Deep Learning Approaches for Automatic Localization in Medical Images

Identification of kidney stones in KUB X-ray images using VGG16 empowered with explainable artificial intelligence

A kidney stone is a solid formation that can lead to kidney failure, severe pain, and reduced quality of life from urinary system blockages. While medical experts can interpret kidney-ureter-bladder (KUB) X-ray images, specific images pose challenges for human detection, requiring significant analysis time. Consequently, developing a detection system becomes crucial for accurately classifying KUB X-ray images. This article applies a transfer learning (TL) model with a pre-trained VGG16 empowered with explainable artificial intelligence (XAI) to establish a system that takes KUB X-ray images and accurately categorizes them as kidney stones or normal cases. The findings demonstrate that the model achieves a testing accuracy of 97.41% in identifying kidney stones or normal KUB X-rays in the dataset used. VGG16 model delivers highly accurate predictions but lacks fairness and explainability in their decision-making process. This study incorporates the Layer-Wise Relevance Propagation (LRP) technique, an explainable artificial intelligence (XAI) technique, to enhance the transparency and effectiveness of the model to address this concern. The XAI technique, specifically LRP, increases the model's fairness and transparency, facilitating human comprehension of the predictions. Consequently, XAI can play an important role in assisting doctors with the accurate identification of kidney stones, thereby facilitating the execution of effective treatment strategies.

MGACA-Net: a novel deep learning based multi-scale guided attention and context aggregation for localization of knee anterior cruciate ligament tears region in MRI images

Anterior cruciate ligament (ACL) tears are a common knee injury that can have serious consequences and require medical intervention. Magnetic resonance imaging (MRI) is the preferred method for ACL tear diagnosis. However, manual segmentation of the ACL in MRI images is prone to human error and can be time-consuming. This study presents a new approach that uses deep learning technique for localizing the ACL tear region in MRI images. The proposed multi-scale guided attention-based context aggregation (MGACA) method applies attention mechanisms at different scales within the DeepLabv3+ architecture to aggregate context information and achieve enhanced localization results. The model was trained and evaluated on a dataset of 917 knee MRI images, resulting in 15265 slices, obtaining state-of-the-art results with accuracy scores of 98.63%, intersection over union (IOU) scores of 95.39%, Dice coefficient scores (DCS) of 97.64%, recall scores of 97.5%, precision scores of 98.21%, and F1 Scores of 97.86% on validation set data. Moreover, our method performed well in terms of loss values, with binary cross entropy combined with Dice loss (BCE_Dice_loss) and Dice_loss values of 0.0564 and 0.0236, respectively, on the validation set. The findings suggest that MGACA provides an accurate and efficient solution for automating the localization of ACL in knee MRI images, surpassing other state-of-the-art models in terms of accuracy and loss values. However, in order to improve robustness of the approach and assess its performance on larger data sets, further research is needed.

Construction of diagnostic and prognostic models based on gene signatures of nasopharyngeal carcinoma by machine learning methods

Background

Diagnostic models based on gene signatures of nasopharyngeal carcinoma (NPC) were constructed by random forest (RF) and artificial neural network (ANN) algorithms. Least absolute shrinkage and selection operator (Lasso)-Cox regression was used to select and build prognostic models based on gene signatures. This study contributes to the early diagnosis and treatment, prognosis, and molecular mechanisms associated with NPC.

Methods

Two gene expression datasets were downloaded from the Gene Expression Omnibus (GEO) database, and differentially expressed genes (DEGs) associated with NPC were identified by gene expression differential analysis. Subsequently, significant DEGs were identified by a RF algorithm. ANN were used to construct a diagnostic model for NPC. The performance of the diagnostic model was evaluated by area under the curve (AUC) values using a validation set. Lasso-Cox regression examined gene signatures associated with prognosis. Overall survival (OS) and disease-free survival (DFS) prediction models were constructed and validated from The Cancer Genome Atlas (TCGA) database and the International Cancer Genome Consortium (ICGC) database.

Results

A total of 582 DEGs associated with NPC were identified, and 14 significant genes were identified by the RF algorithm. A diagnostic model for NPC was successfully constructed using ANN, and the validity of the model was confirmed on the training set AUC =0.947 [95% confidence interval (CI): 0.911-0.969] and the validation set AUC =0.864 (95% CI: 0.828-0.901). The 24-gene signatures associated with prognosis were identified by Lasso-Cox regression, and prediction models for OS and DFS of NPC were constructed on the training set. Finally, the ability of the model was validated on the validation set.

Conclusions

Several potential gene signatures associated with NPC were identified, and a high-performance predictive model for early diagnosis of NPC and a prognostic prediction model with robust performance were successfully developed. The results of this study provide valuable references for early diagnosis, screening, treatment and molecular mechanism research of NPC in the future.

The Establishment of Hypertrophic Cardiomyopathy Diagnosis Model via Artificial Neural Network and Random Decision Forest Method

Hypertrophic cardiomyopathy is a hereditary disease characterized by asymmetric ventricular hypertrophy as the key anatomical feature. Currently, there exists no effective method for the early diagnosis of hypertrophic cardiomyopathy. In this analysis, we incorporated multiple GEO datasets containing RNA profiles of hypertrophic cardiomyopathic patient tissues, identified 642 differentially expressed genes, and performed GO and KEGG analyses. Furthermore, we narrowed down 46 characteristic genes from these differentially expressed genes using random decision forests and conducted transcription factor regulation analysis on them. Using 40 genes that showed overlap between the training set and the verification set, the artificial neural network was trained, and the final MPS scoring model was constructed, and a receiver-operating characteristic (ROC) curve was drawn. We used the MPS model to predict the verification dataset and drew the ROC curve, which demonstrated the good prediction performance of the model. In conclusion, this study combines a random decision forest and artificial neural network to build a diagnostic model for hypertrophic cardiomyopathy to predict the disease, aiming at early detection and treatment, prolonging the survival time, and improving the quality of life of patients.