Miss rate of colorectal neoplastic polyps and risk factors for missed polyps in consecutive colonoscopies

Optimizing colorectal polyp detection and localization: Impact of RGB color adjustment on CNN performance

Colorectal cancer, arising from adenomatous polyps, is a leading cause of cancer-related mortality, making early detection and removal crucial for preventing cancer progression. Machine learning is increasingly used to enhance polyp detection during colonoscopy, the gold standard for colorectal cancer screening, despite its operator-dependent miss rates. This study explores the impact of RGB color adjustment on Convolutional Neural Network (CNN) models for improving polyp detection and localization in colonoscopic images. Using datasets from Harvard Dataverse for training and internal validation, and LDPolypVideo-Benchmark for external validation, RGB color adjustments were applied, and YOLOv8s was used to develop models. Bayesian optimization identified the best RGB adjustments, with performance assessed using mean average precision (mAP) and F1-scores. Results showed that RGB adjustment with 1.0 R-1.0 G-0.8 B improved polyp detection, achieving an mAP of 0.777 and an F1-score of 0.720 on internal test sets, and localization performance with an F1-score of 0.883 on adjusted images. External validation showed improvement but with a lower F1-score of 0.556. While RGB adjustments improved performance in our study, their generalizability to diverse datasets and clinical settings has yet to be validated. Thus, although RGB color adjustment enhances CNN model performance for detecting and localizing colorectal polyps, further research is needed to verify these improvements across diverse datasets and clinical settings.•RGB Color Adjustment: Applied RGB color adjustments to colonoscopic images to enhance the performance of Convolutional Neural Network (CNN) models.•Model Development: Used YOLOv8s for polyp detection and localization, with Bayesian optimization to identify the best RGB adjustments.•Performance Evaluation: Assessed model performance using mAP and F1-scores on both internal and external validation datasets.

Dynamic spectrum-driven hierarchical learning network for polyp segmentation

Accurate automatic polyp segmentation in colonoscopy is crucial for the prompt prevention of colorectal cancer. However, the heterogeneous nature of polyps and differences in lighting and visibility conditions present significant challenges in achieving reliable and consistent segmentation across different cases. Therefore, this study proposes a novel dynamic spectrum-driven hierarchical learning model (DSHNet), the first to specifically leverage image frequency domain information to explore region-level salience differences among and within polyps for precise segmentation. A novel spectral decoupler is advanced to separate low-frequency and high-frequency components, leveraging their distinct characteristics to guide the model in learning valuable frequency features without bias through automatic masking. The low-frequency driven region-level saliency modeling then generates dynamic convolution kernels with individual frequency-aware features, which regulate region-level saliency modeling together with the supervision of the hierarchy of labels, thus enabling adaptation to polyp heterogeneous and illumination variation simultaneously. Meanwhile, the high-frequency attention module is designed to preserve the detailed information at the skip connections, which complements the focus on spatial features at various stages. Experimental results demonstrate that the proposed method outperforms other state-of-the-art polyp segmentation techniques, achieving robust and superior results on five diverse datasets. Codes are available at https://github.com/gardnerzhou/DSHNet.

One in three adenomas could be missed by white-light colonoscopy - findings from a systematic review and meta-analysis

BACKGROUND

White light (conventional) colonoscopy (WLC) is widely used for colorectal cancer screening, diagnosis and surveillance but endoscopists may fail to detect adenomas. Our goal was to assess and synthesize overall and subgroup-specific adenoma miss rates (AMR) of WLC in daily practice.

METHODS

We conducted a systematic review in MEDLINE, EMBASE, Cochrane Library, and grey literature on studies evaluating diagnostic WLC accuracy in tandem studies with novel-colonoscopic technologies (NCT) in subjects undergoing screening, diagnostic or surveillance colonoscopy. Information on study design, AMR overall and specific for adenoma size, histology, location, morphology and further outcomes were extracted and reported in standardized evidence tables. Study quality was assessed using the QUADAS-2 tool. Random-effects meta-analyses and meta-regression were performed to estimate pooled estimates for AMR with 95% confidence intervals (95% CI) and to explain heterogeneity.

RESULTS

Out of 5,963 identified studies, we included sixteen studies with 4,101 individuals in our meta-analysis. One in three adenomas (34%; 95% CI: 30-38%) was missed by WLC in daily practice individuals. Subgroup analyses showed significant AMR differences by size (36%, adenomas 1-5 mm; 27%, adenomas 6-9 mm; 12%, adenomas ≥ 10 mm), histology (non-advanced: 42%, advanced: 21%), morphology (flat: 50%, polypoid: 27%), but not by location (distal: 36%, proximal: 36%).

CONCLUSIONS

Based on our meta-analysis, one in three adenomas could be missed by WLC. This may significantly contribute to interval cancers. Our results should be considered in health technology assessment when interpreting sensitivity of fecal occult blood or other screening tests derived from studies using WLC as "gold standard".

UM-Net: Rethinking ICGNet for polyp segmentation with uncertainty modeling

Automatic segmentation of polyps from colonoscopy images plays a critical role in the early diagnosis and treatment of colorectal cancer. Nevertheless, some bottlenecks still exist. In our previous work, we mainly focused on polyps with intra-class inconsistency and low contrast, using ICGNet to solve them. Due to the different equipment, specific locations and properties of polyps, the color distribution of the collected images is inconsistent. ICGNet was designed primarily with reverse-contour guide information and local-global context information, ignoring this inconsistent color distribution, which leads to overfitting problems and makes it difficult to focus only on beneficial image content. In addition, a trustworthy segmentation model should not only produce high-precision results but also provide a measure of uncertainty to accompany its predictions so that physicians can make informed decisions. However, ICGNet only gives the segmentation result and lacks the uncertainty measure. To cope with these novel bottlenecks, we further extend the original ICGNet to a comprehensive and effective network (UM-Net) with two main contributions that have been proved by experiments to have substantial practical value. Firstly, we employ a color transfer operation to weaken the relationship between color and polyps, making the model more concerned with the shape of the polyps. Secondly, we provide the uncertainty to represent the reliability of the segmentation results and use variance to rectify uncertainty. Our improved method is evaluated on five polyp datasets, which shows competitive results compared to other advanced methods in both learning ability and generalization capability. The source code is available at https://github.com/dxqllp/UM-Net.

Effectiveness of artificial intelligence assisted colonoscopy on adenoma and polyp miss rate: A meta-analysis of tandem RCTs

BACKGROUND AND AIMS

One-fourth of colorectal neoplasia is missed at screening colonoscopy, representing the leading cause of interval colorectal cancer (I-CRC). This systematic review and meta-analysis summarizes the efficacy of computer-aided colonoscopy (CAC) compared to white-light colonoscopy (WLC) in reducing lesion miss rates.

METHODS

Major databases were systematically searched through May 2024 for tandem-design RCTs comparing lesion miss rates in CAC-first followed by WLC vs WLC-first followed by CAC. The primary outcomes were adenoma miss rate (AMR) and polyp miss rate (PMR). The secondary outcomes were advanced AMR (aAMR) and sessile serrated lesion miss rate (SMR).

RESULTS

Six RCTs (1718 patients) were included. AMR was significantly lower for CAC compared to WLC (RR = 0.46; 95 %CI [0.38-0.55]; P < 0.001). PMR was also lower for CAC compared to WLC (RR = 0.44; 95 %CI [0.33-0.60]; P < 0.001). No significant difference in aAMR (RR = 1.28; 95 %CI [0.34-4.83]; P = 0.71) and SMR (RR = 0.44; 95 %CI [0.15-1.28]; P = 0.13) were observed. Sensitivity analysis including only RCTs performed in CRC screening and surveillance setting confirmed lower AMR (RR = 0.48; 95 %CI [0.39-0.58]; P < 0.001) and PMR (RR = 0.50; 95 %CI [0.37-0.66]; P < 0.001), also showing significantly lower SMR (RR = 0.28; 95 %CI [0.11-0.70]; P = 0.007) for CAC compared to WLC.

CONCLUSIONS

CAC results in significantly lower AMR and PMR compared to WLC overall, and significantly lower AMR, PMR and SMR in the screening/surveillance setting, potentially reducing the incidence of I-CRC.

Colonic Polyp Study: Differences in Adenoma Characteristics Based on Colonoscopy History over 5-Year Follow-Up Period

Background: Timely detection and removal of colonic adenomas are critical for preventing colorectal cancer. Methods: This study analyzed differences in colonic adenoma characteristics based on colonoscopy history by reviewing the medical records of 14,029 patients who underwent colonoscopy between January and June 2020 across 40 primary medical institutions in Korea. Results: Adenoma and advanced neoplasia characteristics varied significantly with colonoscopy history (p < 0.05). In the first-time colonoscopy group, adenomas were more frequent in the sigmoid colon (S-colon) and rectum, with Is features and non-granular laterally spreading tumors. Advanced neoplasia was also more common in the S-colon and rectum, with Is and advanced-type features. In the <5-year group, adenomas were predominantly found in the transverse colon (T-colon) and descending colon (D-colon), with IIa and IIb features. Advanced neoplasia in this group was more frequent in the cecum and T-colon, with IIa and IIb features and laterally spreading tumors. In the ≥5-year group, adenomas were more commonly located in the ascending colon (A-colon) and cecum, with Ip features, while advanced neoplasia was more frequent in the A-colon and D-colon, also with Ip features. Conclusions: Although every segment of the colorectum should be carefully observed regardless of colonoscopy history, these findings suggest that prioritizing specific colonic segments for examination based on colonoscopy history may improve adenoma detection rates and reduce the incidence of colorectal cancer. However, further large-scale, prospective studies are needed to confirm these findings and support their application in clinical practice.

Trivariate Joint Modeling for Family Data with Longitudinal Counts, Recurrent Events and a Terminal Event with Application to Lynch Syndrome

Trivariate joint modeling for longitudinal count data, recurrent events, and a terminal event for family data has increased interest in medical studies. For example, families with Lynch syndrome (LS) are at high risk of developing colorectal cancer (CRC), where the number of polyps and the frequency of colonoscopy screening visits are highly associated with the risk of CRC among individuals and families. To assess how screening visits influence polyp detection, which in turn influences time to CRC, we propose a clustered trivariate joint model. The proposed model facilitates longitudinal count data that are zero-inflated and over-dispersed and invokes individual-specific and family-specific random effects to account for dependence among individuals and families. We formulate our proposed model as a latent Gaussian model to use the Bayesian estimation approach with the integrated nested Laplace approximation algorithm and evaluate its performance using simulation studies. Our trivariate joint model is applied to a series of 18 families from Newfoundland, with the occurrence of CRC taken as the terminal event, the colonoscopy screening visits as recurrent events, and the number of polyps detected at each visit as zero-inflated count data with overdispersion. We showed that our trivariate model fits better than alternative bivariate models and that the cluster effects should not be ignored when analyzing family data. Finally, the proposed model enables us to quantify heterogeneity across families and individuals in polyp detection and CRC risk, thus helping to identify individuals and families who would benefit from more intensive screening visits.

Fluorescence-guided tumor visualization of colorectal cancer using tumor-initiating probe yellow in preclinical models

Fluorescence-guided surgery has emerged as an innovative technique with promising applications in the treatment of various tumors, including colon cancer. Tumor-initiating probe yellow (TiY) has been discovered for identifying tumorigenic cells by unbiased phenotypic screening with thousands of diversity-oriented fluorescence library (DOFL) compounds in a patient-derived lung cancer cell model. This study demonstrated the clinical feasibility of TiY for tumor-specific fluorescence imaging in the tissues of patients with colorectal cancer (CRC). To evaluate the efficacy of TiY in tumor imaging, surgical specimens were obtained, consisting of 36 tissues from 18 patients with CRC, for ex vivo molecular fluorescence imaging, histology, and immunohistochemistry. Orthotopic and chemically induced CRC mice models were administered TiY topically, and distinct tumor lesions were observed in 10 min by real-time fluorescence colonoscopy and ex vivo imaging. In a hepatic metastasis mouse model using splenic injection, TiY accumulation was detected in metastatic liver lesions through fluorescence imaging. Correlation analysis between TiY intensity and protein expression, assessed via immunohistochemistry and Western blotting, revealed a positive correlation between TiY and vimentin and Zeb1, which are known as epithelial-mesenchymal transition (EMT) markers of cancers. A comparative analysis of TiY with other FDA-approved fluorescence probes such as ICG revealed greater quantitative differences in TiY fluorescence intensity between tumor and normal tissues than those observed with ICG. Altogether, these results demonstrated that TiY has a strong potential for visualizing CRC by fluorescence imaging in various preclinical models, which can be further translated for clinical use such as fluorescence-guided surgery. Furthermore, our data indicate that TiY is preferentially uptaken by cells with EMT induction and progression, and overexpressing vimentin and Zeb1 in patients with CRC.

Efficacy of colonoscopic re-examination across the entire colon: a randomized controlled trial

BACKGROUND

In standard colonoscopic examinations, some polyps may be missed during the withdrawal phase. Re-examination of the right colon can improve the adenoma detection rate (ADR). However, the effectiveness of applying this re-examination strategy to the entire colon remains unknown. We investigated whether re-examination could increase the detection rate of polyps and adenomas throughout the colon.

METHODS

A randomized, controlled, single-center trial (NCT03268200) was conducted in a university hospital. Patients aged 45-75 years were randomly assigned to either the study or control group. For patients in the control group, observation and polypectomy were performed once using the standard colonoscopy method. For patients in the study group, polypectomy was repeated twice during the withdrawal phase after the initial insertion of the colonoscope. These examinations were performed in the right transverse and left colons. The primary endpoints were the polyp detection rate (PDR) and ADR, defined as the proportion of patients with ≥ 1 polyp and ≥ 1 adenoma, respectively.

RESULTS

Overall, 406 patients were enrolled (study group, n = 210; control group, n = 196) and analyzed. Generally, PDRs and ADRs were similar between the study (withdrawal 1 + 2) and control groups (withdrawal 1), except for the right colon. However, the second withdrawal review increased number of polyps and adenomas in the overall, right, mid, and left colon, respectively.

CONCLUSION

The results of this trial indicated that re-examination of the entire colon during colonoscopy could be beneficial for detecting concealed polyps in patients at risk of interval cancer.

The Utility of Multitarget Stool DNA Testing for Colorectal Cancer Screening After a Normal Colonoscopy

BACKGROUND

Multitarget stool DNA (MT-sDNA) tests (here, Cologuard®) are currently used in average-risk patients as a primary method of screening for colorectal cancer. However, MT-sDNA testing has also been used in patients who previously underwent colonoscopy who wish to avoid repeat colonoscopy. Here, in a large primary care practice setting, our aim was to evaluate the diagnostic performance of MT-sDNA testing in patients with a previously normal colonoscopy.

METHODS

This retrospective cohort study included 5827 patients from 35 different primary locations in South Carolina. Patients aged 45 and above with a previously documented normal, high-quality colonoscopy prior to the MT-sDNA test date were included. High-risk patients and those with a previous negative MT-sDNA result were excluded.

RESULTS

Of 5827 ordered MT-sDNA tests, 248 patients had a prior normal colonoscopy. The average time from initial colonoscopy to MT-sDNA testing was 7.3 years. Of the 63 patients who had a positive MT-sDNA test, 41 patients (65%) completed follow-up colonoscopy and 40 patients had complete colonoscopy data. Of these 40 patients, 12 patients (30%) had advanced adenomas and none had colorectal cancer. Compared to patients without a previous colonoscopy, patients with prior colonoscopies had fewer adenomas of all types (1.6 vs 2.4) and fewer advanced adenomas (1.4 vs 2.0).

CONCLUSION

Patients with a previously negative colonoscopy and subsequent positive MT-sDNA test were found to have a high rate of advanced adenomas on follow-up colonoscopy (30%). Thus, in patients with a previously negative colonoscopy, MT-sDNA testing may be a reasonable alternative screening option.

A lighter hybrid feature fusion framework for polyp segmentation

Colonoscopy is widely recognized as the most effective method for the detection of colon polyps, which is crucial for early screening of colorectal cancer. Polyp identification and segmentation in colonoscopy images require specialized medical knowledge and are often labor-intensive and expensive. Deep learning provides an intelligent and efficient approach for polyp segmentation. However, the variability in polyp size and the heterogeneity of polyp boundaries and interiors pose challenges for accurate segmentation. Currently, Transformer-based methods have become a mainstream trend for polyp segmentation. However, these methods tend to overlook local details due to the inherent characteristics of Transformer, leading to inferior results. Moreover, the computational burden brought by self-attention mechanisms hinders the practical application of these models. To address these issues, we propose a novel CNN-Transformer hybrid model for polyp segmentation (CTHP). CTHP combines the strengths of CNN, which excels at modeling local information, and Transformer, which excels at modeling global semantics, to enhance segmentation accuracy. We transform the self-attention computation over the entire feature map into the width and height directions, significantly improving computational efficiency. Additionally, we design a new information propagation module and introduce additional positional bias coefficients during the attention computation process, which reduces the dispersal of information introduced by deep and mixed feature fusion in the Transformer. Extensive experimental results demonstrate that our proposed model achieves state-of-the-art performance on multiple benchmark datasets for polyp segmentation. Furthermore, cross-domain generalization experiments show that our model exhibits excellent generalization performance.

Significant Missed Polyps in the UK Bowel Cancer Screening Programme (BCSP): A Retrospective Analysis of Prevalence and Contributing Factors

Background Colorectal cancer (CRC) remains a significant public health challenge. Patients having abnormal faecal immunochemical test (FIT) results are offered a colonoscopy. The effectiveness of colonoscopies can, however, often be challenged by the occurrence of missed polyps. This study aims to assess the rate of significant missed polyps in the Bowel Cancer Screening Programme (BCSP) in the UK. Methods A retrospective analysis of BSCP screening data in the Cheshire region in the UK from 2020 to March 2023 was conducted. A significant polyp was defined as a polyp ≥ 10mm. The inclusion criteria included patients (age range: 54-74 years) who had had an index colonoscopy followed by site checks, repeats, or planned polypectomies. Results Out of 2,759 index colonoscopies, 261 (9.5%) met our criteria, and 23 (8.8%) of these had significant polyps. Of the 261, the missed polyp rate was 30% (453/1531 polyps). The overall significant missed polyp rate was 1.6% (24/1531). Of the missed polyps, 5% (24/453) were significant polyps. The majority (71%) of the significant polyps were found on the left of the colon. Men had a higher missed polyp rate (22%) compared to women (7%) (relative risk (RR) = 2.56, 95% CI: 2.1-3.13, p<0.0001). They also had a higher significant missed polyp rate (1.1%) compared to women (0.4%) (RR = 2.41, 95% CI: 1-5.8, p<0.05). A total of 50% of the bowel prep at index colonoscopy was rated as 'adequate/fair' and 79% of the bowel prep at the discovery of the significant polyp was rated as either 'excellent' or 'good' (odds ratio (OR) = 3.8, 95% CI: 1.07-13.5, p<0.05); 92% (22/24) of the significant polyps found were either tubular adenoma (TA) low-grade dysplasia (LGD) or tubular villous adenoma (TVA) LGD, and none were found to be cancerous. Conclusions Almost a third of all polyps detected were missed, and one in 20 of these were significant polyps, putting these patients in the high-risk group for CRC. Improving bowel preparation and monitoring patients with multiple polyps could reduce the rate of missed polyps.

A complete benchmark for polyp detection, segmentation and classification in colonoscopy images

Introduction

Colorectal cancer (CRC) is one of the main causes of deaths worldwide. Early detection and diagnosis of its precursor lesion, the polyp, is key to reduce its mortality and to improve procedure efficiency. During the last two decades, several computational methods have been proposed to assist clinicians in detection, segmentation and classification tasks but the lack of a common public validation framework makes it difficult to determine which of them is ready to be deployed in the exploration room.

Methods

This study presents a complete validation framework and we compare several methodologies for each of the polyp characterization tasks.

Results

Results show that the majority of the approaches are able to provide good performance for the detection and segmentation task, but that there is room for improvement regarding polyp classification.

Discussion

While studied show promising results in the assistance of polyp detection and segmentation tasks, further research should be done in classification task to obtain reliable results to assist the clinicians during the procedure. The presented framework provides a standarized method for evaluating and comparing different approaches, which could facilitate the identification of clinically prepared assisting methods.

Trends in colorectal cancer surgical resection rates during the screening era: a retrospective study in Italy

BACKGROUND

Faecal immunochemical test (FIT)-based screening is effective in reducing colorectal cancer (CRC) incidence, but its sensitivity for proximal lesions remains low.

OBJECTIVES

We compared age-adjusted CRC surgical resection rates across anatomic sites (proximal colon, distal colon, rectum), age groups and sex over 20 years in a large Italian population. We particularly focused on changes in trends following FIT-screening implementation in the target population (50-69 years).

DESIGN

This retrospective study analysed data from the Veneto Region's administrative Hospital Discharge Dataset, involving over 54 000 patients aged 40-89 (43.4% female) who underwent CRC surgery between 2002 and 2021.

RESULTS

Overall, surgery rates increased until 2007 (annual percentage changes: 2.5% in males, 2.9% in females) and then declined (-4.2% in males, -3.4% in females). This decline was steeper for distal and rectal cancers compared with proximal cancer, suggesting a shift towards more right-sided CRC surgery.In males, the prescreening increase in proximal surgery was reversed after screening implementation (slope change: -6%) while the prescreening decline accelerated for distal (-4%) and rectal (-3%) surgeries. In females, stable prescreening trends shifted downward for all sites (-5% for proximal, -8% for distal and -7% for rectal surgery). However, the change in trends between prescreening and postscreening periods was not different across anatomic sites for either sex (all slope change differences in pairwise comparisons were not statistically significant).

CONCLUSION

The shift towards proximal surgery may not be entirely due to the FIT's low sensitivity but may reflect an underlying upward trend in proximal cancers independent of screening.

The Interplay among Wnt/β-catenin Family Members in Colorectal Adenomas and Surrounding Tissues

BACKGROUND

The colorectal adenoma undergoes neoplastic progression via the normal epithelium-adenoma-adenocarcinoma sequence as reported in the Vogelgram. The hazard of developing a tumor is deeply associated with the number and size of adenomas and their subtype. Adenomatous polyps are histologically categorized as follows: approximately 80-90% are tubular, 5-15% are villous, and 5-10% are tubular/villous. Given the higher risk of a malignant transformation observed in tubular/villous adenomas, patients diagnosed with adenomatous polyposis are at an improved risk of developing CRC. The Wnt/β-catenin pathway plays a key role in the onset of colorectal adenoma; in particular, intestinal cells first acquire loss-of-function mutations in the APC gene that induce the formation of adenomas.

METHODS

Wnt/β-catenin pathway APC, Wnt3a, Wnt5a, LEF1, and BCL9 genes and protein expression analyses were conducted by qRT-PCR and western blot in 68 colonic samples (polyps and adjacent mucosa) from 41 patients, of which 17 were affected by FAP. Ten normal colonic mucosal samples were collected from 10 healthy donors.

RESULTS

In this study, both the APC gene and protein were less expressed in the colon tumor compared to the adjacent colonic mucosa. Conversely, the activated β-catenin was more expressed in polyps than in the adjacent mucosa. All results confirmed the literature data on carcinomas. A statistically significant correlation between Wnt3a and BCL9 both in polyps and in the adjacent mucosa underlines that the canonical Wnt pathway is activated in early colon carcinogenesis and that the adjacent mucosa is already altered.

CONCLUSION

This is the first study analyzing the difference in expression of the Wnt/β-catenin pathway in human colorectal adenomas. Understanding the progression from adenomas to colorectal carcinomas is essential for the development of new therapeutic strategies and improving clinical outcomes with the use of APC and β-catenin as biomarkers.

A Hessian-Based Technique for Specular Reflection Detection and Inpainting in Colonoscopy Images

In the field of Computer-Aided Detection (CADx), the use of AI-based algorithms for disease detection in endoscopy images, especially colonoscopy images, is on the rise. However, these algorithms often encounter performance issues due to obstructions like specular reflection, resulting in false positives. This paper presents a novel algorithm specifically designed to tackle the challenges posed by high specular reflection regions in colonoscopy images. The proposed algorithm identifies these regions and applies precise inpainting for restoration. The process entails converting the input image from RGB to HSV color space and focusing on the Saturation (S) component in convex regions detected using a Hessian-based method. This step creates a binary mask that pinpoints areas of specular reflection. The inpainting function then uses this mask to guide the restoration of these identified regions and their borders. To ensure a seamless blend of the restored regions with the background and adjacent pixels, a feathering process is applied to the repaired regions. This enhances both the accuracy and aesthetic coherence of the inpainted images. The performance of our algorithm was rigorously tested on five unique colonoscopy datasets and various endoscopy images from the Kvasir dataset, using an extensive set of evaluation metrics and a comparative analysis with existing methods consistently highlighted the superior performance of our algorithm.

DHAFormer: Dual-channel hybrid attention network with transformer for polyp segmentation

The accurate early diagnosis of colorectal cancer significantly relies on the precise segmentation of polyps in medical images. Current convolution-based and transformer-based segmentation methods show promise but still struggle with the varied sizes and shapes of polyps and the often low contrast between polyps and their background. This research introduces an innovative approach to confronting the aforementioned challenges by proposing a Dual-Channel Hybrid Attention Network with Transformer (DHAFormer). Our proposed framework features a multi-scale channel fusion module, which excels at recognizing polyps across a spectrum of sizes and shapes. Additionally, the framework's dual-channel hybrid attention mechanism is innovatively conceived to reduce background interference and improve the foreground representation of polyp features by integrating local and global information. The DHAFormer demonstrates significant improvements in the task of polyp segmentation compared to currently established methodologies.

Leveraging a realistic synthetic database to learn Shape-from-Shading for estimating the colon depth in colonoscopy images

Colonoscopy is the choice procedure to diagnose, screening, and treat the colon and rectum cancer, from early detection of small precancerous lesions (polyps), to confirmation of malign masses. However, the high variability of the organ appearance and the complex shape of both the colon wall and structures of interest make this exploration difficult. Learned visuospatial and perceptual abilities mitigate technical limitations in clinical practice by proper estimation of the intestinal depth. This work introduces a novel methodology to estimate colon depth maps in single frames from monocular colonoscopy videos. The generated depth map is inferred from the shading variation of the colon wall with respect to the light source, as learned from a realistic synthetic database. Briefly, a classic convolutional neural network architecture is trained from scratch to estimate the depth map, improving sharp depth estimations in haustral folds and polyps by a custom loss function that minimizes the estimation error in edges and curvatures. The network was trained by a custom synthetic colonoscopy database herein constructed and released, composed of 248400 frames (47 videos), with depth annotations at the level of pixels. This collection comprehends 5 subsets of videos with progressively higher levels of visual complexity. Evaluation of the depth estimation with the synthetic database reached a threshold accuracy of 95.65%, and a mean-RMSE of 0.451cm, while a qualitative assessment with a real database showed consistent depth estimations, visually evaluated by the expert gastroenterologist coauthoring this paper. Finally, the method achieved competitive performance with respect to another state-of-the-art method using a public synthetic database and comparable results in a set of images with other five state-of-the-art methods. Additionally, three-dimensional reconstructions demonstrated useful approximations of the gastrointestinal tract geometry. Code for reproducing the reported results and the dataset are available at https://github.com/Cimalab-unal/ColonDepthEstimation.

Polyp Segmentation Using a Hybrid Vision Transformer and a Hybrid Loss Function

Accurate and early detection of precursor adenomatous polyps and their removal at the early stage can significantly decrease the mortality rate and the occurrence of the disease since most colorectal cancer evolve from adenomatous polyps. However, accurate detection and segmentation of the polyps by doctors are difficult mainly these factors: (i) quality of the screening of the polyps with colonoscopy depends on the imaging quality and the experience of the doctors; (ii) visual inspection by doctors is time-consuming, burdensome, and tiring; (iii) prolonged visual inspections can lead to polyps being missed even when the physician is experienced. To overcome these problems, computer-aided methods have been proposed. However, they have some disadvantages or limitations. Therefore, in this work, a new architecture based on residual transformer layers has been designed and used for polyp segmentation. In the proposed segmentation, both high-level semantic features and low-level spatial features have been utilized. Also, a novel hybrid loss function has been proposed. The loss function designed with focal Tversky loss, binary cross-entropy, and Jaccard index reduces image-wise and pixel-wise differences as well as improves regional consistencies. Experimental works have indicated the effectiveness of the proposed approach in terms of dice similarity (0.9048), recall (0.9041), precision (0.9057), and F2 score (0.8993). Comparisons with the state-of-the-art methods have shown its better performance.

Additional 30-Second Observation of the Right-Sided Colon for Missed Polyp Detection With Texture and Color Enhancement Imaging Compared with Narrow Band Imaging: A Randomized Trial

INTRODUCTION

The efficacy of texture and color enhancement imaging (TXI) in the novel light-emitting diode endoscopic system for polyp detection has not been examined. We aimed to evaluate the noninferiority of the additional 30-second (Add-30-s) observation of the right-sided colon (cecum/ascending colon) with TXI compared with narrow band imaging (NBI) for detecting missed polyps.

METHODS

We enrolled 381 patients ≥40 years old who underwent colonoscopy from September 2021 to June 2022 in 3 institutions and randomly assigned them to either the TXI or NBI groups. The right-sided colon was first observed with white light imaging in both groups. Second, after reinsertion from hepatic flexure to the cecum, the right-sided colon was observed with Add-30-s observation of either TXI or NBI. The primary endpoint was to examine the noninferiority of TXI to NBI using the mean number of adenomas and sessile serrated lesions per patient. The secondary ones were to examine adenoma detection rate, adenoma and sessile serrated lesions detection rates, and polyp detection rates in both groups.

RESULTS

The TXI and NBI groups consisted of 177 and 181 patients, respectively, and the noninferiorities of the mean number of adenomas and sessile serrated lesions per patients in the second observation were significant (TXI 0.29 [51/177] vs NBI 0.30 [54/181], P < 0.01). The change in adenoma detection rate, adenoma and sessile serrated lesions detection rate, and polyp detection rate for the right-sided colon between the TXI and NBI groups were not different (10.2%/10.5% [ P = 0.81], 13.0%/12.7% [ P = 0.71], and 15.3%/13.8% [ P = 0.71]), respectively.

DISCUSSION

Regarding Add-30-s observation of the right-sided colon, TXI was noninferior to NBI.

Future Implications of Artificial Intelligence in Medical Education

Artificial intelligence has experienced explosive growth in the past year that will have implications in all aspects of our lives, including medicine. In order to train a physician workforce that understands these new advancements, medical educators must take steps now to ensure that physicians are adequately trained in medical school, residency, and fellowship programs to become proficient in the usage of artificial intelligence in medical practice. This manuscript discusses the various considerations that leadership within medical training programs should be mindful of when deciding how to best integrate artificial intelligence into their curricula.

Wide-field endoscope accessory for multiplexed fluorescence imaging

A wide-field endoscope that is sensitive to fluorescence can be used as an adjunct to conventional white light endoscopy by detecting multiple molecular targets concurrently. We aim to demonstrate a flexible fiber-coupled accessory that can pass forward through the instrument channel of standard medical endoscopes for clinical use to collect fluorescence images. A miniature scan mirror with reflector dimensions of 1.30 × 0.45  mm2 was designed, fabricated, and placed distal to collimated excitation beams at λex = 488, 660, and 785 nm. The mirror was driven at resonance for wide angular deflections in the X and Y-axes. A large image field-of-view (FOV) was generated in real time. The optomechanical components were packaged in a rigid distal tip with dimensions of 2.6 mm diameter and 12 mm length. The scan mirror was driven at 27.6 and 9.04 kHz in the fast (X) and slow (Y) axes, respectively, using a square wave with 50% duty cycle at 60 Vpp to collect fluorescence images at 10 frames per sec. Maximum total divergence angles of ± 27.4° and ± 22.8° were generated to achieve a FOV of 10.4 and 8.4 mm, respectively, at a working distance of 10 mm. Multiplexed fluorescence images were collected in vivo from the rectum of live mice using 3 fluorescently-labeled peptides that bind to unique cell surface targets. The fluorescence images collected were separated into 3 channels. Target-to-background ratios of 2.6, 3.1, and 3.9 were measured. This instrument demonstrates potential for broad clinical use to detect heterogeneous diseases in hollow organs.

Comparative Analysis of Machine Learning Models for Image Detection of Colonic Polyps vs. Resected Polyps

(1) Background: Colon polyps are common protrusions in the colon's lumen, with potential risks of developing colorectal cancer. Early detection and intervention of these polyps are vital for reducing colorectal cancer incidence and mortality rates. This research aims to evaluate and compare the performance of three machine learning image classification models' performance in detecting and classifying colon polyps. (2) Methods: The performance of three machine learning image classification models, Google Teachable Machine (GTM), Roboflow3 (RF3), and You Only Look Once version 8 (YOLOv8n), in the detection and classification of colon polyps was evaluated using the testing split for each model. The external validity of the test was analyzed using 90 images that were not used to test, train, or validate the model. The study used a dataset of colonoscopy images of normal colon, polyps, and resected polyps. The study assessed the models' ability to correctly classify the images into their respective classes using precision, recall, and F1 score generated from confusion matrix analysis and performance graphs. (3) Results: All three models successfully distinguished between normal colon, polyps, and resected polyps in colonoscopy images. GTM achieved the highest accuracies: 0.99, with consistent precision, recall, and F1 scores of 1.00 for the 'normal' class, 0.97-1.00 for 'polyps', and 0.97-1.00 for 'resected polyps'. While GTM exclusively classified images into these three categories, both YOLOv8n and RF3 were able to detect and specify the location of normal colonic tissue, polyps, and resected polyps, with YOLOv8n and RF3 achieving overall accuracies of 0.84 and 0.87, respectively. (4) Conclusions: Machine learning, particularly models like GTM, shows promising results in ensuring comprehensive detection of polyps during colonoscopies.

Near-Infrared Imaging of Colonic Adenomas In Vivo Using Orthotopic Human Organoids for Early Cancer Detection

Colorectal cancer is a leading cause of cancer-related morbidity and mortality worldwide. Premalignant lesions that are flat and subtle in morphology are often missed in conventional colonoscopies. Patient-derived adenoma colonoids with high and low cMet expression and normal colonoids were implanted orthotopically in the colon of immunocompromised mice to serve as a preclinical model system. A peptide specific for cMet was labeled with IRDye800, a near-infrared (NIR) fluorophore. This peptide was administered intravenously, and in vivo imaging was performed using a small animal fluorescence endoscope. Quantified intensities showed a peak target-to-background ratio at ~1 h after intravenous peptide injection, and the signal cleared by ~24 h. The peptide was stable in serum with a half-life of 3.6 h. Co-staining of adenoma and normal colonoids showed a high correlation between peptide and anti-cMet antibody. A human-specific cytokeratin stain verified the presence of human tissues implanted among surrounding normal mouse colonic mucosa. Peptide biodistribution was consistent with rapid renal clearance. No signs of acute toxicity were found on either animal necropsy or serum hematology and chemistries. Human colonoids provide a clinically relevant preclinical model to evaluate the specific uptake of a NIR peptide to detect premalignant colonic lesions in vivo.

Investigation of artificial intelligence integrated fluorescence endoscopy image analysis with indocyanine green for interpretation of precancerous lesions in colon cancer

Indocyanine green (ICG) has been used in clinical practice for more than 40 years and its safety and preferential accumulation in tumors has been reported for various tumor types, including colon cancer. However, reports on clinical assessments of ICG-based molecular endoscopy imaging for precancerous lesions are scarce. We determined visualization ability of ICG fluorescence endoscopy in colitis-associated colon cancer using 30 lesions from an azoxymethane/dextran sulfate sodium (AOM/DSS) mouse model and 16 colon cancer patient tissue-samples. With a total of 60 images (optical, fluorescence) obtained during endoscopy observation of mouse colon cancer, we used deep learning network to predict four classes (Normal, Dysplasia, Adenoma, and Carcinoma) of colorectal cancer development. ICG could detect 100% of carcinoma, 90% of adenoma, and 57% of dysplasia, with little background signal at 30 min after injection via real-time fluorescence endoscopy. Correlation analysis with immunohistochemistry revealed a positive correlation of ICG with inducible nitric oxide synthase (iNOS; r > 0.5). Increased expression of iNOS resulted in increased levels of cellular nitric oxide in cancer cells compared to that in normal cells, which was related to the inhibition of drug efflux via the ABCB1 transporter down-regulation resulting in delayed retention of intracellular ICG. With artificial intelligence training, the accuracy of image classification into four classes using data sets, such as fluorescence, optical, and fluorescence/optical images was assessed. Fluorescence images obtained the highest accuracy (AUC of 0.8125) than optical and fluorescence/optical images (AUC of 0.75 and 0.6667, respectively). These findings highlight the clinical feasibility of ICG as a detector of precancerous lesions in real-time fluorescence endoscopy with artificial intelligence training and suggest that the mechanism of ICG retention in cancer cells is related to intracellular nitric oxide concentration.

Comparison of missed adenomas in deep-sedated and unsedated colonoscopy: A multicenter retrospective study

BACKGROUND

Deep-sedated colonoscopy with propofol is widely used in China. However, its impact on quality metrics remains controversial. We aimed to investigate the effects of deep-sedated colonoscopy on missed adenomas, specifically in each colorectal segment.

METHODS

Data of 3710 individuals from seven hospitals in China who underwent an initial colonoscopy with or without propofol sedation and a second colonoscopy without sedation within six months for surveillance or polypectomy by endoscopist of the same level between October 2020 and September 2021 were retrospectively analyzed.

RESULTS

A total of 1113 missed adenomas in 3710 patients were evaluated. The adenoma miss rate (AMR) was significantly higher in deep-sedated colonoscopy than in unsedated colonoscop [19.14% (578/3020) vs. 16.15% (535/3313), P < 0.05]. The risk of missing adenomas in deep-sedated colonoscopy was 1.229 times higher than in unsedated colonoscopy (OR, 1.229; 95% CI: 1.080-1.399). AMRs of the splenic flexure (26.02% [96/369] vs. 16.04% [47/293], P < 0.05) and descending colon (20.86% [102/489] vs. 13.37% [54/404], P < 0.05) were significantly higher in deep-sedated colonoscopy than in unsedated colonoscopy when performed by middle-level endoscopists rather than high-level endoscopists (P < 0.05).

CONCLUSIONS

AMR was higher in deep-sedated colonoscopy than in unsedated colonoscopy. Furthermore, adenomas in the splenic flexure and descending colon were more frequently missed in deep-sedated colonoscopy than in unsedated colonoscopy, particularly when performed by less experienced endoscopists.

Effect of computer-aided colonoscopy on adenoma miss rates and polyp detection: A systematic review and meta-analysis

BACKGROUND AND AIM

Multiple computer-aided techniques utilizing artificial intelligence (AI) have been created to improve the detection of polyps during colonoscopy and thereby reduce the incidence of colorectal cancer. While adenoma detection rates (ADR) and polyp detection rates (PDR) are important colonoscopy quality indicators, adenoma miss rates (AMR) may better quantify missed lesions, which can ultimately lead to interval colorectal cancer. The purpose of this systematic review and meta-analysis was to determine the efficacy of computer-aided colonoscopy (CAC) with respect to AMR, ADR, and PDR in randomized controlled trials.

METHODS

A comprehensive, systematic literature search was performed across multiple databases in September of 2022 to identify randomized, controlled trials that compared CAC with traditional colonoscopy. Primary outcomes were AMR, ADR, and PDR.

RESULTS

Fourteen studies totaling 10 928 patients were included in the final analysis. There was a 65% reduction in the adenoma miss rate with CAC (OR, 0.35; 95% CI, 0.25-0.49, P < 0.001, I²  = 50%). There was a 78% reduction in the sessile serrated lesion miss rate with CAC (OR, 0.22; 95% CI, 0.08-0.65, P < 0.01, I²  = 0%). There was a 52% increase in ADR in the CAC group compared with the control group (OR, 1.52; 95% CI, 1.39-1.67, P = 0.04, I²  = 47%). There was 93% increase in the number of adenomas > 10 mm detected per colonoscopy with CAC (OR 1.93; 95% CI, 1.18-3.16, P < 0.01, I²  = 0%).

CONCLUSIONS

The results of the present study demonstrate the promise of CAC in improving AMR, ADR, PDR across a spectrum of size and morphological lesion characteristics.

A Real-Time Polyp-Detection System with Clinical Application in Colonoscopy Using Deep Convolutional Neural Networks

Colorectal cancer (CRC) is a leading cause of cancer-related deaths worldwide. The best method to prevent CRC is with a colonoscopy. During this procedure, the gastroenterologist searches for polyps. However, there is a potential risk of polyps being missed by the gastroenterologist. Automated detection of polyps helps to assist the gastroenterologist during a colonoscopy. There are already publications examining the problem of polyp detection in the literature. Nevertheless, most of these systems are only used in the research context and are not implemented for clinical application. Therefore, we introduce the first fully open-source automated polyp-detection system scoring best on current benchmark data and implementing it ready for clinical application. To create the polyp-detection system (ENDOMIND-Advanced), we combined our own collected data from different hospitals and practices in Germany with open-source datasets to create a dataset with over 500,000 annotated images. ENDOMIND-Advanced leverages a post-processing technique based on video detection to work in real-time with a stream of images. It is integrated into a prototype ready for application in clinical interventions. We achieve better performance compared to the best system in the literature and score a F1-score of 90.24% on the open-source CVC-VideoClinicDB benchmark.

A stacking-based artificial intelligence framework for an effective detection and localization of colon polyps

Polyp detection through colonoscopy is a widely used method to prevent colorectal cancer. The automation of this process aided by artificial intelligence allows faster and improved detection of polyps that can be missed during a standard colonoscopy. In this work, we propose to implement various object detection algorithms for polyp detection. To improve the mean average precision (mAP) of the detection, we combine the baseline models through a stacking approach. The experiments demonstrate the potential of this new methodology, which can reduce the workload for oncologists and increase the precision of the localization of polyps. Our proposal achieves a mAP of 0.86, translated into an improvement of 34.9% compared to the best baseline model and 28.8% with respect to the weighted boxes fusion ensemble technique.

Endoscopy Lifetime Systems Architecture: Scoping Out the Past to Diagnose the Future Technology

Systems engineering captures the desires and needs of the customer to conceptualize a system from the overall goal down to the small details prior to any physical development. While many systems projects tend to be large and complicated (i.e., cloud-based infrastructure, long-term space travel shuttles, missile defense systems), systems engineering can also be applied to smaller, complex systems. Here, the system of interest is the endoscope, a standard biomedical screening device used in laparoscopic surgery, screening of upper and lower gastrointestinal tracts, and inspection of the upper airway. Often, endoscopic inspection is used to identify pre-cancerous and cancerous tissues, and hence, a requirement for endoscopic systems is the ability to provide images with high contrast between areas of normal tissue and neoplasia (early-stage abnormal tissue growth). For this manuscript, the endoscope was reviewed for all the technological advancements thus far to theorize what the next version of the system could be in order to provide improved detection capabilities. Endoscopic technology was decomposed into categories, using systems architecture and systems thinking, to visualize the improvements throughout the system's lifetime from the original to current state-of-the-art. Results from this review were used to identify trends in subsystems and components to estimate the theoretical performance maxima for different subsystems as well as areas for further development. The subsystem analysis indicated that future endoscope systems will focus on more complex imaging and higher computational requirements that will provide improved contrast in order to have higher accuracy in optical diagnoses of early, abnormal tissue growth.

Correlation between Surrogate Quality Indicators for Adenoma Detection Rate and Adenoma Miss Rate in Qualified Colonoscopy, CORE Study: KASID Multicenter Study

BACKGROUND/AIMS

The adenoma detection rate (ADR) does not reflect the complete detection of every adenoma during colonoscopy; thus, many surrogate indicators have been suggested. This study investigated whether the ADR and surrogate quality indicators reflect the adenoma miss rate (AMR) when performing qualified colonoscopy.

METHODS

We performed a prospective, multicenter, cross-sectional study of asymptomatic examinees aged 50 to 75 years who underwent back-to-back screening colonoscopies by eight endoscopists. The ADR and surrogate quality indicators, including polyp detection rate, total number of adenomas per colonoscopy, additional adenomas found after the first adenoma per colonoscopy (ADR-Plus), and total number of adenomas per positive participant, were calculated for the prediction of AMR.

RESULTS

A total of 371 back-to-back colonoscopies were performed. There was a significant difference in ADRs (range, 44% to 75.4%; p=0.024), polyp detection rates (range, 56% to 86.9%; p=0.008) and adenomas per positive participants (range, 1.19 to 2.30; p=0.038), and a tendency of a difference in adenomas per colonoscopy (range, 0.62 to 1.31; p=0.051) and ADR-Plus (range, 0.13 to 0.70; p=0.054) among the endoscopists. The overall AMR was 20.1%, and AMRs were not different (range, 13.9 to 28.6; p>0.05) among the endoscopists. No quality indicators were significantly correlated with AMR. The number of adenomas found during the first colonoscopy was an independent factor for increased AMR (odds ratio, 1.79; p<0.001).

CONCLUSIONS

The colonoscopy quality indicators were significantly different among high-ADR endoscopists, and none of the quality indicators reflected the AMR of good quality colonoscopy performances. The only factor influencing AMR was the number of adenomas detected during colonoscopy.

Evaluation of the efficacy of 1 L polyethylene glycol plus ascorbic acid and an oral sodium sulfate solution: A multi-center, prospective randomized controlled trial

INTRODUCTION

Low-volume bowel preparation has been developed to increase patient compliance. We compared 1 L of polyethylene glycol/ascorbic acid (PEG/Asc) and oral sodium sulfate (OSS) with respect to bowel preparation efficacy, compliance, and safety.

METHODS

A multicenter, prospective, randomized, single-blinded, non-inferiority trial was conducted in 3 hospitals. Patients were randomized to receive a bowel-cleansing agent. Bowel-cleansing efficacy was evaluated using the Boston Bowel Preparation Scale (BBPS). Satisfaction, feeling, taste of the bowel cleanser, and adverse events after taking the bowel cleanser were investigated through a questionnaire. Additionally, blood samples were analyzed before and after bowel cleansing.

RESULTS

In total, 172 patients were analyzed (85 with 1 L PEG/Asc and 87 with OSS), and the mean BBPS scores were comparable between agents. The 1L PEG/Asc group tended to have a higher BBPS score in the right colon (2.22 vs 2.02; P = .08). The compliance of 1 L of PEG/Asc was comparable to that of OSS. Patients taking 1 L PEG/Asc reported greater thirst and dizziness (P = .04 and P = .047, respectively) than the OSS cohort. On the other hand, gastrointestinal symptoms such as vomiting and abdominal distension were more common in the OSS group, without statistical significance. In terms of laboratory adverse events, elevation of serum creatinine was found in both groups after taking the bowel cleansing agent (P < .001 for the 1L PEG/Asc group; P = .04 for the OSS group). However, most of the increased values were within the normal ranges.

DISCUSSION

The 1L PEG/Asc treatment was comparable to OSS in terms of bowel preparation efficacy, compliance, and safety.

TGANet: Text-guided attention for improved polyp segmentation

Colonoscopy is a gold standard procedure but is highly operator-dependent. Automated polyp segmentation, a precancerous precursor, can minimize missed rates and timely treatment of colon cancer at an early stage. Even though there are deep learning methods developed for this task, variability in polyp size can impact model training, thereby limiting it to the size attribute of the majority of samples in the training dataset that may provide sub-optimal results to differently sized polyps. In this work, we exploit size-related and polyp number-related features in the form of text attention during training. We introduce an auxiliary classification task to weight the text-based embedding that allows network to learn additional feature representations that can distinctly adapt to differently sized polyps and can adapt to cases with multiple polyps. Our experimental results demonstrate that these added text embeddings improve the overall performance of the model compared to state-of-the-art segmentation methods. We explore four different datasets and provide insights for size-specific improvements. Our proposed text-guided attention network (TGANet) can generalize well to variable-sized polyps in different datasets. Codes are available at https://github.com/nikhilroxtomar/TGANet.

Establishment and evaluation of a nomogram predicting risks of missed diagnoses of colorectal polyps

Background

A missed diagnosis of colorectal polyps during colonoscopy may be associated with the occurrence of interval colorectal cancer. The risk factors for a missed diagnosis or a method to predict the risk of a missed diagnosis of colorectal polyps during colonoscopy remain unidentified.

Methods

The clinical data of patients who underwent two colonoscopies within three months at the Affiliated Hospital of North Sichuan Medical College between February 2017 and August 2019 were retrospectively reviewed. Independent risk factors for missed diagnoses were identified, and a nomogram was established to predict the risk of missed diagnoses. The prediction performance of the nomogram was evaluated using C-index and calibration curves, and its clinical application value was assessed using the Youden index and decision curve analysis.

Results

Independent influencing factors for missed diagnoses included age, endoscopist experience, bowel preparation, retroflected view, withdrawal time, number of polyps in the right colon, and number of polyps ≥ 6 mm. The C-index of the nomogram in the training and validation cohorts was 0.763 (95% confidence interval [CI]: 0.724 − 0.807) and 0.726 (95%CI: 0.657 − 0.794), respectively. The optimal cut-off value of the nomogram calculated using the Youden index was 152.2 points. Under the cut-off value, the sensitivity, specificity, positive predictive value, and negative predictive value were 67.1%, 75.7%, 45.8%, and 88.2%, respectively, in the training cohort, and 57.1%, 79.9%, 53.3%, and 82.3%, respectively, in the validation cohort.

Conclusions

The nomogram provides a reference value for clinicians to analyse the risk of a missed diagnosis of colorectal polyps in individuals, identify high-risk groups, and formulate appropriate follow-up strategies.

Endo-wing versus transparent hood-assisted colonoscopy for colorectal adenoma detection: A randomized controlled trial

BACKGROUND AND AIM

This study aimed to compare the mean number of adenomas in patients undergoing Endo-wing-assisted colonoscopy (EAC) and transparent hood-assisted colonoscopy (TAC).

METHODS

Patients undergoing colonoscopy for positive fecal immunochemical tests, colon polyp surveillance, and evaluation of abdominal symptoms at a single institution were randomly assigned to the EAC or TAC group. The mean number of adenomas per patient, adenoma detection rate, cecal intubation time, withdrawal time, mean number of adenomas per location, and adenoma size were compared.

RESULTS

Overall, 800 patients were enrolled. The EAC and TAC groups comprised 372 and 393 patients, respectively. The groups did not significantly differ with respect to cecal intubation and withdrawal times. The mean number of adenomas per patient was significantly higher in the EAC group (1.13 vs 0.90, P = 0.04), particularly in the sigmoid colon (0.54 [201/372] vs 0.38 [149/393], P = 0.04). The adenoma detection rates were 48.1% and 45.0% in the EAC and TAC groups, respectively, albeit without significant difference between the two groups (P = 0.393). The total number of sessile-type adenomas (0.73 [270/372] vs 0.47 [183/393], P < 0.0001) and small polyps (≤ 5 mm) (0.53 [198/372] vs 0.41 [159/393], P = 0.016) was significantly higher in the EAC group.

CONCLUSION

Endo-wing-assisted colonoscopy is significantly superior to TAC in terms of the mean number of adenomas per patient.

AI and Clinical Decision Making: The Limitations and Risks of Computational Reductionism in Bowel Cancer Screening

Advances in artificial intelligence in healthcare are frequently promoted as ‘solutions’ to improve the accuracy, safety, and quality of clinical decisions, treatments, and care. Despite some diagnostic success, however, AI systems rely on forms of reductive reasoning and computational determinism that embed problematic assumptions about clinical decision-making and clinical practice. Clinician autonomy, experience, and judgement are reduced to inputs and outputs framed as binary or multi-class classification problems benchmarked against a clinician’s capacity to identify or predict disease states. This paper examines this reductive reasoning in AI systems for colorectal cancer (CRC) to highlight their limitations and risks: (1) in AI systems themselves due to inherent biases in (a) retrospective training datasets and (b) embedded assumptions in underlying AI architectures and algorithms; (2) in the problematic and limited evaluations being conducted on AI systems prior to system integration in clinical practice; and (3) in marginalising socio-technical factors in the context-dependent interactions between clinicians, their patients, and the broader health system. The paper argues that to optimise benefits from AI systems and to avoid negative unintended consequences for clinical decision-making and patient care, there is a need for more nuanced and balanced approaches to AI system deployment and evaluation in CRC.

Colonoscopic image synthesis with generative adversarial network for enhanced detection of sessile serrated lesions using convolutional neural network

Computer-aided detection (CADe) systems have been actively researched for polyp detection in colonoscopy. To be an effective system, it is important to detect additional polyps that may be easily missed by endoscopists. Sessile serrated lesions (SSLs) are a precursor to colorectal cancer with a relatively higher miss rate, owing to their flat and subtle morphology. Colonoscopy CADe systems could help endoscopists; however, the current systems exhibit a very low performance for detecting SSLs. We propose a polyp detection system that reflects the morphological characteristics of SSLs to detect unrecognized or easily missed polyps. To develop a well-trained system with imbalanced polyp data, a generative adversarial network (GAN) was used to synthesize high-resolution whole endoscopic images, including SSL. Quantitative and qualitative evaluations on GAN-synthesized images ensure that synthetic images are realistic and include SSL endoscopic features. Moreover, traditional augmentation methods were used to compare the efficacy of the GAN augmentation method. The CADe system augmented with GAN synthesized images showed a 17.5% improvement in sensitivity on SSLs. Consequently, we verified the potential of the GAN to synthesize high-resolution images with endoscopic features and the proposed system was found to be effective in detecting easily missed polyps during a colonoscopy.

Unified Focal loss: Generalising Dice and cross entropy-based losses to handle class imbalanced medical image segmentation

Automatic segmentation methods are an important advancement in medical image analysis. Machine learning techniques, and deep neural networks in particular, are the state-of-the-art for most medical image segmentation tasks. Issues with class imbalance pose a significant challenge in medical datasets, with lesions often occupying a considerably smaller volume relative to the background. Loss functions used in the training of deep learning algorithms differ in their robustness to class imbalance, with direct consequences for model convergence. The most commonly used loss functions for segmentation are based on either the cross entropy loss, Dice loss or a combination of the two. We propose the Unified Focal loss, a new hierarchical framework that generalises Dice and cross entropy-based losses for handling class imbalance. We evaluate our proposed loss function on five publicly available, class imbalanced medical imaging datasets: CVC-ClinicDB, Digital Retinal Images for Vessel Extraction (DRIVE), Breast Ultrasound 2017 (BUS2017), Brain Tumour Segmentation 2020 (BraTS20) and Kidney Tumour Segmentation 2019 (KiTS19). We compare our loss function performance against six Dice or cross entropy-based loss functions, across 2D binary, 3D binary and 3D multiclass segmentation tasks, demonstrating that our proposed loss function is robust to class imbalance and consistently outperforms the other loss functions. Source code is available at: https://github.com/mlyg/unified-focal-loss.

•

Loss function choice is crucial for class-imbalanced medical imaging datasets.

•

Understanding the relationship between loss functions is key to inform choice.

•

Unified Focal loss generalises Dice and cross-entropy based loss functions.

•

Unified Focal loss outperforms various Dice and cross-entropy based loss functions.

Focus U-Net: A novel dual attention-gated CNN for polyp segmentation during colonoscopy

BACKGROUND

Colonoscopy remains the gold-standard screening for colorectal cancer. However, significant miss rates for polyps have been reported, particularly when there are multiple small adenomas. This presents an opportunity to leverage computer-aided systems to support clinicians and reduce the number of polyps missed.

METHOD

In this work we introduce the Focus U-Net, a novel dual attention-gated deep neural network, which combines efficient spatial and channel-based attention into a single Focus Gate module to encourage selective learning of polyp features. The Focus U-Net incorporates several further architectural modifications, including the addition of short-range skip connections and deep supervision. Furthermore, we introduce the Hybrid Focal loss, a new compound loss function based on the Focal loss and Focal Tversky loss, designed to handle class-imbalanced image segmentation. For our experiments, we selected five public datasets containing images of polyps obtained during optical colonoscopy: CVC-ClinicDB, Kvasir-SEG, CVC-ColonDB, ETIS-Larib PolypDB and EndoScene test set. We first perform a series of ablation studies and then evaluate the Focus U-Net on the CVC-ClinicDB and Kvasir-SEG datasets separately, and on a combined dataset of all five public datasets. To evaluate model performance, we use the Dice similarity coefficient (DSC) and Intersection over Union (IoU) metrics.

RESULTS

Our model achieves state-of-the-art results for both CVC-ClinicDB and Kvasir-SEG, with a mean DSC of 0.941 and 0.910, respectively. When evaluated on a combination of five public polyp datasets, our model similarly achieves state-of-the-art results with a mean DSC of 0.878 and mean IoU of 0.809, a 14% and 15% improvement over the previous state-of-the-art results of 0.768 and 0.702, respectively.

CONCLUSIONS

This study shows the potential for deep learning to provide fast and accurate polyp segmentation results for use during colonoscopy. The Focus U-Net may be adapted for future use in newer non-invasive colorectal cancer screening and more broadly to other biomedical image segmentation tasks similarly involving class imbalance and requiring efficiency.

Diagnostic Accuracy of Artificial Intelligence and Computer-Aided Diagnosis for the Detection and Characterization of Colorectal Polyps: Systematic Review and Meta-analysis

BACKGROUND

Colonoscopy reduces the incidence of colorectal cancer (CRC) by allowing detection and resection of neoplastic polyps. Evidence shows that many small polyps are missed on a single colonoscopy. There has been a successful adoption of artificial intelligence (AI) technologies to tackle the issues around missed polyps and as tools to increase the adenoma detection rate (ADR).

OBJECTIVE

The aim of this review was to examine the diagnostic accuracy of AI-based technologies in assessing colorectal polyps.

METHODS

A comprehensive literature search was undertaken using the databases of Embase, MEDLINE, and the Cochrane Library. PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines were followed. Studies reporting the use of computer-aided diagnosis for polyp detection or characterization during colonoscopy were included. Independent proportions and their differences were calculated and pooled through DerSimonian and Laird random-effects modeling.

RESULTS

A total of 48 studies were included. The meta-analysis showed a significant increase in pooled polyp detection rate in patients with the use of AI for polyp detection during colonoscopy compared with patients who had standard colonoscopy (odds ratio [OR] 1.75, 95% CI 1.56-1.96; P<.001). When comparing patients undergoing colonoscopy with the use of AI to those without, there was also a significant increase in ADR (OR 1.53, 95% CI 1.32-1.77; P<.001).

CONCLUSIONS

With the aid of machine learning, there is potential to improve ADR and, consequently, reduce the incidence of CRC. The current generation of AI-based systems demonstrate impressive accuracy for the detection and characterization of colorectal polyps. However, this is an evolving field and before its adoption into a clinical setting, AI systems must prove worthy to patients and clinicians.

TRIAL REGISTRATION

PROSPERO International Prospective Register of Systematic Reviews CRD42020169786; https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42020169786.

Impact of Inadequate Bowel Cleansing on Colonoscopic Findings in Routine Screening Practice

INTRODUCTION

Colonoscopy is an imperfect gold standard for detecting colorectal neoplasms because some proportion of adenomas may be missed, mainly small lesions. This proportion is expected to be higher in case of inadequate bowel cleansing, which is frequently seen in routine practice. We estimated the proportions of neoplasms that are in principle detectable by colonoscopy but might be missed in case of incomplete bowel preparation.

METHODS

For 8,193 participants of screening colonoscopy in South-Western Germany, recruited between 2005 and 2016, the prevalence and numbers of different findings were extracted from colonoscopy reports and compared according to the reported bowel preparation quality.

RESULTS

Bowel preparation quality was reported as good, poor, or was unspecified in 30.3%, 11.1%, and 58.6% of colonoscopy records. Reported prevalences of nonadvanced adenomas (NAAs) were similar among participants with poor and unspecified bowel preparation quality but substantially lower than among participants with good bowel preparation (adjusted prevalence rate ratio [RR] 0.86, 95% confidence interval [CI]: 0.77-0.96). The differences were observed for proximal but not for distal NAAs (RRs 0.82, 95% CI: 0.71-0.95 and 0.95, 95% CI: 0.82-1.10).

DISCUSSION

Our study suggests that a significant proportion of NAAs located in the proximal colon might be missed during colonoscopy if bowel cleansing is not adequate. Major efforts should be made to further facilitate and enhance high-quality bowel preparation in routine screening practice.

Effect of water exchange method on adenoma miss rate of patients undergoing selective polypectomy: A randomized controlled trial

BACKGROUND

Adenomas may be missed in up to 40% of screening colonoscopies. Although the water exchange (WE) method can improve ADR, as shown in several RCTs, it remains uncertain whether it can increase the detection of missing adenomas compared with standard air-insufflated (AI) colonoscopy.

METHODS

Patients aged 18-80 years who underwent selective polypectomy were randomly allocated to the WE or AI group. The primary endpoint was the adenoma miss rate (AMR), defined as the number of patients with one or more additional adenomas during the polypectomy procedure divided by the total number of patients in each group.

RESULTS

A total of 450 patients were enrolled, with 225 in each group. The overall AMRs were 45.8% (103/225) in the WE group and 35.6% (80/225) in the AI group (p = 0.035). More patients in the WE group had at least one missed adenoma in the proximal colon (38.2% vs 24.4%, p = 0.002). The adenoma-level miss rate was also higher in the WE group than in the AI group (35.1% vs 29.0%, p = 0.036). Subgroup analysis showed that patients in the WE group had more missed adenomas located in the proximal colon or with flat shapes.

CONCLUSIONS

This study confirmed that substantial adenomas were missed in patients undergoing selective polypectomy. The WE method significantly improved the detection of missed adenomas, especially those located in the proximal colon or with flat shapes. (ClnicalTrials.gov number: NCT02880748).

Risk factors associated with missed colorectal lesions in colonoscopy and impact of colonoscopy with anesthesia on miss rate

OBJECTIVES

To systematically determine the miss rate and risk factors for polyps, adenomas and advanced adenomas in the same population, and to further analyze the impact of colonoscopy with anesthesia on miss rate.

METHODS

We retrospectively analyzed the information of the patients undergoing the second colonoscopy within 1 year after their first. The patient and lesion miss rate were calculated. The patient and lesion features of missed lesion were compared with non-missed lesion. Finally, the patients were divided into anesthesia group and without anesthesia group, and the impact of colonoscopy with anesthesia on missed lesions was further analyzed.

RESULTS

The patient miss rate of polyps, adenomas and advanced adenomas was 32.8, 25.6 and 10.4%, and the lesions miss rate was 19.6, 15.8 and 7.2%. In multivariable logistic regression analysis, lesion-related factors (large number of lesions, small lesion size, flat shape and location at the right colon) and patient-related factors (male, elder, abdominal symptoms, surgical history, diverticulum, colonoscopy without anesthesia and suboptimal bowel preparation) were found to be independently associated with missed polyps and adenomas (p < .05). Large number of lesions, flat shape and suboptimal bowel preparation were associated with missed advanced adenoma (p < .05). Colonoscopy with anesthesia can reduce the polyp miss rate (PMR) and male and elderly patients are more likely to be missed during colonoscopy without anesthesia.

CONCLUSIONS

Many factors of patients and lesions can affect the lesions miss rate. Colonoscopy with anesthesia can reduce the PMR and male and elderly patients are more likely to be missed during colonoscopy without anesthesia.

Feature Point Tracking-Based Localization of Colon Capsule Endoscope

In large bowel investigations using endoscopic capsules and upon detection of significant findings, physicians require the location of those findings for a follow-up therapeutic colonoscopy. To cater to this need, we propose a model based on tracking feature points in consecutive frames of videos retrieved from colon capsule endoscopy investigations. By locally approximating the colon as a cylinder, we obtained both the displacement and the orientation of the capsule using geometrical assumptions and by setting priors on both physical properties of the intestine and the image sample frequency of the endoscopic capsule. Our proposed model tracks a colon capsule endoscope through the large intestine for different prior selections. A discussion on validating the findings in terms of intra and inter capsule and expert panel validation is provided. The performance of the model is evaluated based on the average difference in multiple reconstructed capsule’s paths through the large intestine. The path difference averaged over all videos was as low as 4±0.7 cm, with min and max error corresponding to 1.2 and 6.0 cm, respectively. The inter comparison addresses frame classification for the rectum, descending and sigmoid, splenic flexure, transverse, hepatic, and ascending, with an average accuracy of 86%.

Comparison of adenoma detection by colonoscopy between polypectomy performed during both insertion and withdrawal versus during withdrawal only: a multicenter, randomized, controlled trial

BACKGROUND AND AIM

In standard colonoscopy, the colonoscope is inserted into the cecum, and inspection of the colonic mucosa and polypectomy are performed during withdrawal. The colon configuration can differ between the insertion and withdrawal phases, and some polyps found in the insertion phase can be missed during withdrawal. A few single-center studies investigated whether detection of polyps during the insertion phase affects the adenoma detection rate (ADR). However, the effectiveness of this strategy is unknown because of conflicting results. We aimed to determine whether polypectomy together with careful inspection during insertion increases the ADR compared with standard colonoscopy.

METHODS

A randomized, controlled, multicenter trial was conducted at three university hospitals. Patients aged 50 to 80 years were randomly assigned to the study group or control group. For patients in the study group, polypectomy was performed together with careful inspection during both colonoscope insertion and withdrawal. In the control group, polyps were inspected and removed only during colonoscope withdrawal. The primary endpoint was the ADR, which was defined as the percentage of patients with ≥ 1 adenoma.

RESULTS

A total of 1142 patients were enrolled (study group, n = 571; control group, n = 571). The ADR was similar in the 2 groups (study group, 44.1%; control group, 43.1%; P = 0.72). In the control group, 12 polyps that had been detected during colonoscope insertion were not found during withdrawal (polyp miss rate: 2.1%, 12/571).

CONCLUSION

Polypectomy and careful inspection during both colonoscope insertion and withdrawal did not improve the overall ADR compared with standard colonoscopy (NCT01925833).

PICCOLO White-Light and Narrow-Band Imaging Colonoscopic Dataset: A Performance Comparative of Models and Datasets

Colorectal cancer is one of the world leading death causes. Fortunately, an early diagnosis allows for effective treatment, increasing the survival rate. Deep learning techniques have shown their utility for increasing the adenoma detection rate at colonoscopy, but a dataset is usually required so the model can automatically learn features that characterize the polyps. In this work, we present the PICCOLO dataset, that comprises 3433 manually annotated images (2131 white-light images 1302 narrow-band images), originated from 76 lesions from 40 patients, which are distributed into training (2203), validation (897) and test (333) sets assuring patient independence between sets. Furthermore, clinical metadata are also provided for each lesion. Four different models, obtained by combining two backbones and two encoder–decoder architectures, are trained with the PICCOLO dataset and other two publicly available datasets for comparison. Results are provided for the test set of each dataset. Models trained with the PICCOLO dataset have a better generalization capacity, as they perform more uniformly along test sets of all datasets, rather than obtaining the best results for its own test set. This dataset is available at the website of the Basque Biobank, so it is expected that it will contribute to the further development of deep learning methods for polyp detection, localisation and classification, which would eventually result in a better and earlier diagnosis of colorectal cancer, hence improving patient outcomes.