Masashi Misawa
Research Interests
Make your next move.
Prepare your application
Review drafts before you send them.
Explore related searches
Articles (8)
Objectives Early detection of colorectal cancer is critical for improving prognosis. However, assessing invasion depth—distinguishing between superficial cancer (T1a) and deep submucosal invasive cancer (T1b)—remains challenging. Recently, artificial intelligence (AI)‐assisted computer‐aided diagnosis (CADx) systems have been introduced to complement conventional endoscopy. This study aims to compare the diagnostic accuracy of endoscopists in predicting deep submucosal invasion in early colorectal cancer under four modalities: white‐light imaging (WLI), magnifying endoscopy (including narrow‐band imaging magnification and pit pattern), endocytoscopy (EC), and CADx support. Methods We conducted a single‐center retrospective study using stored endoscopic images between April 2021 and December 2022. Each lesion was evaluated using white light imaging, magnifying endoscopy, EC, and CADx analysis with the EndoBRAIN‐Plus system. Trainee and expert endoscopists assessed the images sequentially, recording their estimations of invasion depth (T1a vs. T1b) and confidence levels. Sensitivity, specificity, and accuracy were calculated against the pathological reference. We compared performances stratified by confidence level and endoscopist experience. Results During the study period, 66 lesions were eligible. Of them, 27% (18 lesions) were T1b cancers. Diagnostic accuracy improved progressively from white light imaging (82.7% [95% confidence interval {95%CI}: 81.2–86.9]) to EC (85.6% [95%CI: 82.7–88.2]). The highest specificity and accuracy were achieved when AI‐assisted diagnosis was incorporated (accuracy: 88.9% [95%CI: 86.3–91.2], specificity: 93.1% [95%CI: 90.6–95.2]). The proportion of high‐confidence readings rose from 40.2% to 75.5%. This was most pronounced in the trainee group. Conclusions Integrating advanced endoscopic imaging with CADx significantly improved accuracy in assessing invasion depth. This approach may guide treatment decisions in early‐stage colorectal cancer.
Year:
2025
The environmental impact of computer-aided diagnosis (CADx) in colonoscopy is unknown. The potential to reduce greenhouse gas emissions by limiting unnecessary polypectomies may be offset by emissions associated with CADx devices. We quantified and compared the carbon footprint of colonoscopy with and without CADx. Life-cycle carbon emissions associated with endoscopic snares, polyp traps, histopathology, and CADx devices were estimated using life-cycle assessment software. The effect of CADx was estimated by the carbon dioxide equivalent (CO2e) emissions related to outcomes from a prospective cohort of patients who underwent colonoscopy with optical diagnosis under a leave-in-situ strategy (i.e. rectosigmoid polyps ≤5 mm predicted as non-neoplastic were not removed) with and without CADx assistance. We analyzed 1134 individuals (median age 67 years, 59% male) with 1716 colorectal polyps. Colonoscopy with CADx reduced the number of snares and polyp traps from 567 to 543 (difference: 24) and histopathology submissions from 1343 to 1278 (difference: 65). However, the total carbon footprint increased from 832 kg CO2e (95%CI 815 to 847) without CADx to 946 kg CO2e (95%CI 927 to 965) with CADx, corresponding to an absolute increase of 114 kg CO2e (95%CI 80 to 150) or 0.10 kg CO2e per colonoscopy. Using CADx under a leave-in-situ strategy increased the environmental burden of colonoscopy.
Year:
2025
Computer-aided detection (CADe) systems have improved the adenoma detection rate (ADR); however, concerns about its long-term effect on endoscopists’ performance without CADe and potential deskilling remain unaddressed. This study evaluated the impact of CADe on the learning curve of endoscopists for adenoma detection. This propensity score-matching, prospective, single-center, observational study was conducted from January 2021 to December 2023. CADe systems were installed in half of the endoscopy units, and patients were equally distributed between the rooms. Patients aged ≥20 years scheduled for colonoscopy were included, excluding those with polyposis, inflammatory bowel disease, known polyps, incomplete colonoscopy, emergency cases, or previous colorectal surgery, and those examined by novices. Endoscopists were classified as high detectors (ADR ≥25%) or low detectors (ADR <25%) based on the ADR recorded before CADe implementation. To assess skill acquisition and transfer, the primary outcome was the change in ADR over time, as measured by cumulative summation (CUSUM) analysis, in both CADe and non-CADe procedures. Of 18 962 patients who underwent colonoscopy, 13 245 patients were excluded, and of the 5717 patients initially enrolled, 4712 (CADe group, n = 2356; non-CADe group, n = 2356) were analyzed after propensity score matching. CUSUM analysis showed that both high and low detectors achieved enhanced detection performance for CADe procedures. Among non-CADe procedures, high detectors had accelerated learning curves, indicating they maintained a higher ADR, whereas low detectors showed no significant change in their learning trajectory. After CADe implementation, the detection rate in procedures performed without CADe was maintained and did not decline over time.
Year:
2025
Inflammatory bowel disease (IBD), including Crohn's disease and ulcerative colitis, presents substantial diagnostic and management challenges because of its variable clinical course and the limitations of conventional endoscopy. Although endoscopic procedures are crucial for diagnosis and surveillance, their inherent subjectivity and inter‐observer variability complicate disease assessment. Recent advances in artificial intelligence (AI) offer promising solutions to these challenges by enabling automated, precise, and objective image analysis. AI technologies have demonstrated success in diagnosing IBD, distinguishing it from other gastrointestinal disorders, and facilitating early identification of neoplasia in IBD patients, improving clinical decision‐making and potentially reducing the need for invasive procedures. Furthermore, AI applications for evaluating endoscopic images have enhanced the accuracy of disease severity assessments such as the Mayo Endoscopic Score and Ulcerative Colitis Endoscopic Index of Severity by overcoming issues related to observer variability. Integration of AI with advanced endoscopic technologies, including image‐enhanced and magnified endoscopy, further improves lesion characterization and offers insights into mucosal healing, which is crucial for optimizing treatment. While AI's potential in IBD management is substantial, challenges remain in its clinical implementation, necessitating further validation through real‐world data and regulatory approval. This review explores the evolving role of AI in transforming IBD diagnosis, surveillance, and assessment, with a focus on enhancing patient care through improved precision and efficiency.
Year:
2025
BACKGROUND Accurate classification of colorectal polyps during colonoscopy is critical for determining individual cancer risk, guiding treatment, and optimising surveillance, thereby reducing healthcare costs and patient burden. Despite advances in optical diagnosis—such as virtual chromoendoscopy and magnification—its adoption remains limited due to skill requirements and suboptimal accuracy. Current computer-assisted diagnosis (CADx) tools primarily offering binary (neoplastic vs. non-neoplastic) classification of colorectal polyps. OBJECTIVE To address these limitations, we propose developing a multiclass AI-based CADx system capable of real-time, five-class classification of colorectal lesions, including non-neoplastic lesions, low- and high-grade adenomas, sessile serrated lesions (SSLs), and invasive cancers. METHODS This multicentre, international study comprises two phases: development and validation. During the development phase, we will collect and curate a comprehensive dataset of over 17,000 de-identified endoscopic images and videos, sourced prospectively from eight European hospitals and one centre in Japan. The dataset encompasses diverse ethnicities and lesion types, facilitating robust AI training. A semi-supervised learning approach utilising contrastive learning (SimCLR) will be employed to optimise the model’s discriminative capability. Labelling of images will involve consensus from experienced endoscopists, while unlabelled data will be leveraged through contrastive pre-training, enhancing feature extraction even with limited annotations. An integrated CADe component will detect lesions, with heatmaps providing interpretability via Grad-CAM to highlight relevant image features. The proof-of-concept validation will involve 250 unseen lesion images scored by expert endoscopists, with a target accuracy threshold of 85%. Primary outcomes include sensitivity and specificity for cancer prediction, alongside secondary measures such as interobserver agreement, diagnostic accuracy across demographic subgroups, and qualitative analyses of misclassifications. Data management will adhere to strict de-identification protocols, housed within a secure REDCap database, with an open-access resource generated for future research. RESULTS Ethics approval has been secured or is pending across participating sites, with informed consent obtained prospectively and opt-out options available retrospectively. CONCLUSIONS The proposed outcome of this project is a real-time AI tool that provides detailed lesion classification, surpassing current binary systems and supporting personalised endoscopic decision-making. This pioneering effort promises to advance CADx utility in routine practice, ultimately contributing to improved colorectal cancer prevention and patient care worldwide. CLINICALTRIAL
Year:
2025
Background Computer-aided diagnosis (CADx) enables the distinction between neoplastic and non-neoplastic polyps during colonoscopy. We aimed to estimate the patient-level benefit and harm of CADx. Methods We conducted a comparative analysis on data from the EndoBRAIN international clinical trial, evaluating the effect of optical diagnosis during colonoscopy with and without CADx. Three hypothetical scenarios were compared: “endoscopist-alone” and “CADx-assisted” leave-in-situ strategies (leaving non-neoplastic rectosigmoid polyps ≤ 5 mm), and “total removal” (removing all detected polyps). Primary outcomes included patient-level colonoscopy-related cost and surveillance interval agreement (colorectal cancer risk category). Estimates were calculated based on national reimbursement rates and guidelines in four countries. Results We analyzed 1134 patients (59 % men, median age 67 years) with 1716 polyps. Compared with total removal, the endoscopist-alone and CADx-assisted leave-in-situ strategies reduced the removed polyps per patient from 1.51 (95 %CI 1.48–1.54) to 1.18 (95 %CI 1.16–1.20) and 1.12 (95 %CI 1.00–1.14), respectively; however, 0.023 (95 %CI 0.015–0.033) and 0.021 (95 %CI 0.014–0.031) neoplasms per patient were left in situ, respectively. The mean colonoscopy cost decreased by $44 (endoscopist alone) and $46 (CADx assistance) in the USA, $22 and $19 in the UK, $21 and $19 in Japan, and $32 and $30 in Norway, respectively. Surveillance interval agreement decreased to 99.2 % (endoscopist alone) and 99.0 % (CADx assistance) in the USA, 99.8 % and 99.8 % in the UK, 97.9 % and 97.1 % in Japan, and 99.9 % and 99.9 % in Norway, respectively. Conclusions Both endoscopist-alone and CADx-assisted optical diagnosis reduce colonoscopy costs. The risk of missed adenomas and surveillance interval deviations appear marginal.
Year:
2025
Objectives To develop a procedure that matches magnifying endoscopic images with narrow‐band imaging to 3D tissue structures using a tissue‐clearing technique and to qualitatively and quantitatively analyze specified structures in gastrointestinal neoplasms. Methods Endoscopically resected formalin‐fixed paraffin‐embedded gastrointestinal tissues (three esophagus, four stomach, seven colon) were made transparent by ethyl cinnamate. They were then subjected to fluorescent staining of nuclei and blood vessels followed by 3D imaging using a confocal laser scanning microscope. A one‐to‐one correspondence between magnifying endoscopic and 3D reconstructed images was established using vessels and crypts with characteristic shapes as guides, and the depth and caliber of specified vessels were measured. Results All tissues were optically cleared, which allowed 3D visualization of vascular structures and nuclei in all layers. In the esophagus, intraepithelial papillary capillary loops and subepithelial capillary networks were identified. In the upper part of the stomach, polygonal subepithelial capillary loops surrounding the pits were observed, while in the lower part, surface epithelium with ridge‐like structures and coiled vessels were observed. A honeycomb pit structure and surrounding vascular structures were identified in the colon. Quantitative analysis showed the various contrasts of a single continuous vessel in the endoscopic image were due to different depths at which the vessel tortuously ran. Conclusion We established a procedure to allow one‐to‐one correspondence between magnifying endoscopic and 3D reconstructed images and to measure the depth and caliber of endoscopically visualized vessels of interest. This method is expected to improve endoscopic diagnosis and further the development of endoscopic imaging technologies.
Year:
2025
Objectives Lymph node metastasis (LNM) prediction for T1 colorectal cancer (CRC) is critical for determining the need for surgery after endoscopic resection because LNM occurs in 10%. We aimed to develop a novel artificial intelligence (AI) system using whole slide images (WSIs) to predict LNM. Methods We conducted a retrospective single center study. To train and test the AI model, we included LNM status‐confirmed T1 and T2 CRC between April 2001 and October 2021. These lesions were divided into two cohorts: training (T1 and T2) and testing (T1). WSIs were cropped into small patches and clustered by unsupervised K‐means. The percentage of patches belonging to each cluster was calculated from each WSI. Each cluster's percentage, sex, and tumor location were extracted and learned using the random forest algorithm. We calculated the areas under the receiver operating characteristic curves (AUCs) to identify the LNM and the rate of over‐surgery of the AI model and the guidelines. Results The training cohort contained 217 T1 and 268 T2 CRCs, while 100 T1 cases (LNM‐positivity 15%) were the test cohort. The AUC of the AI system for the test cohort was 0.74 (95% confidence interval [CI] 0.58–0.86), and 0.52 (95% CI 0.50–0.55) using the guidelines criteria ( P = 0.0028). This AI model could reduce the 21% of over‐surgery compared to the guidelines. Conclusion We developed a pathologist‐independent predictive model for LNM in T1 CRC using WSI for determination of the need for surgery after endoscopic resection. Trial registration UMIN Clinical Trials Registry (UMIN000046992, https://center6.umin.ac.jp/cgi-open-bin/ctr/ctr_view.cgi?recptno=R000053590 ).
Year:
2023
Collaborators (6)
Katsuro Ichimasa
Yokohama City University Medical Center
cesare hassan
Humanitas University
Yuichi Mori
University of Oslo
Marietta Iacucci
University of Birmingham
Bu'Hussain Hayee
Professor in Gastroenterology
King’s College London
Kento Takenaka
Tokyo Medical and Dental University
Make your next move.
Prepare your application
Review drafts before you send them.