Artificial Intelligence to Predict the Risk of Lymph Node Metastasis in T2 Colorectal Cancer.

OBJECTIVE: To develop and externally validate an updated artificial intelligence (AI) prediction system for stratifying the risk of lymph node metastasis (LNM) in T2 colorectal cancer (CRC). SUMMARY BACKGROUND DATA: Recent technical advances allow complete local excision of T2 CRC, traditionally treated with surgical resection. Yet, the widespread adoption of this approach is hampered by the inability to stratify the risk of LNM. METHODS: Data from pT2 CRC patients undergoing surgical resection between April 2000 and May 2022 at one Japanese and one Italian center were analyzed. Primary goal was AI system development for accurate LNM prediction. Predictors encompassed seven variables: age, sex, tumor size and location, lympho-vascular invasion, histological differentiation, and carcinoembryonic antigen level. The tool's discriminating power was assessed via Area Under the Curve (AUC), sensitivity, and specificity. RESULTS: Out of 735 initial patients, 692 were eligible. Training and validation cohorts comprised of 492 and 200 patients, respectively. The AI model displayed an AUC of 0.75 in the combined validation dataset. Sensitivity for LNM prediction was 97.8% and specificity was 15.6%. The Positive and the Negative Predictive Value were 25.7% and 96% respectively. The False Negative (FN) rate was 2.2%, the False Positive was 84.4%. CONCLUSIONS: Our AI model, based on easily accessible clinical and pathological variables, moderately predicts LNM in T2 CRC. However, the risk of FN needs to be considered. The training of the model including more patients across Western and Eastern centers -differentiating between colon and rectal cancers- may improve its performance and accuracy.

Accuracy Goals in Predicting Preoperative Lymph Node Metastasis for T1 Colorectal Cancer Resected Endoscopically.

Submucosal invasive (T1) colorectal cancer is a significant clinical management challenge, with an estimated 10% of patients developing extraintestinal lymph node metastasis. This condition necessitates surgical resection along with lymph node dissection to achieve a curative outcome. Thus, the precise preoperative assessment of lymph node metastasis risk is crucial to guide treatment decisions after endoscopic resection. Contemporary clinical guidelines strive to identify a low-risk cohort for whom endoscopic resection will suffice, applying stringent criteria to maximize patient safety. Those failing to meet these criteria are often recommended for surgical resection, with its associated mortality risks although it may still include patients with a low risk of metastasis. In the quest to enhance the precision of preoperative lymph node metastasis risk prediction, innovative models leveraging artificial intelligence or nomograms are being developed. Nevertheless, the debate over the ideal sensitivity and specificity for such models persists, with no consensus on target metrics. This review puts forth postoperative mortality rates as a practical benchmark for the sensitivity of predictive models. We underscore the importance of this method and advocate for research to amass data on surgical mortality in T1 colorectal cancer. Establishing specific benchmarks for predictive accuracy in lymph node metastasis risk assessment will hopefully optimize the treatment of T1 colorectal cancer.

Artificial intelligence-assisted video colonoscopy for disease monitoring of ulcerative colitis: A prospective study.

BACKGROUNDS AND AIMS: The Mayo endoscopic subscore (MES) is the most popular endoscopic disease activity measure of ulcerative colitis (UC). Artificial intelligence (AI)-assisted colonoscopy is expected to reduce diagnostic variability among endoscopists. However, no study has been conducted to ascertain whether AI-based MES assignments can help predict clinical relapse, nor has AI been verified to improve the diagnostic performance of non-specialists. METHODS: This open-label, prospective cohort study enrolled 110 patients with UC in clinical remission. The AI algorithm was developed using 74713 images from 898 patients who underwent colonoscopy at three centers. Patients were followed up after colonoscopy for 12 months, and clinical relapse was defined as a partial Mayo score >2. A multi-video, multi-reader analysis involving 124 videos was conducted to determine whether the AI system reduced the diagnostic variability among six non-specialists. RESULTS: The clinical relapse rate for patients with AI-based MES = 1 (24.5% [12/49]) was significantly higher (log-rank test, P = 0.01) than that for patients with AI-based MES = 0 (3.2% [1/31]). Relapse occurred during the 12-month follow-up period in 16.2% (13/80) of patients with AI-based MES = 0 or 1 and 50.0% (10/20) of those with AI-based MES = 2 or 3 (log-rank test, P = 0.03). Using AI resulted in better inter- and intra-observer reproducibility than endoscopists alone. CONCLUSIONS: Colonoscopy using the AI-based MES system can stratify the risk of clinical relapse in patients with UC and improve the diagnostic performance of non-specialists.

Risk factors for lymph node metastasis in T2 colorectal cancer: a systematic review and meta-analysis.

BACKGROUND: Lymph node metastasis (LNM) occurs in 20-25% of patients with T2 colorectal cancer (CRC). Identification of risk factors for LNM in T2 CRC may help identify patients who are at low risk and thereby potential candidates for endoscopic full-thickness resection. We examined risk factors for LNM in T2 CRC with the goal of establishing further criteria of the indications for endoscopic resection. METHODS: MEDLINE, CENTRAL, and EMBASE were systematically searched from inception to November 2023. Studies that investigated the association between the presence of LNM and the clinical and pathological factors of T2 CRC were included. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. Certainty of evidence (CoE) was assessed using the GRADE approach. RESULTS: Fourteen studies (8349 patients) were included. Overall, the proportion of LNM was 22%. The meta-analysis revealed that the presence of lymphovascular invasion (OR, 5.5; 95% CI 3.7-8.3; high CoE), high-grade tumor budding (OR, 2.4; 95% CI 1.5-3.7; moderate CoE), poor differentiation (OR, 2.2; 95% CI 1.8-2.7; moderate CoE), and female sex (OR, 1.3; 95% CI 1.1-1.7; high CoE) were associated with LNM in T2 CRC. Lymphatic invasion (OR, 5.0; 95% CI 3.3-7.6) was a stronger predictor of LNM than vascular invasion (OR, 2.4; 95% CI 2.1-2.8). CONCLUSIONS: Lymphovascular invasion, high-grade tumor budding, poor differentiation, and female sex were risk factors for LNM in T2 CRC. Endoscopic resection of T2 CRC in patients with very low risk for LNM may become an alternative to conventional surgical resection. TRIAL REGISTRATION: PROSPERO, CRD42022316545.

Role of the artificial intelligence in the management of T1 colorectal cancer.

Approximately 10% of submucosal invasive (T1) colorectal cancers demonstrate extraintestinal lymph node metastasis, necessitating surgical intervention with lymph node dissection. The ability to identify T1b (submucosal invasion depth ≥ 1000 µm) as a risk factor for lymph node metastasis via pre-treatment endoscopy is crucial in guiding treatment strategies. Accurately distinguishing T1b from T1a (submucosal invasion depth < 1000 µm) or dysplasia remains a significant challenge for artificial intelligence (AI) systems, which require high and consistent diagnostic capabilities. Moreover, as endoscopic therapies like endoscopic full-thickness resection and endoscopic intermuscular dissection evolve, and the focus on reducing unnecessary surgeries intensifies, the initial management of T1 colorectal cancers via endoscopic treatment is anticipated to increase. Consequently, the development of highly accurate and reliable AI systems is essential, not only for pre-treatment depth assessment but also for post-treatment risk stratification of lymph node metastasis. While such AI diagnostic systems are still under development, significant advancements are expected in the near future to improve decision-making in T1 colorectal cancer management.

Diagnostic Accuracy of Highest-Grade or Predominant Histological Differentiation of T1 Colorectal Cancer in Predicting Lymph Node Metastasis: A Systematic Review and Meta-Analysis.

INTRODUCTION: Treatment guidelines for colorectal cancer (CRC) suggest 2 classifications for histological differentiation-highest grade and predominant. However, the optimal predictor of lymph node metastasis (LNM) in T1 CRC remains unknown. This systematic review aimed to evaluate the impact of the use of highest-grade or predominant differentiation on LNM determination in T1 CRC. METHODS: The study protocol is registered in the International Prospective Register of Systematic Reviews (PROSPERO, registration number: CRD42023416971) and was published in OSF ( https://osf.io/TMAUN/ ) on April 13, 2023. We searched 5 electronic databases for studies assessing the diagnostic accuracy of highest-grade or predominant differentiation to determine LNM in T1 CRC. The outcomes were sensitivity and specificity. We simulated 100 cases with T1 CRC, with an LNM incidence of 11.2%, to calculate the differences in false positives and negatives between the highest-grade and predominant differentiations using a bootstrap method. RESULTS: In 42 studies involving 41,290 patients, the differentiation classification had a pooled sensitivity of 0.18 (95% confidence interval [CI] 0.13-0.24) and 0.06 (95% CI 0.04-0.09) ( P < 0.0001) and specificity of 0.95 (95% CI 0.93-0.96) and 0.98 (95% CI 0.97-0.99) ( P < 0.0001) for the highest-grade and predominant differentiations, respectively. In the simulation, the differences in false positives and negatives between the highest-grade and predominant differentiations were 3.0% (range 1.6-4.4) and -1.3% (range -2.0 to -0.7), respectively. DISCUSSION: Highest-grade differentiation may reduce the risk of misclassifying cases with LNM as negative, whereas predominant differentiation may prevent unnecessary surgeries. Further studies should examine differentiation classification using other predictive factors.

A novel artificial intelligence-assisted "vascular healing" diagnosis for prediction of future clinical relapse in patients with ulcerative colitis: a prospective cohort study (with video).

BACKGROUND AND AIMS: Image-enhanced endoscopy has attracted attention as a method for detecting inflammation and predicting outcomes in patients with ulcerative colitis (UC); however, the procedure requires specialist endoscopists. Artificial intelligence (AI)-assisted image-enhanced endoscopy may help nonexperts provide objective accurate predictions with the use of optical imaging. We aimed to develop a novel AI-based system using 8853 images from 167 patients with UC to diagnose "vascular-healing" and establish the role of AI-based vascular-healing for predicting the outcomes of patients with UC. METHODS: This open-label prospective cohort study analyzed data for 104 patients with UC in clinical remission. Endoscopists performed colonoscopy using the AI system, which identified the target mucosa as AI-based vascular-active or vascular-healing. Mayo endoscopic subscore (MES), AI outputs, and histologic assessment were recorded for 6 colorectal segments from each patient. Patients were followed up for 12 months. Clinical relapse was defined as a partial Mayo score >2 RESULTS: The clinical relapse rate was significantly higher in the AI-based vascular-active group (23.9% [16/67]) compared with the AI-based vascular-healing group (3.0% [1/33)]; P = .01). In a subanalysis predicting clinical relapse in patients with MES ≤1, the area under the receiver operating characteristic curve for the combination of complete endoscopic remission and vascular healing (0.70) was increased compared with that for complete endoscopic remission alone (0.65). CONCLUSIONS: AI-based vascular-healing diagnosis system may potentially be used to provide more confidence to physicians to accurately identify patients in remission of UC who would likely relapse rather than remain stable.

Performance evaluation of a computer-aided polyp detection system with artificial intelligence for colonoscopy.

OBJECTIVES: A computer-aided detection (CAD) system was developed to support the detection of colorectal lesions by deep learning using video images of lesions and normal mucosa recorded during colonoscopy. The study's purpose was to evaluate the stand-alone performance of this device under blinded conditions. METHODS: This multicenter prospective observational study was conducted at four Japanese institutions. We used 326 videos of colonoscopies recorded with patient consent at institutions in which the Ethics Committees approved the study. The sensitivity of successful detection of the CAD system was calculated using the target lesions, which were detected by adjudicators from two facilities for each lesion appearance frame; inconsistencies were settled by consensus. Successful detection was defined as display of the detection flag on the lesion for more than 0.5 s within 3 s of appearance. RESULTS: Of the 556 target lesions from 185 cases, detection success sensitivity was 97.5% (95% confidence interval [CI] 95.8-98.5%). The "successful detection sensitivity per colonoscopy" was 93% (95% CI 88.3-95.8%). For the frame-based sensitivity, specificity, positive predictive value, and negative predictive value were 86.6% (95% CI 84.8-88.4%), 84.7% (95% CI 83.8-85.6%), 34.9% (95% CI 32.3-37.4%), and 98.2% (95% CI 97.8-98.5%), respectively. TRIAL REGISTRATION: University Hospital Medical Information Network (UMIN000044622).

Use of artificial intelligence in the management of T1 colorectal cancer: a new tool in the arsenal or is deep learning out of its depth?

The field of artificial intelligence is rapidly evolving, and there has been an interest in its use to predict the risk of lymph node metastasis in T1 colorectal cancer. Accurately predicting lymph node invasion may result in fewer patients undergoing unnecessary surgeries; conversely, inadequate assessments will result in suboptimal oncological outcomes. This narrative review aims to summarize the current literature on deep learning for predicting the probability of lymph node metastasis in T1 colorectal cancer, highlighting areas of potential application and barriers that may limit its generalizability and clinical utility.

Impact of computer-aided characterization for diagnosis of colorectal lesions, including sessile serrated lesions: Multireader, multicase study.

OBJECTIVES: Computer-aided characterization (CADx) may be used to implement optical biopsy strategies into colonoscopy practice; however, its impact on endoscopic diagnosis remains unknown. We aimed to evaluate the additional diagnostic value of CADx when used by endoscopists for assessing colorectal polyps. METHODS: This was a single-center, multicase, multireader, image-reading study using randomly extracted images of pathologically confirmed polyps resected between July 2021 and January 2022. Approved CADx that could predict two-tier classification (neoplastic or nonneoplastic) by analyzing narrow-band images of the polyps was used to obtain a CADx diagnosis. Participating endoscopists determined if the polyps were neoplastic or not and noted their confidence level using a computer-based, image-reading test. The test was conducted twice with a 4-week interval: the first test was conducted without CADx prediction and the second test with CADx prediction. Diagnostic performances for neoplasms were calculated using the pathological diagnosis as reference and performances with and without CADx prediction were compared. RESULTS: Five hundred polyps were randomly extracted from 385 patients and diagnosed by 14 endoscopists (including seven experts). The sensitivity for neoplasia was significantly improved by referring to CADx (89.4% vs. 95.6%). CADx also had incremental effects on the negative predictive value (69.3% vs. 84.3%), overall accuracy (87.2% vs. 91.8%), and high-confidence diagnosis rate (77.4% vs. 85.8%). However, there was no significant difference in specificity (80.1% vs. 78.9%). CONCLUSIONS: Computer-aided characterization has added diagnostic value for differentiating colorectal neoplasms and may improve the high-confidence diagnosis rate.

Challenges in Implementing Endoscopic Resection for T2 Colorectal Cancer.

The current standard treatment for muscularis propria-invasive (T2) colorectal cancer is surgical colectomy with lymph node dissection. With the advent of new endoscopic resection techniques, such as endoscopic full-thickness resection or endoscopic intermuscular dissection, T2 colorectal cancer, with metastasis to 20%-25% of the dissected lymph nodes, may be the next candidate for endoscopic resection following submucosal-invasive (T1) colorectal cancer. We present a novel endoscopic treatment strategy for T2 colorectal cancer and suggest further study to establish evidence on oncologic and endoscopic technical safety for its clinical implementation.

Differentiation grade as a risk factor for lymph node metastasis in T1 colorectal cancer.

Objectives: Japanese guidelines include high-grade (poorly differentiated) tumors as a risk factor for lymph node metastasis (LNM) in T1 colorectal cancer (CRC). However, whether the grading is based on the least or most predominant component when the lesion consists of two or more levels of differentiation varies among institutions. This study aimed to investigate which method is optimal for assessing the risk of LNM in T1 CRC. Methods: We retrospectively evaluated 971 consecutive patients with T1 CRC who underwent initial or additional surgical resection from 2001 to 2021 at our institution. Tumor grading was divided into low-grade (well- to moderately differentiated) and high-grade based on the least or predominant differentiation analyses. We investigated the correlations between LNM and these two grading analyses. Results: LNM was present in 9.8% of patients. High-grade tumors, as determined by least differentiation analysis, accounted for 17.0%, compared to 0.8% identified by predominant differentiation analysis. A significant association with LNM was noted for the least differentiation method (p < 0.05), while no such association was found for predominant differentiation (p = 0.18). In multivariate logistic regression, grading based on least differentiation was an independent predictor of LNM (p = 0.04, odds ratio 1.68, 95% confidence interval 1.00-2.83). Sensitivity and specificity for detecting LNM were 27.4% and 84.1% for least differentiation, and 2.1% and 99.3% for predominant differentiation, respectively. Conclusions: Tumor grading via least differentiation analysis proved to be a more reliable measure for assessing LNM risk in T1 CRC compared to grading by predominant differentiation.

Additional staining for lymphovascular invasion is associated with increased estimation of lymph node metastasis in patients with T1 colorectal cancer: Systematic review and meta-analysis.

OBJECTIVES: Lymphovascular invasion (LVI) is a critical risk factor for lymph node metastasis (LNM), which requires additional surgery after endoscopic resection of T1 colorectal cancer (CRC). However, the impact of additional staining on estimating LNM is unclear. This systematic review aimed to evaluate the impact of additional staining on determining LNM in T1 CRC. METHODS: We searched five electronic databases. Outcomes were diagnostic odds ratio (DOR), assessed using hierarchical summary receiver operating characteristic curves, and interobserver agreement among pathologists for positive LVI, assessed using Kappa coefficients (κ). We performed a subgroup analysis of studies that simultaneously included a multivariable analysis for other risk factors (deep submucosal invasion, poor differentiation, and tumor budding). RESULTS: Among the 64 studies (18,097 patients) identified, hematoxylin-eosin (HE) and additional staining for LVI had pooled sensitivities of 0.45 (95% confidence interval [CI] 0.32-0.58) and 0.68 (95% CI 0.44-0.86), specificities of 0.88 (95% CI 0.78-0.94) and 0.76 (95% CI 0.62-0.86), and DORs of 6.26 (95% CI 3.73-10.53) and 6.47 (95% CI 3.40-12.32) for determining LNM, respectively. In multivariable analysis, the DOR of additional staining for LNM (DOR 5.95; 95% CI 2.87-12.33) was higher than that of HE staining (DOR 1.89; 95% CI 1.13-3.16) (P = 0.01). Pooled κ values were 0.37 (95% CI 0.22-0.52) and 0.62 (95% CI 0.04-0.99) for HE and additional staining for LVI, respectively. CONCLUSION: Additional staining for LVI may increase the DOR for LNM and interobserver agreement for positive LVI among pathologists.

Whole slide image-based prediction of lymph node metastasis in T1 colorectal cancer using unsupervised artificial intelligence.

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).

Molecular and clinicopathological differences between depressed and protruded T2 colorectal cancer.

BACKGROUND: Colorectal cancer (CRC) can be classified into four consensus molecular subtypes (CMS) according to genomic aberrations and gene expression profiles. CMS is expected to be useful in predicting prognosis and selecting chemotherapy regimens. However, there are still no reports on the relationship between the morphology and CMS. METHODS: This retrospective study included 55 subjects with T2 CRC undergoing surgical resection, of whom 30 had the depressed type and 25 the protruded type. In the classification of the CMS, we first defined cases with deficient mismatch repair as CMS1. And then, CMS2/3 and CMS4 were classified using an online classifier developed by Trinh et al. The staining intensity of CDX2, HTR2B, FRMD6, ZEB1, and KER and the percentage contents of CDX2, FRMD6, and KER are input into the classifier to obtain automatic output classifying the specimen as CMS2/3 or CMS4. RESULTS: According to the results yielded by the online classifier, of the 30 depressed-type cases, 15 (50%) were classified as CMS2/3 and 15 (50%) as CMS4. Of the 25 protruded-type cases, 3 (12%) were classified as CMS1 and 22 (88%) as CMS2/3. All of the T2 CRCs classified as CMS4 were depressed CRCs. More malignant pathological findings such as lymphatic invasion were associated with the depressed rather than protruded T2 CRC cases. CONCLUSIONS: Depressed-type T2 CRC had a significant association with CMS4, showing more malignant pathological findings such as lymphatic invasion than the protruded-type, which could explain the reported association between CMS4 CRC and poor prognosis.