SRGAN Assisted Encoder-Decoder Deep Neural Network for Colorectal Polyp Semantic Segmentation

Early diagnosis is essential for the successful treatment of bowel cancers including colorectal cancer (CRC), and capsule endoscopic imaging with robotic actuation can be a valuable diagnostic tool when combined with automated image analysis. We present a deep learning rooted detection and segmentation framework for recognizing lesions in colonoscopy and capsule endoscopy images. We restructure established convolution architectures, such as VGG and ResNets, by converting them into fully-connected convolution networks (FCNs), fine-tune them and study their capabilities for polyp segmentation and detection. We additionally use shape-from-shading (SfS) to recover depth and provide a richer representation of the tissue’s structure in colonoscopy images. Depth is incorporated into our network models as an additional input channel to the RGB information and we demonstrate that the resulting network yields improved performance. Our networks are tested on publicly available datasets and the most accurate segmentation model achieved a mean segmentation interception over union (IU) of 47.78% and 56.95% on the ETIS-Larib and CVC-Colon datasets, respectively. For polyp detection, the top performing models we propose surpass the current state-of-the-art with detection recalls superior to 90% for all datasets tested. To our knowledge, we present the first work to use FCNs for polyp segmentation in addition to proposing a novel combination of SfS and RGB that boosts performance.

IntroductionMost colon cancers arise within preexisting adenomatous polyps or adenomas. Thereby, the carcinogenesis of colon cancer favors a strategy of early detection. Detection of early stage colorectal cancer, and finding and removal of pre-malignant adenomatous polyps have been shown to beneficially impact on colorectal cancer incidence and mortality. The identification of an early tumor marker that would allow reliable early cancer detection could lead to a diagnostic assay that would greatly aid in the management of this disease. The nuclear matrix is the residual framework scaffolding of the nucleus and consists of the peripheral lamins and pore complexes, an internal ribonucleic protein network and residual nucleoli. NMPs have been demonstrated to participate in many vital cellular functions, such as steroid hormone binding, gene transcription and translation. Most of the nuclear matrix proteins identified to date are common to all cell types, but several identified NMPs are tissue and cell line specific [1, 2]. Cell type-specific „fingerprinting“ of aberrant nuclear matrix proteins and their appearance in cancer development has led to the analysis of nuclear matrix protein composition of a variety of tumors in an effort to determine whether these proteins can be developed as diagnostic markers for cancer.We have identified four specific nuclear matrix proteins (NMP: CC2, CC3, CC4, CC5) that distinguish colon cancer from normal adjacent mucosa, and from normal mucosa in subjects without colon cancer [3]. The objective of this study was to describe and characterize the nuclear matrix protein profile in pre-malignant adenomatous colon polyps. In particular, we wanted to determine when in the adenoma-carcinoma sequence can the unique colon cancer NMP’s be identified.Material and MethodsUsing 2-D gel analysis with following silver staining 20 colon polyps (one juvenile polyp, six tubular adenoma (TA), seven tubulovillous adenoma (TVA), six tubulovillous adenoma with focal high grade dysplasia (HGD), and two human colon cancer cell lines (CaCo2, Cx-1), were analyzed for the presence of these specific nuclear matrix proteins.ResultsThe protein spot CC2 was not seen in any of the pre-malignant polyps, but was present in 80% of colon cancers (N = 10). The protein spot CC5 was present in only 2 pre-malignant TVA with HGD and in one TA, but was present in all colon cancer tissues (10/10). CC3 and CC4 were present in most adenomas, regardless of advancement and in all colon cancer tissues. None of the nuclear matrix proteins were seen in the juvenile polyp, which is not a precursor of colon cancer. The cell line CaCo2 expressed CC3 and CC4 and the cell line CX-1 CC2, CC3 and CC4.ConclusionIt is known that the risk of malignancy within an adenomatous polyp correlates with the sizes, histologic type, and degree of dysplasia. CC2 and CC5 are proteins that are expressed at the junction of an advanced adenoma and invasive colon cancer. CC3 and CC4 are expressed earlier in the development of adenomatous polyps. All four spots are specific for colon cancer and are not found in normal adjacent or normal donor tissue or in the juvenile polyp. The fact that these proteins are shown already in the pre-malignant phase of colon cancer indicates them as useful early detection target. Additionally the functional identification of these proteins could lead to new knowledge in the carciogenesis in colon cancer. The generation of antibodies to them is being actively pursued. Development of an assay to these nuclear matrix proteins may serve as a new promising method for early detection of colon cancer.

IntroductionColonic adenoma detection is one of the quality indicators of endoscopy1 and is measured as present or not, rather than an absolute number of adenomatous polyps per colonoscopy. Several factors...

Purpose: Colonoscopy is performed as a screening procedure for colorectal cancer (CRC). During this procedure, polyps, which could be premalignant lesions, are removed thereby reducing the risk of development of CRC. Gastroenterologists also perform colonoscopy for other indications such as bleeding and anemia. The aim of this study is to determine if there is a difference in the rate of polyp detection between screening and other indications for colonoscopy. Methods: All colonoscopies performed at Henry Ford Hospital for a period of one year between 03/01/1999 to 04/01/2000 were reviewed. The indication for colonoscopy and findings were noted in all cases. The primary end point was the detection of any polyp based on size and the secondary end point was the detection of an adenomatous polyp. The chi-square test was used to calculate the percent differences in polyp detection between groups based on the indication for colonoscopy.Table: Table 1. Indication of colonoscopy and rate of polyp detectionResults: There were a total of 2096 colonoscopies performed on 1547 patients. Of these patients, 48% were male, 67% were African American and 30% were Caucasian. Of the 2096 colonoscopies performed, polyps were detected in 980 (46.7%), and of these 675 (32%) were adenomatous polyps. Major indications for colonoscopy were screening (12%), surveillance (18%), abnormal radiology (4%), bleeding (30%), positive fecal occult blood test (17%), iron deficiency anemia (11%) and bowel habit change (7%). Rate of detection of any polyp, <10 mm polyp and adenomatous polyp are presented in Table-1. Rate of polyp detection (for all three groups) in screening and surveillance colonoscopy were higher in comparison with other indications and were statistically significant (except abnormal radiology). Conclusion: The rate of detection of all polyps, including polyps <10 mm and adenomatous polyps, was significantly higher in screening and surveillance colonoscopy than other indications for colonoscopy. Any program using colonoscopy to screen for colo-rectal cancer that is based on bleeding or anemia has rates of polyp detection inferior to general screening and is inappropriate.

Colorectal cancer remains a leading cause of cancer-related deaths worldwide, with early detection and removal of polyps being critical in preventing disease progression. Automated polyp segmentation, particularly in colonoscopy images, is a challenging task due to the variability in polyp appearance and the low contrast between polyps and surrounding tissues. In this work, we propose an edge-enhanced network (EENet) designed to address these challenges by integrating two novel modules: the covariance edge-enhanced attention (CEEA) and cross-scale edge enhancement (CSEE) modules. The CEEA module leverages covariance-based attention to enhance boundary detection, while the CSEE module bridges multi-scale features to preserve fine-grained edge details. To further improve the accuracy of polyp segmentation, we introduce a hybrid loss function that combines cross-entropy loss with edge-aware loss. Extensive experiments show that the EENet achieves a Dice score of 0.9208 and an IoU of 0.8664 on the Kvasir-SEG dataset, surpassing state-of-the-art models such as Polyp-PVT and PraNet. Furthermore, it records a Dice score of 0.9316 and an IoU of 0.8817 on the CVC-ClinicDB dataset, demonstrating its strong potential for clinical application in polyp segmentation. Ablation studies further validate the contribution of the CEEA and CSEE modules.

&lt;p&gt;Colorectal cancer (CRC) is one of the most prevalent malignant tumors in Colombia and the world. These neoplasms originate in adenomatous lesions or polyps that must be resected to prevent the disease, which can be done with a colonoscopy. It has been reported that during colonoscopy polyps are detected in 40% of men and 30% of women (hyperplastic, adenomatous, serrated, among others), and, on average, 25% of adenomatous polyps (main quality indicator in colonoscopy). However, these lesions are not easy to observe due to the multiplicity of blind spots in the colon and the human error associated with the examination. Objective: to create a computational method for the automatic detection of colorectal polyps using artificial intelligence in recorded videos of real colonoscopy procedures. Methodology: public databases with colorectal polyps and a data collection built in a University Hospital were used. Initially, all the frames of the videos were normalized to reduce the high variability between databases. Subsequently, the polyp detection task is done with a deep learning method using a convolutional neural network. This network is initialized with weights learned on millions of national images from the ImageNet database. The weights of the network are updated using colonoscopy images, following the tuning technique. &lt;strong&gt;&lt;/strong&gt;&lt;/p&gt;

IntroductionThe EndoringsTM is a distal attachment consisting of two layers of circular flexible rings that evert mucosal folds. Aims; to investigate if Endorings Colonoscopy (ER) improves polyp and adenoma detection...

Colorectal cancer (CRC) is one of the most common and deadliest forms of cancer, accounting for nearly 10% of all forms of cancer in the world. Even though colonoscopy is considered the most effective method for screening and diagnosis, the success of the procedure is highly dependent on the operator skills and level of hand-eye coordination. In this work, we propose to adapt fully convolution neural networks (FCN), to identify and segment polyps in colonoscopy images. We converted three established networks into a fully convolution architecture and fine-tuned their learned representations to the polyp segmentation task. We validate our framework on the 2015 MICCAI polyp detection challenge dataset, surpassing the state-of-the-art in automated polyp detection. Our method obtained high segmentation accuracy and a detection precision and recall of 73.61% and 86.31%, respectively.

&lt;div&gt;&lt;p class="4"&gt;&lt;strong&gt;Objective:&lt;/strong&gt; to establish an automatic colonoscopy method based on artificial intelligence. &lt;strong&gt;Methods:&lt;/strong&gt; a public database established by a university hospital was used, including colorectal fat and data collection. Initially, all frames in the video are normalized to reduce the high variability between databases. Then, the convolution neural network is used for full depth learning to complete the detection task of polyps. The network starts with the weights learned from millions of natural images in the ImageNet database. According to the fine-tuning technology, the colonoscopy image is used to update the network weight. Finally, the detection of polyps is performed by assigning the probability of containing Po ́ lipo to each table and determining the threshold defined when polyps appears in the table. &lt;strong&gt;Results:&lt;/strong&gt; 1875 cases were collected from 5 public databases and databases established by university hospitals, with a total of 123046 forms. The method was trained and evaluated. Comparing the results with the scores of different colonoscopy experts, the accuracy was 0.77, the sensitivity was 0.89, the specificity was 0.71, and the ROC curve (re ceiver operation characteristics) was 0.87. &lt;strong&gt;Conclusion:&lt;/strong&gt; compared with experienced gastrointestinal markers, this method overcomes the high variability of different types of lesions and different colonic light conditions (handle, folding or contraction), has very high sensitivity, and can reduce human errors, which is one of the main factors leading to the non detection or leakage of Po lipids in colonoscopy.&lt;/p&gt;&lt;/div&gt;

Adjuvant Chemotherapy Administration and Survival Outcomes in Lymphoma Survivors with Common Solid Tumors: A Population-Based Study

Purpose: The administration of anti-spasmotic drugs during colonoscopy has been proposed to increase the adenoma detection rate (ADR). However, conflicting results have been reported on the impact of Hyoscine N-butyl-bromide (HBB) on ADR. The primary outcome of the study was to test whether the administration of HBB, at time of cecal intubation, increases the polyp and adenoma detection. The HBB safety, the entity of abdominal distension at the end of the procedure and the patient's bloating perception were also evaluated. Methods: Consecutive colonoscopy adult outpatients were considered for the study. Exclusion criteria were represented by co-morbidities regarded as contraindications to HBB administration, previous intestinal resection and history of IBD. After cecal intubation, patients were randomized in a double blind fashion to receive either 20 mg/1ml of HBB i.v. or 1ml i.v. of placebo (saline). The number, size, histology and location of polyps and the occurrence of episodes of significant tachycardia (defined as bpm>140 persisting for more than 30 sec.) were recorded. The air retained in the abdomen at the end of the procedure was indirectly estimated by calculating the δAC (difference in the abdominal circumference measured before and after colonoscopy). At discharge the bloating perceived by the patient was measured by means of a VA scale (range 0-100). A sample size of at least 400 patients was calculated, by hypothesizing an increase of ADR (primary outcome measure) from 28% to 43% (α=0.05, β=0.9). Results: 402 patients (mean age 57.7+12.5 years; 44% men) were randomized to receive either HBB (202 patients) or saline (200 patients). No difference was observed as concerns polyp detection rate, ADR, advanced-ADR, mean number of polyps detected per-patient in the HBB and saline groups (see Table). Significant tachycardia was observed in 2.9% of HBB patients and in 0.5% of those receiving saline (p=ns). The δAC was similar between patients receiving HBB or saline (mean+SD 2.5+2.8 cm and 2.2+2.3 respectively; p=ns) as well as the bloating perception (VAS mean+SD 12.9+13.0 and 13.9+12.9 respectively; p=ns).Table: No Caption available.Conclusion: HBB administered before colonoscopy withdrawal phase does not increase the polyp and adenoma detection. The medication is well tolerated and it does not increase either the abdominal distension or the bloating perceived by patients at discharge.

To evaluate association(s) between withdrawal time and polyp detection in various bowel preparation qualities. Retrospective cohort analysis of screening colonoscopies performed between January 2005 and June 2011 for patients with average risk of colorectal cancer. Exclusion criteria included patients with a personal history of adenomatous polyps or colon cancer, prior colonic resection, significant family history of colorectal cancer, screening colonoscopy after other abnormal screening tests such as flexible sigmoidoscopy or barium enema, and screening colonoscopies during in-patient care. All procedures were performed or directly supervised by gastroenterologists. Main measurements were number of colonic segments with polyps and total number of colonic polyps. Multivariate analysis of 8331 colonoscopies showed longer withdrawal time was associated with more colonic segments with polyps in good (adjusted OR = 1.16; 95%CI: 1.13-1.19), fair (OR = 1.13; 95%CI: 1.10-1.17), and poor (OR = 1.18; 95%CI: 1.11-1.26) bowel preparation qualities. A higher number of total polyps was associated with longer withdrawal time in good (OR = 1.15; 95%CI: 1.13-1.18), fair (OR = 1.13; 95%CI: 1.10-1.16), and poor (OR = 1.20; 95%CI: 1.13-1.29) bowel preparation qualities. Longer withdrawal time was not associated with more colonic segments with polyps or greater number of colonic polyps in bowel preparations with excellent (OR = 1.07, 95%CI: 0.99-1.26; OR = 1.11, 95%CI: 0.99-1.24, respectively) and very poor (OR = 1.02, 95%CI: 0.99-1.12; OR = 1.05, 95%CI: 0.99-1.10, respectively) qualities. Longer withdrawal time is not associated with higher polyp number detected in colonoscopies with excellent or very poor bowel preparation quality.

*Royal Wolverhampton Hospitals NHS Trust, Wolverhampton, UK, †Torbay Hospital, Torquay, UK, ‡Department of General Surgery, Manchester Royal Infirmary, Manchester, UK, §Royal Infirmary, Glasgow, UK, ¶Department of Histopathology, Royal Bolton NHS Foundation Trust, Bolton, UK, **Imperial College Healthcare NHS Trust, Charing Cross Hospital, London, UK, ††The Royal Marsden Hospital, London, UK, ‡‡Salisbury NHS Foundation Trust, Salisbury, UK and §§Department of Histopathology, University Hospital of South Manchester, Manchester, UK

To compare the bowel cleansing efficacy of same day ingestion of 4-L sulfa-free polyethylene glycol (4-L SF-PEG) vs 2-L polyethylene glycol solution with ascorbic acid (2-L PEG + Asc) in patients undergoing afternoon colonoscopy. 206 patients (mean age 56.7 years, 61% male) undergoing outpatient screening or surveillance colonoscopies were prospectively randomized to receive either 4-L SF-PEG (n = 104) or 2-L PEG + Asc solution (n = 102). Colonoscopies were performed by two blinded endoscopists. Bowel preparation was graded using the Ottawa scale. Each participant completed a satisfaction and side effect survey. There was no difference in patient demographics amongst groups. 4-L SF-PEG resulted in better Ottawa scores compared to 2-L PEG + Asc, 4.2 vs 4.9 (P = 0.0186); left colon: 1.33 vs 1.57 respectively (P = 0.0224), right colon: 1.38 vs 1.63 respectively (P = 0.0097). No difference in Ottawa scores was found for the mid colon or amount of fluid. Patient satisfaction was similar for both arms but those assigned to 4-L SF-PEG reported less bloating: 23.1% vs 11.5% (P = 0.0235). Overall polyp detection, adenomatous polyp and advanced adenoma detection rates were similar between the two groups. Morning only 4-L SF-PEG provided superior cleansing with less bloating as compared to 2-L PEG + Asc bowel preparation for afternoon colonoscopy. Thus, future studies evaluating efficacy of morning only preparation for afternoon colonoscopy should use 4-L SF-PEG as the standard comparator.

Postoperative surveillance colonoscopy is regarded as an effective tool for the detection and the diagnoses of metachronous cancer and adenomas. Detection and removal of adenomatous polyps though postoperative colonoscopy may reduce the risk of recurrence of colon cancer and advanced adenomas [1]. National Comprehensive Cancer Network guidelines recommend postoperative surveillance colonoscopy within the first year after a resection of colorectal cancer [2]. Postpolypectomy surveillance guidelines published by the US Multi-society Task Force on Colorectal Cancer and based on the number of polyps, the polyp size, and the pathology were presented in 2006. In 2012, those guidelines added surveillance recommendations for serrated lesions to the previous guidelines. Guidelines are to delineate predictors of advanced pathology, both cancer and advanced adenomas. According to the guidelines, patients can be more definitely stratified at their baseline colonoscopies into those at lower risk and those at increased risk for advanced neoplasia [3]. Guidelines for postoperative surveillance colonoscopy recommend that the presence of adenomas and sessile serrated polyps be considered to determine the follow-up interval [3]. Also recommended is that the surveillance interval be based on the risk of polyps found at index colonoscopy [3]. Ratuapli et al. [4] said that knowledge of the pathology of the actual polyp does not change the postpolypectomy surveillance recommendations. Interval cancers within one year after surgery may also represent lesions that were missed on the index colonoscopy. The experienced endoscopist will miss about 15% of neoplastic polyps smaller than 10 mm in size, but will rarely miss larger polyps [5]. Another study showed an overall miss rate for adenomas of 24%; only 6% of adenomas with sizes larger than 1 cm were overlooked during the first examination [6]. Brenner et al. [7] demonstrated that the protective effect of colonoscopy against colorectal cancer was not uniform throughout the colon and that the right colon had a lower rate of protection. Factors that may contribute to lower adenoma detection, particularly in the right colon, include the prevalence of flat and depressed lesions, as well as serrated adenomas, in the more proximal segments, which are more difficult to visualize [7]. A higher proportion of polyps were adenomatous in the right and the proximal colon compared with the distal colon [8]. The adenoma detection rates for the proximal colon (22.9%) and the right colon (17.1%) were significantly higher than the adenoma detection rate for the left colon (13.5%) (both P < 0.001) [8]. A considerable variation in adenoma detection exists among endoscopists. Colonoscopies performed by endoscopists with higher adenoma detection rates are associated with lower rates of interval cancer [8]. In this study, the authors said that in the 116 enrolled patients, the incidence of total polyps during postoperative surveillance colonoscopy was 53.4% and that the incidence of neoplastic polyps was 25.9%. They also said that the miss rate of total polyps was 37.4% and the miss rate of neoplastic polyps was 24.2% during postoperative colonoscopy. The missed neoplastic polyps were located in the right colon and the proximal part from the primary tumor at a higher percentage. When more neoplastic polyps were detected during preoperative colonoscopy, more neoplastic polyps were missed during postoperative colonoscopy [9]. Repeat surveillance colonoscopy earlier than recommended may increase healthcare expenditures, and delayed surveillance colonoscopy may increase the number of patients with an interval cancer [4]. Baseline postoperative surveillance colonoscopy with excellent patient preparation and adequate withdrawal time could reduce the miss rate for colorectal cancer and adenomas. Postoperative surveillance colonoscopy is very important because it can have a great effect on reducing the incidence of colorectal cancer in patients with adenomatous polyps. Especially, the right side of Correspondence to: Byung Chun Kim, M.D. Department of Surgery, Hallym University Kangnam Sacred Heart Hospital, Hallym University College of Medicine, 1 Singil-ro, Yeongdeungpo-gu, Seoul 07441, Korea Tel: +82-2-829-5130, Fax: +82-2-849-4469 E-mail: bckimgs@hallym.or.kr