclusion: Insertion of SEMS for acute MBO has high success rates and a good safety profile. Most patients in this audit underwent minimally invasive surgery and primary anastomosis after successful BTS.

Resected pancreatic adenocarcinoma: An Asian institution's experience.

BACKGROUND: Pancreatic adenocarcinoma (PDAC) is highly lethal. Surgery offers the only chance of cure, but 5-year overall survival (OS) after surgical resection and adjuvant therapy remains dismal. Adjuvant trials were mostly conducted in the West enrolling fit patients. Applicability to a general population, especially Asia has not been described adequately. AIM: We aimed to evaluate the clinical outcomes, prognostic factors of survival, pattern, and timing of recurrence after curative resection in an Asian institution. METHODS AND RESULTS: The clinicopathologic and survival outcomes of 165 PDAC patients who underwent curative resection between 1998 and 2013 were reviewed retrospectively. Median age at surgery was 62.0 years. 55.2% were male, and 73.3% had tumors involving the head of pancreas. The median OS of the entire cohort was 19.7 months. Median OS of patients who received adjuvant chemotherapy was 23.8 months. Negative predictors of survival include lymph node ratio (LNR) of >0.3 (HR = 3.36, P = .001), tumor site involving the body or tail of pancreas (HR = 1.59, P = .046), presence of perineural invasion (PNI) (HR = 2.36, P = .018) and poorly differentiated/undifferentiated tumor grade (HR = 1.86, P = .058). The median time to recurrence was 8.87 months, with 66.1% and 81.2% of patients developing recurrence at 12 months and 24 months respectively. The most common site of recurrence was the liver. CONCLUSION: The survival of Asian patients with resected PDAC who received adjuvant chemotherapy is comparable to reported randomized trials. Clinical characteristics seem similar to Western patients. Hence, geographical locations may not be a necessary stratification factor in RCTs. Conversely, lymph node ratio and status of PNI ought to be incorporated.