The most prevalent medical disorders in the healthcare industry are lung ailments. Lung disease is often diagnosed by a professional utilizing image from a chest X-ray and visual inspection. Because it is manual, it might take a lot of time and result in incorrect diagnoses. Deep learning techniques require a lot of data. Small amounts of data, however, won't produce accurate results. This research seeks to employ meta learning technique to overcome this problem. Few-shot learning, one of the meta learning techniques, has wide scope in machine learning applications in recent times. Very initially, our proposed framework would perform pre-processing over the raw dataset, which utilize Contrast Limited Adaptive Histogram Equalization (CLAHE). The pre-processed image will be given input as segmentation. The UNET++ model is used in the segmentation step to give results that can precisely separate the lung nodules. The segmented lung nodules will then be fed into a feature extraction transfer learning model. The necessary features are extracted using the feature extraction. The CheXNet is used as a feature extraction model which was pre-trained using the Chest X-ray14 dataset, which contains 14 chest abnormalities. Finally, the extracted features are given as input to the classifier to detect the various abnormality and normal condition. Few-shot learning techniques are utilized here as the classifier, which would efficiently classify the images using the limited data.