Medical Image Classification Using Thermal Images and Diagnosis of Carcinogen

Lung and Colon (L&C) tumors are lethal sicknesses that can foster in a few organs all the while and, in specific circumstances, jeopardize human existence. Even though it is highly unlikely that these two types of cancer will develop simultaneously, delaying diagnosis significantly increases the likelihood of metastasis between the organs that are affected. To really treat specific sorts of malignant growth, histological determination is fundamental. In the past, doctors had to go through a long and difficult process to look at thermal images and figure out if a patient had cancer; however, this procedure may now be completed much more quickly thanks to the new technology options. Histological images of L&C tumors were classified using a hybrid Deep Learning (DL) model with an attention mechanism and a multipath network in this study. To zero in on the main attributes and negligence the less significant ones, a consideration component was utilized. In a multipath network, data travels over a number of channels before each channel is converted and the output from all of the branches is combined. The multipath network is similar to grouped convolution when simplified. The five thermal image categories of the LC25000 dataset were utilized. Among these classes were two for colon disease and three for cellular breakdown in the lungs. The proposed model was compared to a number of well-known DL models, such as ResNet-50, VGG-16, and AlexNet. The proposed approach showed the best exhibition concerning exactness (99.2%), particularity (99.12%), responsiveness (99.28%), accuracy (99.12%), and F1 score (99.2%), as indicated by the exploratory discoveries