scholarly journals Deep learning model for classifying endometrial lesions

2021 ◽  
Vol 19 (1) ◽  
Author(s):  
YunZheng Zhang ◽  
ZiHao Wang ◽  
Jin Zhang ◽  
CuiCui Wang ◽  
YuShan Wang ◽  
...  

Abstract Background Hysteroscopy is a commonly used technique for diagnosing endometrial lesions. It is essential to develop an objective model to aid clinicians in lesion diagnosis, as each type of lesion has a distinct treatment, and judgments of hysteroscopists are relatively subjective. This study constructs a convolutional neural network model that can automatically classify endometrial lesions using hysteroscopic images as input. Methods All histopathologically confirmed endometrial lesion images were obtained from the Shengjing Hospital of China Medical University, including endometrial hyperplasia without atypia, atypical hyperplasia, endometrial cancer, endometrial polyps, and submucous myomas. The study included 1851 images from 454 patients. After the images were preprocessed (histogram equalization, addition of noise, rotations, and flips), a training set of 6478 images was input into a tuned VGGNet-16 model; 250 images were used as the test set to evaluate the model’s performance. Thereafter, we compared the model’s results with the diagnosis of gynecologists. Results The overall accuracy of the VGGNet-16 model in classifying endometrial lesions is 80.8%. Its sensitivity to endometrial hyperplasia without atypia, atypical hyperplasia, endometrial cancer, endometrial polyp, and submucous myoma is 84.0%, 68.0%, 78.0%, 94.0%, and 80.0%, respectively; for these diagnoses, the model’s specificity is 92.5%, 95.5%, 96.5%, 95.0%, and 96.5%, respectively. When classifying lesions as benign or as premalignant/malignant, the VGGNet-16 model’s accuracy, sensitivity, and specificity are 90.8%, 83.0%, and 96.0%, respectively. The diagnostic performance of the VGGNet-16 model is slightly better than that of the three gynecologists in both classification tasks. With the aid of the model, the overall accuracy of the diagnosis of endometrial lesions by gynecologists can be improved. Conclusions The VGGNet-16 model performs well in classifying endometrial lesions from hysteroscopic images and can provide objective diagnostic evidence for hysteroscopists.

2021 ◽  
Vol 81 (01) ◽  
pp. 75-85
Author(s):  
Ernesto Lara ◽  

Endometrial cancer represents worldwide the sixth most common malignant pathology in the female population, the endometroid type constitutes the most common form, usually developed from a typical sequence of endometrial hyperplasia secondary to sustained exposure to unopposed estrogens balanced by progestogens. Different classification systems for endometrial hyperplasia have been described, the most recent, published by the World Health Organization in 2014, proposes two categories: 1) hyperplasia without atypia, and 2) atypical hyperplasia or endometrial intraepithelial neoplasia. This classification avoids confusion due to the different terms in use and reflects a better understanding of the pathology behavior. Atypical hyperplasia or endometrial intraepithelial neoplasia is considered a precursor lesion to endometrial carcinoma type I. Health professionals must handle standardized terminology, accurately diagnose this entity, and ensure proper treatment of it. Keywords: Endometrial intraepithelial neoplasia, Endometrial hyperplasia, Atypical hyperplasia, Endometrial cancer.


2021 ◽  
pp. 59-61
Author(s):  
Bansi Kavar ◽  
Neeru Dave

Background: Endometrial hyperplasia is the precursor lesion of most endometrial cancers of endometrioid type. The most commonly used classication system for endometrial hyperplasia is WHO 1994 classication system in which architecture disruption and cytological atypia are used to identify four types of endometrial hyperplasia including simple or complex hyperplasia with or without atypia. Newer EIN diagnosis by cytological atypia is of great consideration for the progression to endometrial cancer. Material And Methods: The study consists of 100 cases of WHO classied endometrial hyperplasia for period of 4 yrs from 2015 to 2019. Type of sampling procedures- dilation & curettage, endometrial biopsy and fractional curettage. Objective: 1. To discuss revised criteria for recognition of endometrial intraepithelial neoplasia (EIN). 2. To nd out the sensitivity of endometrial intraepithelial neoplasia (EIN) classication in predicting the risk of malignancy. Results: This study consists of 100 cases of endometrial hyperplasia. Patients were mostly postmenopausal & presented with abnormal vaginal bleeding. From WHO classied endometrial lesions, 2 out of 35 cases of simple typical hyperplasia, 10 out of 14 cases of complex typical hyperplasia,12 out of 20 cases of simple atypical hyperplasia and 20 out of 21 cases of complex atypical hyperplasia were reclassied as EI N. Conclusion: To estimate the risk of progression to carcinoma and guide clinical management, the histo-pathologic diagnosis of endometrial hyperplastic lesion is very important, specially the diagnosis of EIN lesions. EIN carries a much greater risk of progression to endometrial cancer than other WHO classied endometrial hyperplasia.


2015 ◽  
Vol 25 (6) ◽  
pp. 1010-1014 ◽  
Author(s):  
Stanislav Mikhailovich Pronin ◽  
Olga Valerievna Novikova ◽  
Julia Yurievna Andreeva ◽  
Elena Grigorievna Novikova

ObjectiveTo evaluate oncologic and reproductive outcome with levonorgestrel-releasing intrauterine system combined with gonadotropin-releasing hormone agonist in women with grade 1 endometrial carcinoma, and the levonorgestrel monotherapy in women with complex atypical hyperplasia.Materials/MethodsA prospective study was conducted. We analyzed the clinical characteristics of 70 patients younger than 42 years (mean age, 33 years) with a diagnosis of complex atypical endometrial hyperplasia (AEH) or grade 1 endometrial adenocarcinoma who were treated with hormonal therapy at the Division of Gynecologic Oncology of P.A. Hertsen Moscow Cancer Research Institute from February 2009 to December 2012. Patients with complex AEH received monotherapy with levonorgestrel-releasing intrauterine system (Mirena, Shering, Finland; 52 mg). Patients with a diagnosis of grade 1 endometrial cancer were treated with levonorgestrel-releasing intrauterine system combined with gonadotropin-releasing hormone agonist (Zoladex; AstraZeneca UK Limited, UK; 3.6-mg depot). All the patients received hormonal therapy for a minimum of 6 months. Pretreatment evaluation consisted of transabdominal and transvaginal ultrasound in grayscale, color Doppler ultrasound, contrast-enhanced magnetic resonance imaging,cervical hysteroscopy, Pipelle endometrial biopsy, and morphological and immunohistochemical characteristics of the tissue.ResultsSeventy patients were included in study analyses. Twenty three (72%) of 32 patients with adenocarcinoma and 35 (92%) of 38 patients with AEH had complete remission, defined as the absence of any carcinoma or hyperplasia on endometrial sampling specimens. Among these cases, 2 patients with adenocarcinoma and 1 patient with AEH had recurrence after their complete response. Nine patients had persistent disease. Eight patients had 10 conceptions, resulting in 8 live births.ConclusionsThe suggested conservative treatment strategy can be considered as a valid therapeutic option for young women of childbearing potential with atypical endometrial hyperplasia and grade 1 endometrial adenocarcinoma who wish to preserve their fertility and thus may be recommended as an alternative to hysterectomy. Close follow-up during and after the treatment period is strictly required.


GYNECOLOGY ◽  
2018 ◽  
Vol 20 (2) ◽  
pp. 14-17
Author(s):  
G E Chernukha ◽  
I A Ivanov ◽  
Z N Efendieva ◽  
M R Dumanovskaya ◽  
A V Asaturova

Abnormal uterine bleeding (AUB) is one of the most common indications for hysteroscopy. Most of the AUB cases occur due to endometrial or myometrium pathology. Among it, endometrial polyps (EP) and chronic endometritis (CE) prevalent in reproductive age, while endometrial hyperplasia (EH) and EP dominate in perimenopause. It was determined that EP and CE are characterized with menorrhagia and metrorrhagia approximately equally, whereas EH reveals AUB with oligomenorrhoea. Verification of exact endometrial pathology by ultrasound examination is hindered, that results in deviations of ultrasound and histological diagnosis. The usage of ultrasound data and AUB’s characteristics may improve the diagnostic accuracy on preadmission period.


2020 ◽  
Vol 103 (6) ◽  
pp. 585-593

Objective: To evaluate the accuracy of transvaginal ultrasonography (TVS) and saline infusion sonography (SIS) in use for the diagnosis of endometrial polyps and submucous myoma compared to hysteroscopy. Histopathology was considered as the gold standard for final diagnosis. Materials and Methods: The present retrospective study was conducted at Bhumibol Adulyadej Hospital, Bangkok, Thailand between January 2014 and December 2017. Medical records of 150 patients who attended for hysteroscopy and histopathological diagnosis were reviewed. The accuracy of TVS and SIS for the diagnosis of endometrial polyps and submucous myoma were determined. Results: Out of 150 enrolled cases, endometrial polyp was the most frequent hysteroscopic finding in participants of the present study (92/150). Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of TVS, SIS, and hysteroscopy compared to pathologic reports for detection endometrial polyps were 71.7% versus 93.5% versus 97.8%, 38.5% versus 52.2% versus 68.2%, 80.5% versus 88.7% versus 92.8%, 27.8% versus 66.7% versus 88.2%, and 64.4% versus 85.2% versus 92.1%, respectively. The sensitivity, specificity, PPV, NPV, and accuracy of TVS, SIS, and hysteroscopy for detection of submucous myoma were 81.6% versus 92.1% versus 94.7%, 66.7% versus 86.9% versus 100%, 77.5% versus 92.1% versus 100%, 72.0% versus 86.9% versus 90.9%, and 75.4% versus 90.2% versus 96.6%, respectively. The kappa value from TVS, SIS, and hysteroscopy when the histopathologic reports were overall intrauterine abnormalities, endometrial polyps and submucous myoma were 0.45/0.43/0.72, 0.77/0.76/0.89, and 0.92/0.92/1.00, respectively. Conclusion: Sensitivity, specificity, PPV, NPV, accuracy, and kappa value of SIS for detecting endometrial polyps and submucous myoma were better than TVS. Keywords: Ultrasonography, Saline infusion sonography, Hysteroscopy, Accuracy


2020 ◽  
Vol 19 ◽  
pp. 153303382096558
Author(s):  
Lixia Shan ◽  
Tao Zhao ◽  
Yu Wang

Objective: Long non-coding RNAs (lncRNAs) play a critical role in tumorigenesis. Upregulation of lncRNA deleted in lymphocytic leukemia 1 (DLEU1) has been reported in endometrial cancer (EC) tissues. This prospective study aimed to determine the potential clinical significance of serum lncRNA DLEU1 in EC. Methods: The serum lncRNA DLEU1 level was detected in EC patients, patients with endometrial hyperplasia and healthy controls by reverse transcription-quantitative polymerase chain reaction (RT-qPCR). Then its clinical value in EC was further evaluated. Results: Our results demonstrated that serum lncRNA DLEU1 levels were significantly increased in patients with EC, and serum lncRNA DLEU1 showed good performance for discriminating EC patients from patients with endometrial hyperplasia and healthy controls. In addition, EC patients with advanced clinicopathological features had higher circulating lncRNA DLEU1 level than those with favorable clinical characteristics. Moreover, EC patients in the high serum lncRNA DLEU1 group suffered worse overall survival and disease-free survival than those in the low serum lncRNA DLEU1 group. Furthermore, multivariate cox regression analysis displayed that the serum lncRNA DLEU1 served as an independent prognostic factor for EC. Conclusions: Collectively, our study suggests that serum lncRNA DLEU1 is a novel and promising biomarker for prognostic estimation of EC.


Diagnostics ◽  
2021 ◽  
Vol 11 (3) ◽  
pp. 442
Author(s):  
Norbert Stachowicz ◽  
Agata Smoleń ◽  
Michał Ciebiera ◽  
Tomasz Łoziński ◽  
Paweł Poziemski ◽  
...  

Background: Abnormal uterine bleeding (AUB) represents a common diagnostic challenge, as it might be related to both benign and malignant conditions. Endometrial cancer may not be detected with blind uterine cavity sampling by dilatation and curettage or suction devices. Several scoring systems using different ultrasound image characteristics were recently proposed to estimate the risk of endometrial cancer (EC) in women with AUB. Aim: The aim of the present study was to externally validate the predictive value of the recently proposed scoring systems including the Risk of Endometrial Cancer scoring model (REC) for EC risk stratification. Material and methods: It was a retrospective cohort study of women with postmenopausal bleeding. From June 2012 to June 2020 we studied a group of 394 women who underwent standard transvaginal ultrasound examination followed by power Doppler intrauterine vascularity assessment. Selected ultrasound features of endometrial lesions were assessed in each patient. Results: The median age was 60.3 years (range ±10.7). The median body mass index (BMI) was 30.4 (range ± 6.0). Histological examination revealed 158 cases of endometrial hyperplasia (EH) and 236 cases of EC. Of the studied ultrasound endometrial features, the highest areas under the curve (AUCs) were found for endometrial thickness (ET) (AUC = 0.76; 95% CI: 0.71–0.81) and for interrupted endomyometrial junction (AUC = 0.70, 95% CI: 0.65–0.75). Selected scoring systems presented moderate to good predictive performance in differentiating EC and EH. The highest AUC was found for REC model (AUC = 0.75, 95% CI: 0.70–0.79) and for the basic model that included ET, Doppler score and interrupted endometrial junction (AUC = 0.77, 95% CI: 0.73–0.82). REC model was more accurate than other scoring systems and selected single features for differentiating benign hyperplasia from EC at early stages, regardless of menopausal status. Conclusions: New scoring systems, including the REC model may be used in women with AUB for more efficient differentiation between benign and malignant conditions.


2014 ◽  
Vol 21 (2) ◽  
pp. 296-302 ◽  
Author(s):  
Daniel Spadoto Dias ◽  
Flávia Neves Bueloni-Dias ◽  
Rogério Dias ◽  
Jorge Nahás-Neto ◽  
Eliana Aguiar Petri Nahás ◽  
...  

Sign in / Sign up

Export Citation Format

Share Document