Introduction

Anorectal functional disorders are common in clinical practice, impacting the life of over 5% of the world population. High-resolution anorectal manometry (HR-ARM) is considered the gold standard for evaluating anorectal functional disorders. However, progress is hindered by the limited accessibility of the procedure, which is not available across all centers, and by the complexity of the data analysis involved. The London Protocol and Classification were developed as a standardized method for anorectal motility patterns classification. Artificial intelligence models have increasingly been developed in medical areas with a strong imaging component and Gastroenterology is not an exception. However, when it comes to functional anorectal disorders it remains uncharted territory, with only a few research articles published. This study aims to develop and validate an AI model for the identification and differentiation of disorders of anal tone and contractility during HR-ARM, based on the London Classification.

Aims & Methods

A dataset of 701 HR-ARM exams from a tertiary center was used for developing and testing of multiple machine learning (ML) algorithms. All HR-ARM procedures were performed with Solar™ GI High-Resolution Anorectal Manometry HRAM single use Water Perfused Catheters Cathethers (Laborie^TM, Enschede, The Netherlands). Classification of anorectal motility disorders was based on the London Classification. The training dataset had 80% of the total exams, with the remaining 20% being used for the testing dataset. The testing dataset was used for the models’ evaluation through its accuracy, f1-weight score, sensitivity, specificity, positive and negative predictive values and area under the receiving-operating characteristic curve.

Results

After models’ fine tuning, LGBM Classifier had the best performance, with a mean accuracy of 87.0% and a F1 weight score of 85.0% for identifying disorders of anal tone and contractility. Different ML models excelled in distinguishing specific disorders of anal tone and contractility, with mean accuracy over 90.0%. LGBMClassifier identified anal hypotension with normal contractility with a F1 weight score of 91.0%. On the other hand, AdaBoostClassifier had a F1 weight score of 97.0% for combined anal hypotension and hypocontractility. Anal normotension with hypocontractility was detected with 90.0% F1 weight score by LGBM Classifier. Finally, DecisionTreeClassifier excelled in the detection of anal hypertension, with a F1 weight score of 97.0%.

Conclusion

This is the first worldwide study proving the accuracy of ML models for differentiation of motility patterns in HR-ARM, based on London Classification. The evaluation of disorders of anal tone and contractility with HR-ARM is an indirect measure of the internal and external anal sphincter function and accurate identification of these disorders is of utmost importance for guiding clinical practice. Artificial intelligence has the potential to enhance accessibility to HR-ARM by facilitating the evaluation of the data and, perhaps, making it more cost-efficient. Finally, interoperability, a key point of interest, can be achieved using AI models.

References

1. Heitmann PT, Vollebregt PF, Knowles CH, et al. Understanding the physiology of human defaecation and disorders of continence and evacuation. Nat Rev Gastroenterol Hepatol 2021;18:751-769.
2. Drossman DA, Hasler WL. Rome IV-Functional GI Disorders: Disorders of Gut-Brain Interaction. Gastroenterology 2016;150:1257-61.
3. Carrington EV, Knowles CH, Grossi U, et al. High-resolution Anorectal Manometry Measures Are More Accurate Than Conventional Measures in Detecting Anal Hypocontractility in Women With Fecal Incontinence. Clin Gastroenterol Hepatol 2019;17:477-485.e9.
4. Scott SM, Carrington EV. The London Classification: Improving Characterization and Classification of Anorectal Function with Anorectal Manometry. Curr Gastroenterol Rep 2020;22:55.
5. Carrington EV, Heinrich H, Knowles CH, et al. The international anorectal physiology working group (IAPWG) recommendations: Standardized testing protocol and the London classification for disorders of anorectal function. Neurogastroenterol Motil 2020;32:e13679.

Introduction

High-resolution anoscopy (HRA) is the gold standard for assessing HPV-related anal lesions, though its diagnostic accuracy remains limited. While most AI models focus on binary classification, this study aims to develop and evaluate a deep learning model for detection and classification of HRA frames into high-grade (HSIL), low-grade (LSIL), and non-dysplastic lesions (e.g., inflammation).

Aims & Methods

This multicenter retrospective study included 191,951 frames from 107 HRA procedures across five centers using five imaging devices. Frames were labeled as HSIL, LSIL, or non-dysplastic based on matched histopathology. A YOLOv11-based object detection model was trained to detect these lesions using bounding boxes. The dataset was randomly split into training, validation, and test sets. Lesion detection was defined by correct bounding box placement. Classification metrics included recall (sensitivity), precision (PPV), and accuracy, reported as weighted averages across confidence thresholds of 0.45, 0.47, 0.50, 0.52 and 0.55.

Results

The model achieved an overall accuracy of 95.2% (94.7- 95.7) in classifying detected lesions. For HSIL, recall was 97.9% (97.6- 98.2) and precision was 93.9% (92.8, 95.1)). For LSIL, recall was 98.9% (98.8- 99.0) and precision was 98.5% (98.3, 98.7). For non-dysplastic lesions, recall was 98.7% (98.5- 98.8) and precision was 96.0% (94.9- 97.2).

Conclusion

This study introduces the first worldwide AI model capable of simultaneous detection and trinary classification (HSIL, LSIL, non-dysplastic mimickers) of anal canal lesions using heterogeneous, interoperable datasets, a key advancement toward real-world deployment and precision-guided care in AI-assisted HRA.

Introduction

Anorectal disorders, causing fecal incontinence or evacuation dysfunction are common, affecting up to 5% of the general population (1). Most of these disorders result from an intricate combination of disturbances, including neurologic and pelvic floor musculature dysfunction, and have a significant impact in the quality of life of patients (2,3). Anorectal manometry (ARM) is the gold standard for the evaluation of anorectal functional disorders. Nevertheless, the accessibility to this exam is limited, and the complexity of data analysis and report is a significant drawback (4).

Aims & Methods

This pilot study aimed to develop and validate an artificial intelligence (AI) model to automatically differentiate motility patterns of fecal incontinence (FI) from obstructed defecation (OD) using ARM data.
We included ARM exams performed at a large gastrointestinal motility clinic in Brazil (Pelvia – Gastrointestinal Motility and Continence, Paraná, Brazil) between 2015 and 2021. All ARM procedures were performed using water-perfused probes with eight radial transducers with a distal latex balloon. We developed and tested multiple machine learning algorithms for the automatic interpretation of ARM data. Four models were tested: k-nearest neighbors (KNN), support vector machines (SVM), random forests (RF) and gradient boosting (xGB). These models were trained using a stratified 5-fold strategy. Their performance was assessed after fine-tuning of each model’s hyperparameters, using 90% of data for training and 10% of data for testing.

Results

A total of 827 ARM exams were used in this study, 493 of patients diagnosed with obstructed defecation (OD), while the remaining 334 presented FI. The performance marks of the different tuned models, including the precision, sensitivity and the F1-score, is summarized on table 1. After fine-tuning, the xGB model presented an overall accuracy (84.6 ± 2.9%), similar to that of RF (82.7 ± 4.8%) and SVM (81.0 ± 8.0%), and higher that of KNN (74.4 ± 3.8%). The xGB models showed the highest discriminating performance between OD and FI, with an area under the curve of 0.939.

Conclusion

The tested machine learning algorithms, particularly the xGB model, accurately differentiated between FI and OD manometric patterns. Subsequent development of these tools may optimize the access to ARM studies, which may significantly impact on the management of patients with anorectal functional diseases.

References

1. Whitehead WE, Wald A, Diamant NE, et al. Functional disorders of the anus and rectum. Gut 1999;45 Suppl 2:Ii55-9.
2. Maeda Y, Vaizey CJ, Hollington P, et al. Physiological, psychological and behavioural characteristics of men and women with faecal incontinence. Colorectal Dis 2009;11:927-32.
3. Remes-Troche JM, Rao SS. Defecation disorders: neuromuscular aspects and treatment. Curr Gastroenterol Rep 2006;8:291-9.
4. Brown HW, Guan W, Schmuhl NB, et al. If We Don't Ask, They Won't Tell: Screening for Urinary and Fecal Incontinence by Primary Care Providers. J Am Board Fam Med 2018;31:774-782.

Introduction

High-resolution anoscopy (HRA) plays a central role in the detection and treatment of precursors of anal squamous cell carcinoma (ASCC). The visual characterization of high-grade intraepithelial lesions (HSIL) is hampered by a limited interobserver agreement. Artificial intelligence (AI) algorithms have shown high levels of efficiency in detecting and differentiating HSIL from low-grade squamous intraepithelial lesions (LSIL) in HRA images. Nevertheless, these studies have been conducted using a digital videoproctoscope, whose use is not predicted in the standards of the International Anal Neoplasia Society (IANS). Our aim was to develop a deep learning system for automatic detection and differentiation of HSIL versus LSIL using HRA images from both conventional and digital proctoscopes.

Aims & Methods

A convolutional neural network (CNN) was developed based on 151 HRA exams from 137 patients undergoing HRA at two large volume centres (GH Paris Saint-Joseph, Paris, France and Emílio Ribas Infecciology Institute, São Paulo, Brazil), using conventional HRA (KLP 200 LED®, Kolplast) and digital HRA (THD Proctostation®, THD SpA). A total of 58,498 images were included, 29,212 images containing HSIL and 29,286 LSIL. Partial subanalyses were performed to evaluate the performance of the CNN in the subset of images with no staining (n=2,820), acetic acid (n=27,191), lugol (n=10,011), and after treatment of the anal canal (n=11,047). For these subanalyses, training, and testing datasets comprised 90% and 10% of each class's total number of images. The sensitivity, specificity, accuracy, positive and negative predictive values, and area under the curve (AUC) were calculated.

Results

The overall accuracy of the CNN in distinguishing HSIL from LSIL during the testing stage was 98.6%. The algorithm had an overall sensitivity and specificity of 97.3% and 99.1%, respectively. The overall AUC was 0.99. For staining with acetic acid, HSIL was differentiated from LSIL with an overall accuracy of 96.4%, while for lugol and after therapeutic manipulation these values were 96.6% and 99.3%, respectively. The AUC ranged between 0.98 and 1.00 in these particular subgroups.

Conclusion

The introduction of AI algorithms to HRA may enhance the early diagnosis of ASCC precursors, and this system showed to perform adequately across conventional and digital HRA interfaces. Improving the detection of ASCC precursors may improve the treatment outcomes and prognosis. The developed AI model is the first model capable of performing adequately in different clinical sub settings and in different systems available on the market.

References

1. Hillman, R.J.; Cuming, T.; Darragh, T.et al. 2016 IANS International Guidelines for Practice Standards in the Detection of Anal Cancer Precursors. J Low Genit Tract Dis 2016; 20: 283-291.
2. Mistrangelo, M.; Salzano, A. Progression of LSIL to HSIL or SCC: is anoscopy and biopsy good enough? Tech Coloproctol 2019; 23: 303-304.
3. Deshmukh, A.A.; Suk, R.; Shiels, M.S.et al. Recent Trends in Squamous Cell Carcinoma of the Anus Incidence and Mortality in the United States, 2001-2015. J Natl Cancer Inst 2020; 112: 829-838.
4. Palefsky, J.M.; Lee, J.Y.; Jay, N.et al. Treatment of Anal High-Grade Squamous Intraepithelial Lesions to Prevent Anal Cancer. New England Journal of Medicine 2022; 386: 2273-2282.
5. Neukam, K.; Milanés Guisado, Y.; Fontillón, M.et al. High-resolution anoscopy in HIV-infected men: Assessment of the learning curve and factors that improve the performance. Papillomavirus Res 2019; 7: 62-66.
6. Saraiva, M.M.; Spindler, L.; Fathallah, N.et al. Artificial intelligence and high-resolution anoscopy: automatic identification of anal squamous cell carcinoma precursors using a convolutional neural network. Tech Coloproctol 2022; 26: 893-900.
7. Pathologists, C.o.A. Protocol for the Examination of Excision Specimens From Patients With Carcinoma of the Anus. 2021; 4.2.0.

Introduction

Esophageal motility disorders are common in the clinical practice, with impact in patient’s quality of life. High-resolution esophageal manometry (HREM) is the gold standard for evaluation of patients with functional esophageal disorders. The Chicago Classification offers a standardized approach HREM, reducing bias and enabling a precise diagnosis of these disorders. Despite these, HREM remains a complex exam, both in data analysis and in accessibility. This proof-of-concept study aims to develop and validate an artificial intelligence model for identification of motility disorders in HREM according to the Chicago Classification.

Aims & Methods

618 HREM exams from two different devices and three centers from Spain and the United States of America were used for model development. Several machine learning models were trained and evaluated for detection of motility disorders. The exams were classified according to the Chicago Classification 4.0, based on a consensus reached by two independent HREM experts. The training and testing dataset were divided in a 80%/20% patient split design based on label frequency. Models’ performance was evaluated through their accuracy and area under the receiver-operating characteristic curve (AUC-ROC).

Results

The GradientBoostingClassifier model outperformed the remaining machine learning models with an accuracy of 94.2% and an AUCROC of 0.921 for identifying disorders of esophagogastric junction outflow. The xGBClassifier model detected disorders of peristalsis with an accuracy of 80.9% and an AUC-ROC of 0.871.

Conclusion

In this pilot study with integration of exams from different devices and demographic contexts, several machine learning models accurately identified disorders of the esophagogastric junction outflow and peristalsis in HREM according to Chicago Classification.
The integration of AI into the study of esophageal motility disorders through HREM-driven diagnosis is expected to revolutionize the way these conditions are managed, optimizing both exam accuracy and accessibility, transforming patient management.