DETERMINATION OF AREA OF THE TUMOR WITH DEEP LEARNING TECHNIQUES

Abstract

In the realm of medical imaging, tumor identification and categorization are crucial for a variety of reasons. First and foremost, effective cancer treatment and management depend greatly on a prompt and accurate tumor diagnosis. The recommended therapy will be very effective due to the early tumor discovery, and there may be a chance of increase in the patient's survival rate. The clinical professionals can access the tumor's size, form, and location with the tumor segmentation method. This information will be very crucial for planning the better treatment which may include surgery, radiation therapy or chemo therapy. The tumor's stage can be determined by computing the tumor's area. Patients may experience variations in the tumor's minute characteristics as well as variations in tumor size. In some cases an MRI picture of the cerebral fluid may also appear as a mass of tissue. The identification of the smallest tumor along with the minute details of the tumor is required to be obtained from the image and this can be obtained using U-Net which is used for semantic segmentation. The Modified U-Net using ResNet models are employed for the smallest area calculation and the tumor as small as 105 pixels is obtained using this method. These small tumors are less likely to create visible symptoms and may not even show up on conventional imaging methods until they are greatly expanded in size. The Modified U-Net with ResNet as encoder and decoder are employed for the tumor area calculation by which the patient can undergo effective treatment and improve prognosis.