Artificial intelligence and computer-aided diagnosis in colonoscopy: current evidence and future directions

Cost savings in colonoscopy with artificial intelligence-aided polyp diagnosis: an add-on analysis of a clinical trial (with video)

BACKGROUND AND AIMS

Artificial intelligence (AI) is being implemented in colonoscopy practice, but no study has investigated whether AI is cost saving. We aimed to quantify the cost reduction using AI as an aid in the optical diagnosis of colorectal polyps.

METHODS

This study is an add-on analysis of a clinical trial that investigated the performance of AI for differentiating colorectal polyps (ie, neoplastic versus non-neoplastic). We included all patients with diminutive (≤5 mm) rectosigmoid polyps in the analyses. The average colonoscopy cost was compared for 2 scenarios: (1) a diagnose-and-leave strategy supported by the AI prediction (ie, diminutive rectosigmoid polyps were not removed when predicted as non-neoplastic), and (2) a resect-all-polyps strategy. Gross annual costs for colonoscopies were also calculated based on the number and reimbursement of colonoscopies conducted under public health insurances in 4 countries.

RESULTS

Overall, 207 patients with 250 diminutive rectosigmoid polyps (104 neoplastic, 144 non-neoplastic, and 2 indeterminate) were included. AI correctly differentiated neoplastic polyps with 93.3% sensitivity, 95.2% specificity, and 95.2% negative predictive value. Thus, 105 polyps were removed and 145 were left under the diagnose-and-leave strategy, which was estimated to reduce the average colonoscopy cost and the gross annual reimbursement for colonoscopies by 18.9% and US$149.2 million in Japan, 6.9% and US$12.3 million in England, 7.6% and US$1.1 million in Norway, and 10.9% and US$85.2 million in the United States, respectively, compared with the resect-all-polyps strategy.

CONCLUSIONS

The use of AI to enable the diagnose-and-leave strategy results in substantial cost reductions for colonoscopy.

Clinical Trials for Artificial Intelligence in Cancer Diagnosis: A Cross-Sectional Study of Registered Trials in ClinicalTrials.gov

Objective: Clinical trials are the most effective way to judge the merits of diagnosis and treatment strategies. The in-depth mining of clinical trial data enables us to grasp the application trend of artificial intelligence (AI) for cancer diagnosis. The aim of this study was to analyze the characteristics of registered trials on AI for cancer diagnosis. Methods: Clinical trials on AI for cancer diagnosis registered on the ClinicalTrials.gov database were searched and downloaded. Statistical analysis was performed by using SPSS 20.0 software. Results: A total of 97 registered trials were included. Of them, only 27 (27.8%) were interventional trials and 70 (72.1%) were observational trials. Fifteen (15.4%) trials had been completed. Fifty trials were in recruitment, and another 18 remained unrecruited. The number of cases included in the clinical trials tended to be large, 31 (32.0%) trials including samples ranging from 100 to 499 cases and 17 (17.5%) trials including samples ranging from 500 to 999 cases. Of the 27 interventional trials, only two trials reported trials' phase. Most (85.2%) interventional trials were for diagnosis, and a few (3.7%) were for the purpose of both the diagnosis and therapy of cancers. For the observational clinical trials, 46 (65.7%) were cohort studies, and 11 (15.7%) were case-only studies. Among the observational trials, 46 (65.7%) were prospective studies and 13 (18.6%) were retrospective studies. Among 97 trials, 37 (38.1%) involved colorectal cancer, 11 (11.3%) involved breast cancer, 43 (44.3%) were for imaging diagnosis, 33 (34.0%) were for endoscopic diagnosis, and 11 (11.3%) were for pathological diagnosis. For the interventional trials, 11 trials were parallel assignment (40.7%), and 14 were single group assignment (51.9%). Among the 27 interventional trials, 18 (66.7%) trials were performed without masking, 6 (22.2%) trials were performed with single masking, only 1 (3.7%) was performed with double masking, and 2 (7.4%) was performed with triple masking. Conclusion: It appears that most registered trials on AI for cancer diagnosis are observational design, and more trials are needed in this field.

Deep learning to find colorectal polyps in colonoscopy: A systematic literature review

Colorectal cancer has a great incidence rate worldwide, but its early detection significantly increases the survival rate. Colonoscopy is the gold standard procedure for diagnosis and removal of colorectal lesions with potential to evolve into cancer and computer-aided detection systems can help gastroenterologists to increase the adenoma detection rate, one of the main indicators for colonoscopy quality and predictor for colorectal cancer prevention. The recent success of deep learning approaches in computer vision has also reached this field and has boosted the number of proposed methods for polyp detection, localization and segmentation. Through a systematic search, 35 works have been retrieved. The current systematic review provides an analysis of these methods, stating advantages and disadvantages for the different categories used; comments seven publicly available datasets of colonoscopy images; analyses the metrics used for reporting and identifies future challenges and recommendations. Convolutional neural networks are the most used architecture together with an important presence of data augmentation strategies, mainly based on image transformations and the use of patches. End-to-end methods are preferred over hybrid methods, with a rising tendency. As for detection and localization tasks, the most used metric for reporting is the recall, while Intersection over Union is highly used in segmentation. One of the major concerns is the difficulty for a fair comparison and reproducibility of methods. Even despite the organization of challenges, there is still a need for a common validation framework based on a large, annotated and publicly available database, which also includes the most convenient metrics to report results. Finally, it is also important to highlight that efforts should be focused in the future on proving the clinical value of the deep learning based methods, by increasing the adenoma detection rate.

The application of deep learning based diagnostic system to cervical squamous intraepithelial lesions recognition in colposcopy images

Background Deep learning has presented considerable potential and is gaining more importance in computer assisted diagnosis. As the gold standard for pathologically diagnosing cervical intraepithelial lesions and invasive cervical cancer, colposcopy-guided biopsy faces challenges in improving accuracy and efficiency worldwide, especially in developing countries. To ease the heavy burden of cervical cancer screening, it is urgent to establish a scientific, accurate and efficient method for assisting diagnosis and biopsy. Methods The data were collected to establish three deep-learning-based models. For every case, one saline image, one acetic image, one iodine image and the corresponding clinical information, including age, the results of human papillomavirus testing and cytology, type of transformation zone, and pathologic diagnosis, were collected. The dataset was proportionally divided into three subsets including the training set, the test set and the validation set, at a ratio of 8:1:1. The validation set was used to evaluate model performance. After model establishment, an independent dataset of high-definition images was collected to further evaluate the model performance. In addition, the comparison of diagnostic accuracy between colposcopists and models weas performed. Results The sensitivity, specificity and accuracy of the classification model to differentiate negative cases from positive cases were 85.38%, 82.62% and 84.10% respectively, with an AUC of 0.93. The recall and DICE of the segmentation model to segment suspicious lesions in acetic images were 84.73% and 61.64%, with an average accuracy of 95.59%. Furthermore, 84.67% of high-grade lesions were detected by the acetic detection model. Compared to colposcopists, the diagnostic system performed better in ordinary colposcopy images but slightly unsatisfactory in high-definition images. Implications The deep learning-based diagnostic system could help assist colposcopy diagnosis and biopsy for HSILs.

How Artificial Intelligence Will Impact Colonoscopy and Colorectal Screening

Artificial intelligence may improve value in colonoscopy-based colorectal screening and surveillance by improving quality and decreasing unnecessary costs. The quality of screening and surveillance as measured by adenoma detection rates can be improved through real-time computer-assisted detection of polyps. Unnecessary costs can be decreased with optical biopsies to identify low-risk polyps using computer-assisted diagnosis that can undergo the resect-and-discard or diagnose-and-leave strategy. Key challenges include the clinical integration of artificial intelligence-based technology into the endoscopists' workflow, the effect of this technology on endoscopy center efficiency, and the interpretability of the underlying deep learning algorithms. The future for image-based artificial intelligence in gastroenterology will include applications to improve the diagnosis and treatment of cancers throughout the gastrointestinal tract.

Diagnostic evaluation of a deep learning model for optical diagnosis of colorectal cancer

Colonoscopy is commonly used to screen for colorectal cancer (CRC). We develop a deep learning model called CRCNet for optical diagnosis of CRC by training on 464,105 images from 12,179 patients and test its performance on 2263 patients from three independent datasets. At the patient-level, CRCNet achieves an area under the precision-recall curve (AUPRC) of 0.882 (95% CI: 0.828-0.931), 0.874 (0.820-0.926) and 0.867 (0.795-0.923). CRCNet exceeds average endoscopists performance on recall rate across two test sets (91.3% versus 83.8%; two-sided t-test, p < 0.001 and 96.5% versus 90.3%; p = 0.006) and precision for one test set (93.7% versus 83.8%; p = 0.02), while obtains comparable recall rate on one test set and precision on the other two. At the image-level, CRCNet achieves an AUPRC of 0.990 (0.987-0.993), 0.991 (0.987-0.995), and 0.997 (0.995-0.999). Our study warrants further investigation of CRCNet by prospective clinical trials.

Frontiers of Robotic Colonoscopy: A Comprehensive Review of Robotic Colonoscopes and Technologies

Flexible colonoscopy remains the prime mean of screening for colorectal cancer (CRC) and the gold standard of all population-based screening pathways around the world. Almost 60% of CRC deaths could be prevented with screening. However, colonoscopy attendance rates are affected by discomfort, fear of pain and embarrassment or loss of control during the procedure. Moreover, the emergence and global thread of new communicable diseases might seriously affect the functioning of contemporary centres performing gastrointestinal endoscopy. Innovative solutions are needed: artificial intelligence (AI) and physical robotics will drastically contribute for the future of the healthcare services. The translation of robotic technologies from traditional surgery to minimally invasive endoscopic interventions is an emerging field, mainly challenged by the tough requirements for miniaturization. Pioneering approaches for robotic colonoscopy have been reported in the nineties, with the appearance of inchworm-like devices. Since then, robotic colonoscopes with assistive functionalities have become commercially available. Research prototypes promise enhanced accessibility and flexibility for future therapeutic interventions, even via autonomous or robotic-assisted agents, such as robotic capsules. Furthermore, the pairing of such endoscopic systems with AI-enabled image analysis and recognition methods promises enhanced diagnostic yield. By assembling a multidisciplinary team of engineers and endoscopists, the paper aims to provide a contemporary and highly-pictorial critical review for robotic colonoscopes, hence providing clinicians and researchers with a glimpse of the major changes and challenges that lie ahead.

AI-doscopist: a real-time deep-learning-based algorithm for localising polyps in colonoscopy videos with edge computing devices

We have designed a deep-learning model, an "Artificial Intelligent Endoscopist (a.k.a. AI-doscopist)", to localise colonic neoplasia during colonoscopy. This study aims to evaluate the agreement between endoscopists and AI-doscopist for colorectal neoplasm localisation. AI-doscopist was pre-trained by 1.2 million non-medical images and fine-tuned by 291,090 colonoscopy and non-medical images. The colonoscopy images were obtained from six databases, where the colonoscopy images were classified into 13 categories and the polyps' locations were marked image-by-image by the smallest bounding boxes. Seven categories of non-medical images, which were believed to share some common features with colorectal polyps, were downloaded from an online search engine. Written informed consent were obtained from 144 patients who underwent colonoscopy and their full colonoscopy videos were prospectively recorded for evaluation. A total of 128 suspicious lesions were resected or biopsied for histological confirmation. When evaluated image-by-image on the 144 full colonoscopies, the specificity of AI-doscopist was 93.3%. AI-doscopist were able to localise 124 out of 128 polyps (polyp-based sensitivity = 96.9%). Furthermore, after reviewing the suspected regions highlighted by AI-doscopist in a 102-patient cohort, an endoscopist has high confidence in recognizing four missed polyps in three patients who were not diagnosed with any lesion during their original colonoscopies. In summary, AI-doscopist can localise 96.9% of the polyps resected by the endoscopists. If AI-doscopist were to be used in real-time, it can potentially assist endoscopists in detecting one more patient with polyp in every 20-33 colonoscopies.

Review of computational methods for the detection and classification of polyps in colonoscopy imaging

Computer-aided diagnosis (CAD) is a tool with great potential to help endoscopists in the tasks of detecting and histologically classifying colorectal polyps. In recent years, different technologies have been described and their potential utility has been increasingly evidenced, which has generated great expectations among scientific societies. However, most of these works are retrospective and use images of different quality and characteristics which are analysed off line. This review aims to familiarise gastroenterologists with computational methods and the particularities of endoscopic imaging, which have an impact on image processing analysis. Finally, the publicly available image databases, needed to compare and confirm the results obtained with different methods, are presented.

The evolution of lower gastrointestinal endoscopy: where are we now?

Lower gastrointestinal endoscopy has evolved over time, fulfilling a widening diagnostic and therapeutic remit. As our understanding of colorectal cancer and its prevention has improved, endoscopy has progressed with improved diagnostic technologies and advancing endoscopic therapies. Despite this, the fundamental design of the endoscope has remained similar since its inception. This review presents the important role lower gastrointestinal endoscopy serves in the prevention of colorectal cancer and the desirable characteristics of the endoscope that would enhance this. A brief history of the endoscope is presented. Current and future robotic endoscopic platforms, which may fulfil these desirable characteristics, are discussed. The incorporation of new technologies from allied scientific disciplines will help the endoscope fulfil its maximum potential in preventing the increasing global burden of colorectal cancer. There are a number of endoscopic platforms under development, which show significant promise.

Application of Artificial Intelligence to Gastroenterology and Hepatology

Since 2010, substantial progress has been made in artificial intelligence (AI) and its application to medicine. AI is explored in gastroenterology for endoscopic analysis of lesions, in detection of cancer, and to facilitate the analysis of inflammatory lesions or gastrointestinal bleeding during wireless capsule endoscopy. AI is also tested to assess liver fibrosis and to differentiate patients with pancreatic cancer from those with pancreatitis. AI might also be used to establish prognoses of patients or predict their response to treatments, based on multiple factors. We review the ways in which AI may help physicians make a diagnosis or establish a prognosis and discuss its limitations, knowing that further randomized controlled studies will be required before the approval of AI techniques by the health authorities.

An overview of deep learning algorithms and water exchange in colonoscopy in improving adenoma detection

Introduction: Among the Gastrointestinal (GI) Endoscopy Editorial Board top 10 topics in advances in endoscopy in 2018, water exchange colonoscopy and artificial intelligence were both considered important advances. Artificial intelligence holds the potential to increase and water exchange significantly increases adenoma detection.Areas covered: The authors searched MEDLINE (1998-2019) using the following medical subject terms: water-aided, water-assisted and water exchange colonoscopy, adenoma, artificial intelligence, deep learning, computer-assisted detection, and neural networks. Additional related studies were manually searched from the reference lists of publications. Only fully published journal articles in English were reviewed. The latest date of the search was Aug10, 2019. Artificial intelligence, machine learning, and deep learning contribute to the promise of real-time computer-aided detection diagnosis. By emphasizing near-complete suction of infused water during insertion, water exchange provides salvage cleaning and decreases cleaning-related multi-tasking distractions during withdrawal, increasing adenoma detection. The review will address how artificial intelligence and water exchange can complement each other in improving adenoma detection during colonoscopy.Expert opinion: In 5 years, research on artificial intelligence will likely achieve real-time application and evaluation of factors contributing to quality colonoscopy. Better understanding and more widespread use of water exchange will be possible.

Gastrointestinal diagnosis using non-white light imaging capsule endoscopy

Capsule endoscopy (CE) has proved to be a powerful tool in the diagnosis and management of small bowel disorders since its introduction in 2001. However, white light imaging (WLI) is the principal technology used in clinical CE at present, and therefore, CE is limited to mucosal inspection, with diagnosis remaining reliant on visible manifestations of disease. The introduction of WLI CE has motivated a wide range of research to improve its diagnostic capabilities through integration with other sensing modalities. These developments have the potential to overcome the limitations of WLI through enhanced detection of subtle mucosal microlesions and submucosal and/or transmural pathology, providing novel diagnostic avenues. Other research aims to utilize a range of sensors to measure physiological parameters or to discover new biomarkers to improve the sensitivity, specificity and thus the clinical utility of CE. This multidisciplinary Review summarizes research into non-WLI CE devices by organizing them into a taxonomic structure on the basis of their sensing modality. The potential of these capsules to realize clinically useful virtual biopsy and computer-aided diagnosis (CADx) is also reported.