Can a deep learning model trained on laparoscopic video from expert endometriosis centres automatically detect and classify endometriosis lesions during surgery, and does performance vary according to lesion subtype?<\/p>\n
Study Design and Population<\/strong><\/p>\n This retrospective multicentre proof-of-concept study included 112 women undergoing laparoscopy for suspected endometriosis at expert centres in France, Hungary, Brazil and Denmark.<\/p>\n The authors extracted 19,721 laparoscopic frames containing 61,797 lesion annotations. Lesions were categorised into nine visual classes: superficial black, red, white and subtle lesions; filmy and dense adhesions; deep endometriosis; ovarian endometrioma; and ovarian chocolate fluid.<\/p>\n Importantly, annotations were based on expert visual assessment rather than histopathological confirmation. While this introduces limitations, it also reflects the reality that intraoperative diagnosis of endometriosis remains heavily dependent on surgical expertise and visual pattern recognition.<\/p>\n Methods and Outcomes<\/strong><\/p>\n A YOLOv5 object-detection neural network was trained using a development dataset and evaluated on separate patient-level validation and test datasets to minimise overfitting and data leakage. Results<\/strong><\/p>\n Performance was strongest for visually distinctive lesions, including superficial black lesions (F1 0.94), superficial subtle lesions (F1 0.74) and ovarian chocolate fluid (F1 0.75). Dense adhesions (F1 0.70), ovarian endometrioma (F1 0.63) and deep endometriosis (F1 0.62) demonstrated moderate detection performance despite their anatomical and morphological heterogeneity.<\/p>\n In contrast, superficial red lesions (F1 0.25), superficial white lesions (F1 0.18) and filmy adhesions (F1 0.02) were poorly detected, largely because these lesions were frequently missed rather than incorrectly classified.<\/p>\n Interestingly, the confusion matrix (below) demonstrated relatively little misclassification between lesion types. The dominant problem was lesion omission rather than incorrect categorisation.<\/p>\n Strengths<\/strong><\/p>\nThis is one of the first multicentre studies to apply real-time computer vision to endometriosis surgery and represents an important step beyond binary disease detection (endometriosis present \/ absent) towards lesion-specific recognition. Limitations and Clinical Relevance<\/strong><\/p>\n The central limitation is the absence of a true reference standard. The model was trained on surgeon-defined appearances rather than histologically confirmed disease, meaning the algorithm effectively learns expert visual recognition rather than definitive endometriosis diagnosis. Whether this represents a limitation or simply reflects the realities of contemporary endometriosis surgery remains open to debate.<\/p>\n Additional limitations include the: relatively small dataset by modern A.I. standards; inclusion of only positive surgical cases; lack of external prospective validation; and performance assessment on individual frames rather than complete video sequences.<\/p>\n Conventional computer vision metrics may underestimate clinical usefulness. Surgeons do not require every lesion to be identified in every frame; a system that highlights suspicious pathology during a surgical survey may still provide meaningful assistance.<\/p>\n Implications for Practice<\/strong><\/p>\n This technology is not ready for clinical deployment, and cannot be used to determine disease clearance, nor to reassure patients that endometriosis is absent.<\/p>\n However, the study demonstrates that automated visual recognition of endometriosis is feasible. Future applications may include surgical education, lesion mapping, and quality assurance.<\/p>\n Take-Home Messages<\/strong><\/p>\n This multicentre proof-of-concept demonstrates that A.I. can identify several common endometriosis phenotypes during laparoscopy, particularly visually distinctive lesions such as black peritoneal disease, endometriomas and deep lesions. Performance remains poor for red, white and filmy disease, highlighting the ongoing challenge of detecting subtle pathology. While not yet clinically applicable, this study represents an important milestone in the development of A.I.-assisted endometriosis surgery and raises important questions about how we define and standardise visual diagnosis in our field.<\/p>\n\t Artificial intelligence (A.I.) is rapidly entering the operating theatre, but can it recognise endometriosis as reliably as an experienced surgeon? This Journal Club reviews a multicentre proof-of-concept study exploring A.I.-based lesion detection during laparoscopy, and discusses why the paper’s most important contribution may be the attempt to define the visual language of endometriosis surgery itself.<\/p>\n","protected":false},"author":1,"featured_media":72,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"site-sidebar-layout":"no-sidebar","site-content-layout":"","ast-site-content-layout":"full-width-container","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[1],"tags":[],"class_list":["post-62","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorised"],"acf":[],"_links":{"self":[{"href":"https:\/\/escope.ages.com.au\/july-2026\/wp-json\/wp\/v2\/posts\/62","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/escope.ages.com.au\/july-2026\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/escope.ages.com.au\/july-2026\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/escope.ages.com.au\/july-2026\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/escope.ages.com.au\/july-2026\/wp-json\/wp\/v2\/comments?post=62"}],"version-history":[{"count":6,"href":"https:\/\/escope.ages.com.au\/july-2026\/wp-json\/wp\/v2\/posts\/62\/revisions"}],"predecessor-version":[{"id":216,"href":"https:\/\/escope.ages.com.au\/july-2026\/wp-json\/wp\/v2\/posts\/62\/revisions\/216"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/escope.ages.com.au\/july-2026\/wp-json\/wp\/v2\/media\/72"}],"wp:attachment":[{"href":"https:\/\/escope.ages.com.au\/july-2026\/wp-json\/wp\/v2\/media?parent=62"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/escope.ages.com.au\/july-2026\/wp-json\/wp\/v2\/categories?post=62"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/escope.ages.com.au\/july-2026\/wp-json\/wp\/v2\/tags?post=62"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
The primary outcomes were precision, recall and F1 score for each lesion subtype. Performance was assessed on a frame-by-frame basis using standard computer vision metrics. [F1 score is a single measure that combines precision (how often a detected lesion is truly present) and recall (how often lesions present are successfully detected). An F1 score of 1.0 represents perfect performance.]<\/p>\n![\"Table](\"https:\/\/escope.ages.com.au\/july-2026\/wp-content\/uploads\/sites\/13\/2025\/05\/Table-Article-5-1.png\" "\\"Table")
\n\t<\/figure>\n\t
\nPerhaps the most important contribution is the creation of a large, surgeon-annotated visual dataset and structured visual taxonomy of endometriosis lesions. The inclusion of complex operative scenes and heterogeneous disease makes the dataset more representative of real-world surgery.\n![\"\"](\"https:\/\/escope.ages.com.au\/july-2026\/wp-content\/uploads\/sites\/13\/2025\/05\/Photo-Tristan-McCaughey-237x300.png\")
<\/p>\nDr Tristan McCaughey
\nMBBS (Hons), BMedSc (Hons)
\nThe Mercy Hospital for Women (Melbourne)\n\t![\"\"](\"https:\/\/escope.ages.com.au\/july-2026\/wp-content\/uploads\/sites\/13\/2025\/05\/Al-profile-pic-300x300.jpg\")
<\/p>\nDr Alison Bryant-Smith
\nMBBS\/BA, MPH, MSurgEd, MRCOG, FRANZCOG, AGES ATP
\nThe Royal Women’s Hospital (Melbourne); Northern Health (Melbourne)\n\n","protected":false},"excerpt":{"rendered":"