DIFFERENCES IN REGIONS-OF-INTEREST TO IDENTIFY DEEPLY INVASIVE COLORECTAL CANCERS: COMPUTER-AIDED DIAGNOSIS COMPARED WITH ENDOSCOPISTS

INTRODUCTION: Colonoscopy reduces colorectal cancer mortality via the identification and removal of neoplastic polyps. In clinical trials, computer aided detection (CADe) improves polyp detection, but there is limited data of CADe implementation in routine practice. We aimed to assess the impact of CADe upon polyp detection in a large cohort of high-volume colonoscopists.

METHODS: A CADe system (GI Genius, Medtronic) system was implemented in staggered fashion in a single large academic medical center to pragmatically assess its impact over a 6-month period (March 2022 to August 2022). Four CADe units were placed in a twelve-room endoscopy unit where colonoscopists rotate through different rooms. Thus, a colonoscopist may be able to utilize CADe when performing colonoscopy on one day (“CADe room”) but perform colonoscopy in a room without CADe the next day (“non-CADe room”). Colonoscopists who performed at least 100 colonoscopies over the 6-month period were included in this analysis. Colonoscopists were encouraged but not mandated to utilize CADe. The primary outcome was screening and surveillance colonoscopy polypectomy rate. Secondary outcomes were screening colonoscopy adenoma detection rate (ADR) and serrated detection rate (SDR). Results were further stratified by self-reported utilization of CADe: CADe majority users (self-reported use in > 50% of cases) and CADe minority users (self-reported use in < 50% of cases).

RESULTS: Over the 6-month study period, 21 colonoscopists performed 4,820 colonoscopies (Screening: 2,459, Surveillance: 1,472, and Diagnostic: 889). Of 21 colonoscopists, 9 were CADe majority users. Screening and surveillance polypectomy rates significantly increased in CADe rooms compared to non-CADe rooms (60.5% versus 51.7%, p<0.0001; Table). When stratified by CADe use, CADe majority users had a significant increase in polypectomy rate in CADe compared to non-CADe rooms (66.5% versus 53.4%, p<0.0001); in contrast, CADe minority users did not have a significant increase in polypectomy in CADe compared to non-CADe rooms (54.3% versus 50.4%, p=0.2).
When CADe was available, screening colonoscopy ADR (50.6% versus 41.6%, p<0.0002) and SDR (19.4% versus 14.7%, P=0.006) significantly increased. However, as expected, this significant increase in ADR and SDR was only seen in CADe majority users but not minority users.

DISCUSSION: In this pragmatic assessment of the impact of CADe upon colonoscopy quality, CADe significantly increased polypectomy rates for both screening and surveillance colonoscopy as well as screening colonoscopy ADR and SDR. As the impact of CADe is somewhat blunted by only half of colonoscopists using CADe in a majority of cases, further work is needed to improve CADe utilization in practice.

ACKNOWLEDGEMENTS: Nives and Joseph Rizza and the Digestive Health Foundation for their generous gifts to support AI research.

Aims. Since sessile serrated lesions (SSL) were introduced, multiple AI studies have tried to classify them alongside hyperplastic and adenomatous polyps, with varying results. A major hurdle for the classification of SSL is the low prevalence and difficult endoscopic recognition. Furthermore, the classification based on histology can also be subjective, as demonstrated by the inter-rater variability amongst pathologists. This study aims to improve the baseline training method for classification by training multiple models on subtasks before taking an ensemble vote for final classification.

Methods. 784 unique polyps (24% hyperplastic, 69% adenomas and 7% sessile serrated polyps) were recorded in different endoscopic imaging modalities as white light, blue light imaging and linked-color imaging. The ground truth was based on the histology of the polyp, assessed as hyperplastic (hyp), adenoma (adn) or SSL. The videos containing on average 125 frames were split into training, validation and test sets without overlapping patients to remove any possible data contamination. Subtasks such as 1-vs-all and 1-vs-1 strategies were trained on each of the class combinations. The outputs of the subtasks were used to vote the outcome using different combinations of the subtasks. The results were compared to a model that directly classified the three classes (hyp-vs-adn-vs-ssl).

Results. The averaged frame-based accuracy, sensitivity and positive predictive value (PPV) per class are shown in the tables. Table 1 shows the mean and standard deviation over the three classes for a selection of ensemble model. Table 2 shows the results per class for all ensemble models. An improvement can be seen comparing the results of the ensemble models with the baseline (table 1). There is a large improvement of both sensitivity and PPV for SSL (see table 2). there are slight variations in the metrics per class depending on the choice of ensemble.

Conclusions. The proposed method for ensemble voting is a valid approach for improving results for characterizing sessile serrated lesions with AI. The ensemble models have similar metrics, with slight variations depending on the choice of subtasks.

Backgrounds and aims: The optical distinction between T1a colorectal cancer (CRC) and T1b (invasion depth >1000 μm) CRC is crucial in determining subsequent treatment, but the diagnosis is difficult. We developed and reported on a computer-aided optical diagnosis (CAD) system for the diagnosis of CRC with deep invasion, using only white light imaging (WLI). This study evaluated the optical diagnostic accuracy of T1b CRC in T1 CRC by the CAD system, magnifying narrow-band imaging endoscopy (M-NBI), and dye-based magnifying chromoendoscopy (MCE).
Methods: Among 1173 lesions registered in a multicenter prospective study (NBI-CV study), 365 images of 37 T1 CRCs (16 T1a CRCs and 21 T1b CRCs) registered at our hospital were included. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy rate of the CAD system for diagnosing T1b CRC were calculated, and compared with those of M-NBI and MCE. Optical diagnoses by M-NBI and MCE were performed by experts in real time.
Results: The sensitivity, specificity, PPV, NPV, and accuracy rate of CAD were 66.9%, 76.7%, 85.4%, 46.8%, and 70.1%, respectively. On the other hand, those for M-NBI were 79.2%, 87.5%, 92.8%, 32.7%, and 81.9%, respectively, and those for MCE were 94.7%, 79.2%, 90.3%, 12.0%, and 89.6%, respectively. The PPV of CAD was relatively high (85.4%), and its specificity was similar to that of MCE (p = 0.72), but significantly lower than that of M-NBI and MCE in terms of sensitivity and accuracy rates (p < 0.001, p < 0.001). In comparing M-NBI and MCE, MCE had significantly higher sensitivity and accuracy rates (p < 0.001, p < 0.001).
Conclusion: Although WLI-based CAD showed good PPV for T1b diagnosis, its sensitivity and accuracy rates were inferior to those of M-NBI and MCE. Therefore, M-NBI and MCE are still required for the diagnosis of T1 CRC with deep invasion.

AIMS: A previous multicenter study showed that the diagnostic performance of a computer-aided diagnosis (CADx) system for deep (≥1mm) submucosally invasive (T1b) colorectal cancer using white light endoscopic images was excellent (sensitivity 87.3%, specificity 94.4%, presented at DDW 2021). However, the “regions of interest” (ROI) within images are not obvious. Class activation mapping (CAM) enables identification of the ROI that CADx utilizes for diagnosis. The purpose of this substudy was a quantitative investigation of the difference between CADx and endoscopists.
METHODS: Endoscopic images collected for validation of a previous study were used, including histologically proven T1b colorectal cancers (n=82; morphology: flat 36, polypoid 46; median maximum diameter 20mm, interquartile range 15-25mm; histological subtype: papillary 5, well 51, moderate 24, poor 2; location: proximal colon 26, distal colon 27, rectum 29). Application of CAM was limited to one white light endoscopic image (per lesion) to demonstrate findings of T1b cancers. The CAM images were generated from the weights of the previously fine-tuned ResNet50. Two expert endoscopists (YN, DN) depicted the ROI in identical images through discussion. Concordance of the ROI was rated by intersection over union (IoU) analysis (the ratio of overlapping areas of ROIs of both the CADx and endoscopist). Area was measured in pixels. Features of lesions with low IoU were explored. The CADx system also generated a probability score (range 0-1) for T1b cancers (probability score >0.5 is defined as a positive diagnosis). Quantitative variables were described by means and standard deviations (SD).
RESULTS: Pixel counts of ROIs were significantly lower using CADx (188.9K[x10³] ± 109.1K) than by endoscopists (354.5K ± 223.6K; p<0.0001). Mean ± SD of the IoU was 0.203 ± 0.170, range 0.000 to 0.700. IoU was significantly higher in correctly identified lesions (n=54, 0.234 ± 0.172) than incorrect ones (n=28, 0.144 ± 0.153, p=0.0215). Association of IoU with lesion morphology, size and location was not found, but IoU was significantly higher for moderately or poorly differentiated adenocarcinoma (0.262 ± 0.201), compared with papillary or well differentiated adenocarcinoma (0.176 ± 0.148, p=0.0330).
CONCLUSIONS: ROIs using the CADx system was smaller than that by endoscopists. IoU was larger in correctly diagnosed T1b colorectal cancers. These observations suggest that optimal annotation of the ROI may be the key to improve the diagnostic accuracy of the CADx for T1b colorectal cancers.