Classification of cervical neoplasms on colposcopic photography using deep learning

Abstract

Colposcopy is widely used to detect cervical cancers, but experienced physicians who are needed for an accurate diagnosis are lacking in developing countries. Artificial intelligence (AI) has been recently used in computer-aided diagnosis showing remarkable promise. In this study, we developed and validated deep learning models to automatically classify cervical neoplasms on colposcopic photographs. Pre-trained convolutional neural networks were fine-tuned for two grading systems: the cervical intraepithelial neoplasia (CIN) system and the lower anogenital squamous terminology (LAST) system. The multi-class classification accuracies of the networks for the CIN system in the test dataset were 48.6 ± 1.3% by Inception-Resnet-v2 and 51.7 ± 5.2% by Resnet-152. The accuracies for the LAST system were 71.8 ± 1.8% and 74.7 ± 1.8%, respectively. The area under the curve (AUC) for discriminating high-risk lesions from low-risk lesions by Resnet-152 was 0.781 ± 0.020 for the CIN system and 0.708 ± 0.024 for the LAST system. The lesions requiring biopsy were also detected efficiently (AUC, 0.947 ± 0.030 by Resnet-152), and presented meaningfully on attention maps. These results may indicate the potential of the application of AI for automated reading of colposcopic photographs. Cervical cancer is the fourth most common cancer in women worldwide, and the second most common cancer among females in developing countries 1. Screening is the principal prevention method aimed at reducing mortality rates. Screening includes certain steps, including population-based Papanicolaou (Pap) testing, colposcopydirected biopsy of suspicious lesions, and the treatment of confirmed pre-cancer lesions 2,3. In women with low-grade intraepithelial lesions (LSIL) or high-grade intraepithelial lesions (HSIL), the risk of pre-cancer is medium to high, and immediate referral for colposcopy is necessary. However, referring all women with atypical squamous cells of undetermined significance (ASC-US) is considered inefficient, as the risk of such cases being pre-cancerous is lower 4. Screening programs have been successful in the developed countries, leading to an approximately 80% decrease in the cervical cancer incidence over the past 4 decades. In contrast, the increase in cervical cancer incidence reported in developing countries 5 has been attributed to the unsuccessful implementation of screening programs. This, has been attributed to logistics in health systems, infrastructural inadequacies, and the lack of expert physicians capable of introducing screening programs and follow-up 6 .

Diagnosis of Cervical Cancer based on Ensemble Deep Learning Network using Colposcopy Images

“…The proposed model uses a different filter size to extract distinct features and works well for unexpected data. It offers a 92.40% sensitivity and a 96.20% specificity, which improves sensitivity and specificity by approximately 25% compared with Inception-Resnet-v2 in [ 45 ]. The proposed method has trouble distinguishing the false positive samples from those with fewer false positives.…”

The role of artificial intelligence in healthcare: a structured literature review

“…Another critical area of medicine where AI is making an impact is clinical decision-making and disease diagnosis. AI technologies can ingest, analyse, and report large volumes of data across different modalities to detect disease and guide clinical decisions [ 3 , 8 ]. AI applications can deal with the vast amount of data produced in medicine and find new information that would otherwise remain hidden in the mass of medical big data [ 9 – 11 ].…”

“…However, as Meskò et al [ 7 ] find, the technology will potentially reduce care costs and repetitive operations by focusing the medical profession on critical thinking and clinical creativity. As Cho et al and Doyle et al [ 8 , 9 ] add, the AI perspective is exciting; however, new studies will be needed to establish the efficacy and applications of AI in the medical field [ 10 ].…”

The Use of Artificial Intelligence in Automation in the Fields of Gynaecology and Obstetrics – an Assessment of the State of Play

“…The authors optimised pre-trained CNNs in two scoring systems: the cervical intraepithelial neoplasia (CIN) system and the lower anogenital squamous terminology (LAST) system. The CNNs were capable of efficiently identifying biopsy-worthy findings (AUC 0.947) 18 . Shanthi et al were able to correctly classify microscopic cervical cell smears as normal, mild, moderate, severe and carcinomatous with an accuracy of 94.1%, 92.1% and 85.1%, respectively, using various CNNs trained with augmented data sets (original colposcopy, contour-extracted and binary image data) 19 .…”