To construct a pathological image diagnostic model of cervical adenocarcinoma in situ(CAIS) based on deep learning algorithm.

Pathological tissue sections of CAIS and normal cervical canal and gland sections of chronic cervicitis stored in the Pathology Department of Shengjing Hospital, China Medical University from January 2019 to December 2021 were retrospectively collected. After image collection, they were randomly divided into training set, validation set and test set with a ratio of 4∶3∶3. The data of training set and validation set were used to conduct transfer learning training and parameter debugging for 6 network models, including VGG16, VGG19, Inception V3, Xception, ResNet50 and DenseNet201, and the convolutional neural network binary classification model that could recognize pathological images of CAIS was constructed. The models were combined to build the ensemble learning model. Based on the test set data, the performance of pathological image recognition of single model and ensemble learning model was evaluated. The results were expressed by operation time, accuracy, precision, recall, F1 score and area under the curve(AUC) of receiver operating characteristic.

A total of 104 pathological sections of CAIS and 90 pathological sections of normal cervical duct and gland with chronic cervicitis were selected. A total of 500 pathological images of CAIS and normal cervical duct glands were collected, including 400 images of training set, 300 images of validation set and 300 images of test set, respectively. Among the 6 models, ResNet50 model, with the highest accuracy(87.33%), precision(90.00%), F1 score(86.90%) and AUC(0.87), second highest recall(84.00%) and shorter operation time(2062.04 s), demonstrated the best overall performance; VGG19 model was the second; and Inception V3 and Xception model had the worst performance.Among the 6 kinds of ensemble learning models, ResNet50 and DenseNet201 showed the best overall performance, and their accuracy, precision, recall, F1 score and AUC were 89.67%, 84.67%, 94.07%, 89.12% and 0.90, respectively. VGG19 and ResNet50 ensemble model followed.

It is feasible to construct CAIS pathological image recognition models by deep learning algorithm, among which ResNet50 models has higher overall performance. Ensemble learning can improve the recognition effect on pathological images by single model.

To evaluate the clinical application value of the compression and storage algorithm of histopathological images based on matrix computing, and to seek the best image compression ratio.

Two classical matrix algorithms, principal component analysis (PCA) and singular value decomposition (SVD), were used to compress and reconstruct the immunohistochemical images and HE staining images of low, medium and high differentiated cervical cancer tissues. The peak signal-to-noise ratio(PSNR) and structural similarity (SSIM) were used to analyze and evaluate the quality of image reconstruction.

When the compression ratio of PCA reconstruction image was 10.18 (53 principal components were retained), the mean PSNR of immunohistochemical images of low, medium and high differentiated cervical cancer tissues were 43.84±0.43, 43.27±0.25 and 43.71±0.49, respectively, and the SSIM were 0.964±0.004, 0.963±0.006 and 0.965±0.005, respectively. Meanwhile, the mean PSNR of HE staining images of low, medium and high differentiated cervical cancer tissues were 43.41±0.78, 42.95±1.03 and 43.52±0.69, respectively, and the SSIM were 0.953±0.010, 0.949±0.015 and 0.960±0.007, respectively. When the compression ratio of SVD reconstruction image was 10.00(128 singular values were retained), the mean PSNR of immunohistochemical images of low, medium and high differentiated cervical cancer tissues were 39.89±1.69, 38.20±2.19 and 40.90±0.50, respectively, and the SSIM were 0.949±0.006, 0.938±0.011 and 0.955±0.004, respectively. Meanwhile, the mean PSNR of HE staining images of low, medium and high differentiated cervical cancer tissues were 40.31±0.98, 39.46±1.59 and 40.77±1.67, respectively, and the SSIM were 0.965±0.006, 0.943±0.010 and 0.969±0.005, respectively.

PCA and SVD can compress and store histopathological images and obtain better image quality, which provides a solution to the problem of hospital image storage.