Survival Risk Prediction of Esophageal Cancer Based on the Kohonen Network Clustering Algorithm and Kernel Extreme Learning Machine

Prediction of five-year survival among esophageal cancer patients using machine learning

Background and aim

Considering the silent progression of esophageal cancer, the survival prediction of this disease is crucial in enhancing the quality of life of these patients globally. So far, no prediction solution has been introduced for the survival of EC in Iran based on the machine learning approach. So, this study aims to develop a prediction model for the five-year survival of EC based on the ML approach to promote clinical outcomes and various treatment and preventive plans.

Material and methods

In this retrospective study, we investigated the 1656 cases of survived and non-survived EC patients belonging to Imam Khomeini Hospital in Sari City from 2013 to 2020. The multivariable regression analysis was used to select the best predictors of five-year survival. We leveraged random forest, eXtreme Gradient Boosting, support vector machine, artificial neural networks, Bayesian networks, J-48 decision tree, and K-nearest neighborhood to develop the prediction models. To get the best model for predicting the five-year survival of EC, we compared them using the area under the receiver operator characteristics.

Results

The age at diagnosis, body mass index, smoking, obstruction, dysphagia, weight loss, lymphadenopathy, chemotherapy, radiotherapy, family history of EC, tumor stage, type of appearance, histological type, grade of differentiation, tumor location, tumor size, lymphatic invasion, vascular invasion, and platelet albumin ratio were considered as the best predictors associated with the five-year survival of EC based on the regression analysis. In this respect, the random forest with the area under the receiver operator characteristics of 0.95 was identified as a superior model.

Conclusion

The experimental results of the current study showed that the random forest could have a significant role in enhancing the quality of care in EC patients by increasing the effectiveness of follow-up and treatment measures introduced by care providers.

Early prediction of response to palliative chemotherapy in patients with stage-IV gastric and esophageal cancer

BACKGROUND

The goal of therapy for many patients with advanced stage malignancies, including those with metastatic gastric and esophageal cancers, is to extend overall survival while also maintaining quality of life. After weighing the risks and benefits of treatment with palliative chemotherapy (PC) with non-curative intent, many patients decide to pursue treatment. It is known that a subset of patients who are treated with PC experience significant side effects without clinically significant survival benefits from PC.

METHODS

We use data from 150 patients with stage-IV gastric and esophageal cancers to train machine learning models that predict whether a patient with stage-IV gastric or esophageal cancers would benefit from PC, in terms of increased survival duration, at very early stages of the treatment.

RESULTS

Our findings show that machine learning can predict with high accuracy whether a patient will benefit from PC at the time of diagnosis. More accurate predictions can be obtained after only two cycles of PC (i.e., about 4 weeks after diagnosis). The results from this study are promising with regard to potential improvements in quality of life for patients near the end of life and a potential overall survival benefit by optimizing systemic therapy earlier in the treatment course of patients.

The Technology-Oriented Pathway for Auxiliary Diagnosis in the Digital Health Age: A Self-Adaptive Disease Prediction Model

The advent of the digital age has accelerated the transformation and upgrading of the traditional medical diagnosis pattern. With the rise of the concept of digital health, the emerging information technologies, such as machine learning (ML) and data mining (DM), have been extensively applied in the medical and health field, where the construction of disease prediction models is an especially effective method to realize auxiliary medical diagnosis. However, the existing related studies mostly focus on the prediction analysis for a certain disease, using models with which it might be challenging to predict other diseases effectively. To address the issues existing in the aforementioned studies, this paper constructs four novel strategies to achieve a self-adaptive disease prediction process, i.e., the hunger-state foraging strategy of producers (PHFS), the parallel strategy for exploration and exploitation (EEPS), the perturbation-exploration strategy (PES), and the parameter self-adaptive strategy (PSAS), and eventually proposes a self-adaptive disease prediction model with applied universality, strong generalization ability, and strong robustness, i.e., multi-strategies optimization-based kernel extreme learning machine (MsO-KELM). Meanwhile, this paper selects six different real-world disease datasets as the experimental samples, which include the Breast Cancer dataset (cancer), the Parkinson dataset (Parkinson's disease), the Autistic Spectrum Disorder Screening Data for Children dataset (Autism Spectrum Disorder), the Heart Disease dataset (heart disease), the Cleveland dataset (heart disease), and the Bupa dataset (liver disease). In terms of the prediction accuracy, the proposed MsO-KELM can obtain ACC values in analyzing these six diseases of 94.124%, 84.167%, 91.079%, 72.222%, 70.184%, and 70.476%, respectively. These ACC values have all been increased by nearly 2-7% compared with those obtained by the other models mentioned in this paper. This study deepens the connection between information technology and medical health by exploring the self-adaptive disease prediction model, which is an intuitive representation of digital health and could provide a scientific and reliable diagnostic basis for medical workers.

Atom Search Optimization with the Deep Transfer Learning-Driven Esophageal Cancer Classification Model

Esophageal cancer (EC) is a commonly occurring malignant tumor that significantly affects human health. Earlier recognition and classification of EC or premalignant lesions can result in highly effective targeted intervention. Accurate detection and classification of distinct stages of EC provide effective precision therapy planning and improve the 5-year survival rate. Automated recognition of EC can aid physicians in improving diagnostic performance and accuracy. However, the classification of EC is challenging due to identical endoscopic features, like mucosal erosion, hyperemia, and roughness. The recent developments of deep learning (DL) and computer-aided diagnosis (CAD) models have been useful for designing accurate EC classification models. In this aspect, this study develops an atom search optimization with a deep transfer learning-driven EC classification (ASODTL-ECC) model. The presented ASODTL-ECC model mainly examines the medical images for the existence of EC in a timely and accurate manner. To do so, the presented ASODTL-ECC model employs Gaussian filtering (GF) as a preprocessing stage to enhance image quality. In addition, the deep convolution neural network- (DCNN-) based residual network (ResNet) model is applied as a feature extraction approach. Besides, ASO with an extreme learning machine (ELM) model is utilized for identifying the presence of EC, showing the novelty of the work. The performance of the ASODTL-ECC model is assessed and compared with existing models under several medical images. The experimental results pointed out the improved performance of the ASODTL-ECC model over recent approaches.

A computational method for large-scale identification of esophageal cancer-related genes

The incidence of esophageal cancer has obvious genetic susceptibility. Identifying esophageal cancer-related genes plays a huge role in the prevention and treatment of esophageal cancer. Through various sequencing methods, researchers have found only a small number of genes associated with esophageal cancer. In order to improve the efficiency of esophageal cancer genetic susceptibility research, this paper proposes a method for large-scale identification of esophageal cancer-related genes by computational methods. In order to improve the efficiency of esophageal cancer genetic susceptibility research, this paper proposes a method for large-scale identification of esophageal cancer-related genes by computational methods. This method fuses graph convolutional network and logical matrix factorization to effectively identify esophageal cancer-related genes through the association between genes. We call this method GCNLMF which achieved AUC as 0.927 and AUPR as 0.86. Compared with other five methods, GCNLMF performed best. We conducted a case study of the top three predicted genes. Although the association of these three genes with esophageal cancer has not been reported in the database, studies by other reseachers have shown that these three genes are significantly associated with esophageal cancer, which illustrates the accuracy of the prediction results of GCNLMF.

Clustering Algorithm with a Greedy Agglomerative Heuristic and Special Distance Measures

Automatic grouping (clustering) involves dividing a set of objects into subsets (groups) so that the objects from one subset are more similar to each other than to the objects from other subsets according to some criterion. Kohonen neural networks are a class of artificial neural networks, the main element of which is a layer of adaptive linear adders, operating on the principle of “winner takes all”. One of the advantages of Kohonen networks is their ability of online clustering. Greedy agglomerative procedures in clustering consistently improve the result in some neighborhood of a known solution, choosing as the next solution the option that provides the least increase in the objective function. Algorithms using the agglomerative greedy heuristics demonstrate precise and stable results for a k-means model. In our study, we propose a greedy agglomerative heuristic algorithm based on a Kohonen neural network with distance measure variations to cluster industrial products. Computational experiments demonstrate the comparative efficiency and accuracy of using the greedy agglomerative heuristic in the problem of grouping of industrial products into homogeneous production batches.