An Experimental Analysis of Lung Cancer Classification Based on Autoencoder on CT Scan Images

When it comes to medical picture recognition, convolutional neural networks are now the gold standard. The mistake in recreating the CT picture is responsible for the low overall accuracy in lung tumor prediction and the lengthy processing time. The use of multispace images in the pooling layer of a convolutional neural network is proposed here in an effort to boost overall prediction accuracy while simultaneously decreasing processing time. Using multispace images in the pooling layer of a convolutional neural network and an Adam Procedure for optimization, the suggested technique is able to detect lung cancer with high precision. The Autoencoder system contributes to this success. The CT images were first fed into a convolution filter and down sampled using max pooling as a first step in the processing pipeline. Autoencoder models based on convolutional neural networks are utilized to extract features, and multispace image reconstruction techniques are employed to minimize reconstruction errors, resulting in greater precision for predicting lung cancer. Python Tensor Flow was used to analyze the state-of-the-art and recommended remedies, and the results show a considerable improvement (from 91 to 87) in accuracy when classifying lung cancer. The suggested approach boasts the lowest error rate alongside the best accuracy, sensitivity, and specificity.The suggested technique reduces processing time while maintaining superior categorization accuracy, as contrasted to the state of the art. Low-pixel images may be analyzed for categorization evaluation now that the huge dataset is operational.