Classification of Kidney Diseases Using Transfer Learning

Classification of Kidney Diseases Using Transfer Learning

Sachin Kumar Saxena, Jitendra Nath Shrivastava, Gaurav Agarwal, Sanjay Kumar
DOI: 10.4018/978-1-6684-6821-0.ch004
OnDemand:
(Individual Chapters)
Available
$37.50
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

A urologist confirms high risks of kidney stones just because of diabetic mellitus; however, other factors also exist, but a major cause is type 2 diabetes. Renal cyst and diabetes clinical features show 58% of affected subjects as the same. Research findings prove the high risk of renal cancer among diabetes patients. All these patients underwent abdominal MRI or CT scan to extract kidney high-definition 3D images. The dataset was gathered from two hospitals: the first is the SRMS IMS, and the second is the Bareilly MRI & CT Scan Centre, both located in the city of Bareilly in the state of Uttar Pradesh of India. Research has been analyzed to note the classification among four classes using seven transfer learning methods. Results have been compared with seven transfer learning methods. The methods are EfficientNetB0, Xception, VGG16, ResNet50, MobileNet, InceptionV3, DenseNet121. Out of these deep learning-based algorithms, EfficientNetB0 shows the best accuracy of 96.02%.
Chapter Preview
Top

Introduction

There is a strong correlation between chronic kidney diseases and medical history showing that diabetic Mellitus causes diabetic symptoms. A high level of blood glucose level leads to CKD, cardiac issues, liver problems, nerve damage, etc. The asymptotic phase of renal disorder is crucial because the patient is not aware of the consequences and avoids taking precautionary steps to protect his kidney. In the later stage of CKD, symptoms are worse than vomiting, cardiac attack, retina damage, leg swelling, and congestive cardiac failure. The renal calculi, also called kidney stones, are hard particles stored at a single place or multi places, resulting in diminished functionality of the urinary system. Various kinds of kidney stones are existing in the human body, such as calcium depository, which is due to the excess amount of calcium in urine, the second type is uric acid stone, due to high level of acid existing in urine, third is struvite stone, which is due to infection in the urinary system, the fourth type is cysteine, which is due to family hierarchy disorder. Hypertension, older age, family history of kidney cancer, smoke, obesity, etc. are the main cause of kidney cancer, which is the worst case of kidney tumor. However, medical research does not claim any certain reason why kidney tumor particles are developed in the renal system.

In recent research, the association between kidney stones and diabetic Mellitus has come into the picture. Patients with type 2 diabetes are more prone to develop uric acid stones. There is a 33.9 percent risk factor of renal calculi or Nephrolithiasis inside the kidney of patients who are also suffering from type 2 diabetes (Daudon et al., 2006). A near about, 2.5 times is the risk factor to be developed kidney stones in the renal of type 2 diabetic patients. In addition, HbA1c ranges from 5.7 to 6.4 percent have a moderate chance of 34 percent, and HbA1c ranges above 6.4 percent have 92 percent to develop kidney stones (NICResearch, 2019). Based on the study of causes of renal calculi, an imbalance of insulin instances in the body, and high blood pressure levels are the major roots of Nephrolithiasis (Vieira, 2020). Kidney cancer or renal tumor is also one of the causes of kidney function deterioration. Tseng et al. (2015) extensively examined one million kidney cancer patients with diabetic and non-diabetic symptoms. They found a high number of cancer patients, who were also diagnosed with diabetic Mellitus. The ratio later hiked, when other parameters such as age, the high blood pressure were also added. The patients with type 2 diabetic Mellitus have much higher risk for kidney cancer as compared to other factors (Undzyte et al., 2020).

Figure 1.

Cross section MRI image view

978-1-6684-6821-0.ch004.f01
Figure 2.

Patient abdominal raw CT scan image

978-1-6684-6821-0.ch004.f02
Top

Motivation

In the training stage, Kears API integrated with Python code to achieve the Deep Learning feature extractctions.

All the models have been trained on NVIDIA GeForce Experience recetly purchased by Shri Ram Murti Smarak College of Engineering and Technology, Bareilly, Uttar Pradesh, India. Further details are given in Table 1 GPU hardware configuration .

Table 1.
GPU hardware configuration
Hardware DetailConfiguration
NVIDIA model number3060
RAM12 GB
SSD Graphic card500 NVME
System Opearting systemWindows 10 Pro 64 Bits
Processor ModelIntel (R) Core (TM) i7-10700K
CPU prcessor3.8 GHz
System RAM32 GB

Complete Chapter List

Search this Book:
Reset