IOT Based Deep Learning framework to Diagnose Breast Cancer over Pathological Clinical Data

Abstract

Metastasis of breast cancer cells is a critical element in determining a patient's prognosis. A sentinel lymph node biopsy may be used to determine metastases. The standard pathologist examination procedure, on the other hand, is redundant and time consuming, and it is easy to overlook micro metastatic lesions. At the moment, the findings of employing a convolutional neural network to research breast cancer sentinel lymph node metastases have been obtained. Nonetheless, the accuracy rate is low, and the micro metastasis detection impact is poor. A multichannel convolutional neural network model was constructed and suggested in answer to the aforesaid challenges using the sentinel lymph node pathological imaging dataset of breast cancer (PCam). The model employs stacked multichannel convolutional units and IOT based CNN modules, as well as skip cross-layer connections, a mix of conventional and depth wise separable convolutions, and a combination of sum and concatenation operations. Iteratively train 50% of the photos 35 times to produce the model weights. Then, using the accuracy and area under the curve (AUC) values, evaluate the test pictures. Accuracy is 97.32 percent and AUC is 98.05 percent. When compared to the findings of previous research and mainstream convolutional network models, the model scores first in AUC values for 49 percent, 51 percent, and 100% test sets. The findings indicate that the model is very accurate at recognizing lymph node metastasis and performs well at detecting micro metastases.