Autism spectrum disorder (ASD), a neurodevelopmental condition, affects approximately 1% of children globally and their social and cognitive abilities. This leads to difficulties in communication, repetitive behaviors, psychomotor skills, and eye contact maintenance. In recent years, there has been an increased utilization of Artificial Intelligence (AI) for the early detection of autism. Using knowledge gathered from 116 peer-reviewed publications, this study assessed algorithmic efficacy, model performance, multimodal data integration, classification metrics, generalization ability, and clinical usefulness. Machine Learning, Deep Learning (DL), Graph Neural Networks, Federated Learning, auto encoders, and Natural Language Processing, the Attention Mechanism seemed to be among the several AI techniques that were investigated in this study. To protect a child's developmental progress, this survey investigates how early detection of autism facilitates prompt therapeutic treatments, minimizes intellectual disabilities, improves mobility, and supports customized care. The research emphasizes the effectiveness of predicting ASD with minimal time using different data modalities, including EEG microstates, ABIDE I & II, to assess DL models such as GoogleNet, Xception, AlexNet, ResNet, VGG, and DenseNet. Three types of data were analyzed: biochemical (biomarkers and physiological metrics), behavioral (facial features, eye gazing, and audio-video cues), and structural and functional (MRI, EEG, and ECG) images. The research demonstrated strong diagnostic performance with models attaining accuracy rates ranging from 90% to 96% across diverse datasets. Classification measures, such as accuracy, sensitivity, specificity, precision, recall, and F1-score, were used in the performance evaluation. Error and statistical metrics, such as RMSE, MSE, R2, Kappa, and G-mean, were also employed. The dependability and efficiency of the models in detecting ASD were enhanced by validation methods, such as confusion matrix, receiver operating characteristics (ROC) curves, and AUC. Based on the evaluated studies, transfer learning with diverse datasets and modalities has great promise for early ASD diagnosis. Even with a minimal data size, these techniques increase robustness, accuracy, and generalization. For real-time clinical applications, a hybrid transfer learning-based framework is advised to assist clinicians and therapists in accurately diagnosing and assessing ASD severity.
Stomach abnormalities such as ulcers, tumors, and inflammation often progress silently, remaining undetected until they reach advanced stages, which significantly reduce treatment effectiveness. Traditional diagnostic methods rely heavily on invasive procedures and often fail to detect early signs, especially when symptoms are minimal or absent. There is a critical need for non-invasive, accurate, and early detection techniques that can analyze diverse patient data effectively. Therefore, this research proposes an innovative ensemble architecture combining Bi-GRU, DBN, and RNN (En-BiDBRNN) models for stomach abnormality detection with an improved segmentation model. Each modality of input data (patient data) and iris images undergoes individual preprocessing steps. Subsequently, improved balanced iterative reducing and clustering using hierarchy-based segmentation is applied to the filtered images. Moreover, feature extraction and augmentation take place. Then, the features are processed independently using bidirectional gated recurrent unit, deep belief network, and recurrent neural network for stomach abnormality detection. Finally, the scores obtained from these models are combined using an enhanced score-level fusion method. Experimental results demonstrate that En-BiDBRNN significantly outperforms existing models across various training data sizes, achieving up to 96% accuracy at 90% training data. These findings validate the model's robustness, positioning En-BiDBRNN as a promising solution for early stomach abnormality detection.
Skin disease is a disorder that damages the skin by causing irritation, inflammation, or abnormal growth. In the olden days, there were limited diagnostic tools, which made it difficult to identify and treat the disease accurately. Therefore, a novel Convolutional Dense Network (Conv-DenNet) is designed for the detection of skin cancer. The Conv-DenNet model combines Convolutional eXtreme Gradient Boosting (ConvXGB) and Dense Network (DenseNet). Initially, the input skin image is denoised using the Medav filter. Following this, skin lesion segmentation is performed via the Recurrent Residual U-Net integrated with TopK Focal loss (TKF + R2U-Net). Here, the R2U-Net model is fine-tuned via Artificial Namib Beetle Optimization (ANBO), which integrates the Artificial Hummingbird Algorithm (AHA) with Namib Beetle Optimization (NBO). Moreover, image augmentation is applied using a combination of geometric and color space transformations, along with Mixup and CutMix. Following this, the images are analyzed to extract statistical and Haralick features. Finally, skin cancer detection is performed using the Conv-DenNet model. Experimental results show that Conv-DenNet achieves 93.46% accuracy, 93.78% True Positive Rate (TPR), 94.37% precision and 6.38% False Positive Rate (FPR), under 5 leave-one-out cross-validation. The study shows that the Conv-DenNet model improves the detection of skin cancer relative to the conventional methods.
The process of natural language processing (NLP) to classify and identify sentiments as negative, positive, or neutral automatically is called sentiment classification. Sentiment Analysis improves the service and product based on customer feedback. Sentiment Analysis is beneficial for business operations for analyzing customer feedback to improve customer satisfaction and product development. However, there are some limitations, such as overfitting issues, flexibility, interpretability, generalization, stability, low energy efficiency, and complexity. The traditional methods do not capture the contextual semantics of the sentence, which leads to poor sentiment classification in the presence of sarcasm, irony, or idiomatic expressions. Also, many of the methods ignore the textural cues like hashtags, all-caps words, punctuation, and negations. To overcome above-mentioned issues, an efficient hierarchical Kronecker deep learning network (HKDL-Net) is established. Here, HKDL-Net is developed by the combination of hierarchical deep learning for text classification (HDLTex) and deep Kronecker network (DKN). Initially, reviews are subjected to bidirectional encoder representation from transformers (BERT) tokenization to split text into specific words. Then, aspect term extraction (ATE) is performed. Afterwards, feature extraction is executed by using extracted aspect terms for mining features, such as emotion, hashtag, all-caps, number of sentences in review, elongated units, numerical words, a bag of units, question marks, number of words in a review, and negation. The sentiment classification is ultimately achieved through the proposed HKDL-Net. Moreover, HKDL-Net attained better results in sentiment classification with a precision of 91.877%, recall of 93.899%, F1-score of 92.877%, and accuracy of 97.568%, respectively. This shows that the devised model effectively detects the non-standard language and enhances the classification performance of sentiments in subtle or complex cases. Also, it can capture contextual relationships between words, which helps to improve its ability to understand sarcasm, irony, or idiomatic expressions.
China boasts a diverse range of crops. The yield and the quality of crops exert an influence on economic development and farmers’ income. The healthy and stable development of agriculture has consistently been a goal pursued by the country. Therefore, it is of great significance to rapidly and precisely identify the types of crop diseases and pests and offer timely and accurate prevention and control measures. The recognition of traditional crop pest and disease images mainly relies on agricultural experts to recognize the color and shape of disease images. This approach is not only labor-intensive and inefficient but also subjective, and it can easily lead to misjudgment and cause the optimal treatment time to be missed. Potatoes are crucial for food processing and daily consumption. They are a representative crop that is also prone to significant economic losses due to pests and diseases. In order to solve the problem of identifying crop pest and disease images, this article combines image processing and VGG deep-learning techniques to implement an improved computer vision model and conducts experiments on seven types of pest and disease images of potato leaves. The experimental results show that compared with the traditional VGG16 model, the new model has a higher identification rate of pests and diseases on potato leaves, and it also has significant reference value for the pest and disease identification methods of other leaf crops.
