IA2SKAbs: Intelligent automated and accurate system for classification of kidney abnormalities

The occurrence of kidney abnormalities (K.Abs) prevalence is elevating which have noxious and deleterious affects all over the world. An automated detection of K.Abs accurately detects and classify with labels by implementing latest DL method with transfer learning approaches. The prior methods were expensive, tedious, arduous and less efficient to detect K.Abs from noisy low frequency images with limited no. of classes and small size dataset. The proposed IA2SKAbs research introduced two effective automated model's 1st was Efficient-b0 and 2nd ResNet-18 which resolve all abovementioned problems expeditiously. Both models enhance the overall accuracy and classification accuracy with large dataset (12446) of CT images along with a greater number of classes which efficiently work in contrast to prior research. Both pre-trained models were trained by altering last three layers or using them from scratch like fully Connected, SoftMax and Classification output layers according to the size of classification problem. Before training of model's dataset was pre-processed with
$224\times 224$
dimensions which are according to the used models. The training accuracy of Efficient-b0 and ResNet-18 99.92% and 99.97%; the overall classification accuracy of both models was 99.99% and 98.1% respectively. The implementation detail of system architecture is discussed below in section 3.