Cervical cancer is still an important cause of mortality among women in a number of countries. There are effective methods of prevention and early diagnosis, but they require well-trained medical professionals including cytologists. Within this project, we built a prototype of a new device together with implemented software using U-NET and CNN architectures of neural networks (ANN), to convert the currently used optical microscopes into fully independent scanning and evaluating systems for cytological samples. To evaluate the specificity and sensitivity of the system, 2058 (2000 normal and 58 abnormal samples) consecutive liquid-based cytology (LBC) samples were analysed. The observed sensitivity and specificity to distinguish normal and abnormal samples was 100%. We observed slight incompatibility in the evaluation of the type of abnormality. The use of ANN is promising for increasing the effectiveness of cervical screening. The low cost of neural network usage further increases the potential areas of application of the presented method. Further refinement of neural networks on a larger sample size is required to evaluate the software.
went preoperative 18F-FDG PET/CT were considered. SUV, MTV, TLG, geometrical shape, histograms and texture features were computed inside tumor contours. In group 1 (87 patients), univariate association with LN metastases was computed by Mann-Whitney test and a neural network multivariate model was developed. Univariate and multivariate models were assessed with leave one out on 20 training sessions and on group 2 (29 patients). Results Sensitivity and specificity of LN visual detection were 50% and 99% on group 1 and 33% and 95% on group 2. The lower sensitivity of visual detection in group 2 is mainly related to the higher rate of micrometastases (25% vs 13%). A unique heterogeneity feature computed on the primary tumor (GLSZM ZP) was able to predict LN metastases better than any other feature, or multivariate model (sensitivity and specificity of 75% and 81% in group 1 and of 89% and 80% in group 2). Tumors with LN metastases generally demonstrated a lower GLSZM ZP value, i.e. by the co-presence of high-uptake and low-uptake areas. Conclusions In our study the computation of imaging features on the primary tumor increases nodal staging for detection sensitivity in 18F-FDG PET and can be considered for a better planning of the surgical treatment.
e18017 Background: The incidence and mortality of cervical cancer are high in Poland. There are effective methods of the prevention and the early diagnosis, however they require well-trained medical professionals including cytologists. Within this project we built a prototype of a new device together with implemented software to convert the currently used optical microscopes to fully independent scanning systems for cytological samples. The use of the device is intended to improve the effectiveness of cytological screening, and registration of cytological tests’ results. The features of the software include digital backup as well as transmission and telemedicine evaluation. Methods: The software uses the artificial neural network (U-NET architecture) designed to be able to recognize suspicious regions and enhanced CNN neural network (VGG) allowing to determine the type of disorder such as: ASCUS, ASC-H, HIS, AGC, cancer. 7128 liquid based (LBC) and 1700 conventional cytology samples were evaluated by trained cyto-sreeners. Cytological abnormalities like: ASCUS, ASC-H, HIS, AGC, cancer were found in 254 (3.6%) LBC cases and 51 (3.0%) conventional cytology cases. All samples were scanned and archived. Selected samples with diagnosed abnormality were a model to teach the artificial neural networks. Results: Preliminary results obtained with use of U-NET and VGG (CNN networks) so far indicate 90-96% (LBC samples) and 85-95% (conventional cytology) compliance with results obtained using standard methods. Conclusions: Further refinement of neural networks is necessary to reduce the number of false positives and false negatives. A study with a larger sample size is required to evaluate the software. This project is co-financed with European Regional Development Fund within the Priority Axis I, Support for R&D activities for companies, Measure 1.2, Sectoral R&D Programmes, Sectoral Programme: “INNOMED – scientific research and development programme for innovative medicine economy sector”.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.