Skin cancer is one of the widespread diseases among existing cancer types. More importantly, the detection of lesions in early diagnosis has tremendously attracted researchers’ attention. Thus, artificial intelligence (AI)-based techniques have supported the early diagnosis of skin cancer by investigating deep-learning-based convolutional neural networks (CNN). However, the current methods remain challenging in detecting melanoma in dermoscopic images. Therefore, in this paper, we propose an ensemble model that uses the vision of both EfficientNetV2S and Swin-Transformer models to detect the early focal zone of skin cancer. Hence, we considerthat the former architecture leads to greater accuracy, while the latter model has the advantage of recognizing dark parts in an image. We have modified the fifth block of the EfficientNetV2S model and have included the Swin-Transformer model. Our experiments demonstrate that the constructed ensemble model has attained a higher level of accuracy over the individual models and has also decreased the losses as compared to traditional strategies. The proposed model achieved an accuracy score of 99.10%, a sensitivity of 99.27%, and a specificity score of 99.80%.

Infectious disease-related illness has always posed a concern on a global scale. Each year, pneumonia (viral and bacterial pneumonia), tuberculosis (TB), COVID-19, and lung opacity (LO) cause millions of deaths because they all affect the lungs. Early detection and diagnosis can help create chances for better care in all circumstances. Numerous tests, including molecular tests (RT-PCR), complete blood count (CBC) tests, Monteux tuberculin skin tests (TST), and ultrasounds, are used to detect and classify these diseases. However, these tests take a lot of time, have a 20% mistake rate, and are 80% sensitive. So, with the aid of a doctor, radiographic tests such as computed tomography (CT) and chest radiograph images (CRIs) are used to detect lung disorders. With CRIs or CT-scan images, there is a danger that the features of various lung diseases’ diagnoses will overlap. The automation of such a method is necessary to correctly classify diseases using CRIs. The key motivation behind the study was that there is no method for identifying and classifying these (LO, pneumonia, VP, BP, TB, COVID-19) lung diseases. In this paper, the DeepLungNet deep learning (DL) model is proposed, which comprises 20 learnable layers, i.e., 18 convolution (ConV) layers and 2 fully connected (FC) layers. The architecture uses the Leaky ReLU (LReLU) activation function, a fire module, a maximum pooling layer, shortcut connections, a batch normalization (BN) operation, and group convolution layers, making it a novel lung diseases classification framework. This is a useful DL-based method for classifying lung disorders, and we tested the effectiveness of the suggested framework on two datasets with a variety of images from different datasets. We have performed two experiments: a five-class classification (TB, pneumonia, COVID-19, LO, and normal) and a six-class classification (VP, BP, COVID-19, normal, TB, and LO). The suggested framework’s average accuracy for classifying lung diseases into TB, pneumonia, COVID-19, LO, and normal using CRIs was an impressive 97.47%. We have verified the performance of our framework on a different publicly accessible database of images from the agriculture sector in order to further assess its performance and validate its generalizability. This study offers an efficient and automated method for classifying lung diseases that aids in the early detection of lung disease. This strategy significantly improves patient survival, possible treatments, and limits the transmission of infectious illnesses throughout society.