Cervical lesion assessment using real‐time microendoscopy image analysis in Brazil: The CLARA study

Abstract

We conducted a prospective evaluation of the diagnostic performance of high-resolution microendoscopy (HRME) to detect cervical intraepithelial neoplasia (CIN) in women with abnormal screening tests. Study participants underwent colposcopy, HRME and cervical biopsy. The prospective diagnostic performance of HRME using an automated morphologic image analysis algorithm was compared to that of colposcopy using histopathologic detection of CIN as the gold standard. To assess the potential to further improve performance of HRME image analysis, we also conducted a retrospective analysis assessing performance of a multi-task convolutional neural network to segment and classify HRME images. One thousand four hundred eighty-six subjects completed the study; 435 (29%) subjects had CIN Grade 2 or more severe (CIN2+) diagnosis. HRME with morphologic image analysis for detection of CIN Grade 3 or more severe diagnoses (CIN3+) was similarly sensitive (95.6% vs 96.2%, P = .81) and specific (56.6% vs 58.7%, P = .18) as colposcopy. HRME with morphologic image analysis for detection of CIN2+ was slightly less sensitive (91.7% vs 95.6%, P < .01) and specific (59.7% vs 63.4%, P = .02) than colposcopy. Images from 870 subjects were used to train a multi-task convolutional neural network-based algorithm and images from the remaining 616 were used to validate its performance. There were no significant differences in the sensitivity and specificity of HRME with neural network analysis vs colposcopy for detection of CIN2+ or CIN3+.Using a neural network-based algorithm, HRME has comparable sensitivity and specificity to colposcopy for detection of CIN2+. HRME could provide a low-cost, point-of-care alternative to colposcopy and biopsy in the prevention of cervical cancer.

The Impact of Artificial Intelligence on Health Equity in Oncology: Scoping Review

“…Studies from our review investigated a wide range of cancers, with general oncological applications being the dominant category (28/133, 21.1%), followed by gynecologic (19/133, 14.3%), breast (16/133, 12%), oral (12/133, 9%), prostate (12/133, 9%), and dermatologic cancers (12/133, 9%). Among the articles on gynecologic cancers, 84% (16/19) were categorized under theme 1, discussing the use of AI technologies to address disparities in gynecologic cancer screening (11/16, 70%) [23,[84][85][86][87][88][89][90][91][92][93], diagnosis (4/16, 25%) [94][95][96][97], and treatment (1/16, 6%) [98]. Of the 16 articles, 15 (94%) developed AI technologies to target gynecologic cancer disparities in lowand middle-income countries (LMICs) [84][85][86][87][88][89][90][91][92][93][94][95][96][97][98], while 1 (6%) did so for implementation in high-income countries (HICs) [23].…”

“…Among the articles on gynecologic cancers, 84% (16/19) were categorized under theme 1, discussing the use of AI technologies to address disparities in gynecologic cancer screening (11/16, 70%) [23,[84][85][86][87][88][89][90][91][92][93], diagnosis (4/16, 25%) [94][95][96][97], and treatment (1/16, 6%) [98]. Of the 16 articles, 15 (94%) developed AI technologies to target gynecologic cancer disparities in lowand middle-income countries (LMICs) [84][85][86][87][88][89][90][91][92][93][94][95][96][97][98], while 1 (6%) did so for implementation in high-income countries (HICs) [23]. The other 3 (n=19, 16%) articles fell under theme 3, discussing the use of AI to explore the genetic (1/3, 33%) [99] and social (2/3, 67%) determinants of health outcomes in gynecologic cancers [100,101].…”

“…The other 3 (n=19, 16%) articles fell under theme 3, discussing the use of AI to explore the genetic (1/3, 33%) [99] and social (2/3, 67%) determinants of health outcomes in gynecologic cancers [100,101]. Moreover, most of these articles were clinical studies (14/19, 74%) [84][85][86][87][88][89][93][94][95][97][98][99][100][101], 16% (3/19) were commentaries [23,90,91], 5% (1/19) was an epidemiological study [96], and 5% (1/19) was a review [92].…”

“…When the AI algorithms were directly compared with the standard of care, the results varied. Most of the articles noted no difference between AI algorithms and the standard of care [93,97,[123][124][125], whereas others observed that the accuracy of AI algorithms was lower than that of human physicians [85,126]. One study noted that AI outperformed its human counterpart when detecting and staging prostate cancer in a higher number of patients [117].…”

Cervical Imaging in the Low Resource Setting: A Review

“…The HRME has been deployed in clinical settings such as Botswana, Brazil, the United States (Texas), and El Salvador [ 44 , 47 , 48 , 49 , 50 , 51 ]. A human study in Botswana was performed by first conducting a routine colposcopic examination.…”