March 11-12, 2023, Virtual Conference
Shuang Qian1, Yi Zhang, Li Liu, Jun Liao and School of Big Data & Software Engineering, Chongqing University, Chongqing 400030, China
Ki-67 proliferation status is an important clinical indicator for determining the prognosis and treatment of prostate cancer. Most of the existing models are classified based on single-modal data (such as hand crafted features, images), which limits the performance of the model. In this paper, we propose a feature fusion network based on functional constraints to fuse deep learning features and radiomics features to improve the predictive performance of clinical indicators such as Ki-67. Specifically, we design a deep learning network to extract deep features from MP-MRI T2 weighted images, use the feature selectio algorithm RFE to select representative features from high-dimensional radiomics features, and finally the deep features and radiomics features are fused to provide complementary information for classification. Meanwhile, we add the segmentation loss to force the network to pay more attention to the lesion area, and then combine the segmentation loss with the classification loss and integrate it into the backpropagation process of model training. The experimental results on three real datasets show that the predictive model using fusion features has more advantages in the performance of clinical indicators such as Ki-67 status than traditional machine learning methods or deep CNN models relying on single-modal data. We use the class activation map to show the focus of model classification. Compared with other CNN models, the focus area of the proposed model is closer to the actual lesion area, which improves the interpretability of the model classification results to a certain extent.
Prostate Cancer, Ki-67, Radiomics Features, Deep Learning Features, Feature Fusion.
Yanrong Li1,2,3, Shibiao Xu1,2,3, LiGuo1,2,3, Qing Zhang4, Hailong Jin4, 1 School of Artificial Intelligence, University of Posts and Telecommunications, Beijing 100876, China 2National Engineering Research Center for Mobile Internet Security Technology, Beijing University of Posts and Telecommunications, Beijing 100876, China 3 Key Laboratory of Universal Wireless Communications, Ministry of Education, Beijing University of Posts and Telecommunications, Beijing 100876, China 4Department of Organ Transplantation, The Third Medical Center of PLA General Hospital, Beijing 100039, China
Infection prediction after kidney transplantation is significant. Most existing models for predicting kidney transplant infection are statistical, unintelligent, and straightforward. The foremost task of this paper is to analyze kidney transplantation data, introduce existing traditional machine learning and deep learning methods from non-temporal and temporal scenarios, respectively, and comprehensively evaluate the predictive power of the methods for kidney transplantation infection. Specifically, in the non-temporal scenario, we use Naïve Bayes, K-Nearest Neighbor, Support Vector Machines, and Random Forest models. In addition, in the temporal scenario, we propose an MTAN model based on a sliding window algorithm to exploit the hidden information of adjacent time series fully. Experimental results show that the kidney transplantation prediction models built by Naïve Bayes and Support Vector Machines have better stability than those constructed by K-Nearest Neighbor and Random Forest. The MTAN model with sliding windows can better mine the hidden temporal information.
Kidney Transplant Infection Prediction, Traditional Machine Learning, Deep Learning, Sliding Window
Anuj Singh1 and Deepak Ohri2, 1Amity Institute of Biotechnology, Amity University Uttar Pradesh, Lucknow, (U.P), 226028, India 2Amity Institute of Biotechnology, Amity University Uttar Pradesh, Lucknow, (U.P), 226028, India
Lung cancer is the leading cause of cancer-related deaths worldwide, and current pharmacological medicines have limited efficacy due to resistance caused by mutations. This study focuses on finding compounds that can inhibit the BRAF mutation, responsible for 2% to 5% of lung cancer cases. The research utilizes high throughput virtual screening of marine compounds, which have been recognized as a promising research resource. A pharmacophore model was developed to screen a library of 31,561 natural compounds compounds taken from the various marine databases, followed by molecular docking and ADMET investigation. Seven compounds showed high -CDOCKER energy and potential as lead molecules for effective BRAF inhibition..
Structure based Pharspacohore, BRAF, Lung Cancer, Natural compounds