This study aims to solve the overfitting problem caused by insufficient labeled images in the automatic image annotation field. We propose a transfer learning model called CNN-2L that incorporates the label localization strategy described in this study. The model consists of an InceptionV3 network pretrained on the ImageNet dataset and a label localization algorithm. First, the pretrained InceptionV3 network extracts features from the target dataset that are used to train a specific classifier and fine-tune the entire network to obtain an optimal model. Then, the obtained model is used to derive the probabilities of the predicted labels. For this purpose, we introduce a squeeze and excitation (SE) module into the network architecture that augments the useful feature information, inhibits useless feature information, and conducts feature reweighting. Next, we perform label localization to obtain the label probabilities and determine the final label set for each image. During this process, the number of labels must be determined. The optimal K value is obtained experimentally and used to determine the number of predicted labels, thereby solving the empty label set problem that occurs when the predicted label values of images are below a fixed threshold. Experiments on the Corel5k multilabel image dataset verify that CNN-2L improves the labeling precision by 18% and 15% compared with the traditional multiple-Bernoulli relevance model (MBRM) and joint equal contribution (JEC) algorithms, respectively, and it improves the recall by 6% compared with JEC. Additionally, it improves the precision by 20% and 11% compared with the deep learning methods Weight-KNN and adaptive hypergraph learning (AHL), respectively. Although CNN-2L fails to improve the recall compared with the semantic extension model (SEM), it improves the comprehensive index of the F1 value by 1%. The experimental results reveal that the proposed transfer learning model based on a label localization strategy is effective for automatic image annotation and substantially boosts the multilabel image annotation performance.
BACKGROUND: In China, traditional Chinese herbal medicine (TCHM) has been widely used for pancreatic cancer. This retrospective, matched case-control study aimed to assess factors affecting the survival time of patients with pancreatic cancer.
METHODS: From 2004 to 2012, a total of 411 patients with pathologically confirmed pancreatic cancer were enrolled, and 272 patients were matched and divided into TCHM and non-TCHM groups (control group) based on received TCHM or not. The match was according to gender, age of onset, radiotherapy, and chemotherapy. Both groups received comprehensive treatments, the TCHM group simultaneously received the TCHM spleen-invigorating compound for more than 3 months. The Cox model was used for prognostic factor analysis and the Kaplan-Meier method for estimating median overall survival (OS) and disease-free survival (DFS).
RESULTS: In 130 patients with advanced pancreatic cancer, COX analysis showed the Karnofsky Performance Scale (KPS; P = .000), radiotherapy (P = .003), and TCHM (P = .001) were independent prognostic factors for OS, with median OS of 12.7 and 9.9 months in TCHM and non-TCHM groups, respectively (hazard ratio [HR] = 0.520; 95% confidence interval [CI] = 0.353-0.766; P = .033). In 142 patients undergoing radical surgery, KPS (P = .000) and TCHM (P = .000) were independent prognostic factors for OS and DFS, median OS was 23.8 and 12.4 months in TCHM and non-TCHM groups, respectively (HR = 0.373; 95% CI = 0.251-0.554; P = .000), and the median DFS was 21.5 and 10.2 months in TCHM and non-TCHM groups, respectively (HR = 0.352; 95% CI = 0.237-0.522; P = .000).
CONCLUSIONS: KPS was an important prognostic factor of pancreatic cancer. Spleen-invigorating compounds could have an effect on improving the prognosis of pancreatic cancer patients.