Volume 10, Issue 3, March – 2025 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/25mar1609

Streamlining Kidney Stone Detection through

Image Processing and Deep Learning
B Raju1; G Abhilash2
1
Department of ECE, S R Gudlavalleru Engineering College, Gudlavalleru, India
2
Department of ECE, S R Gudlavalleru Engineering College, Gudlavalleru, India

Publication Date: 2025/04/11

Abstract: In this paper an automated system for precise kidney stone identification in computed tomography CT scans is
introduced. The system consists of two essential components custom CNN for image classification and fuzzy c-means FCM
clustering for localizing stones, the CNN architecture is trained to classify kidney CT scans as normal or abnormal scans
based on a dataset collected from Kaggle. Subsequently FCM clustering is then applied on the abnormal images to
automatically detect and localize kidney stones by segmenting pixels of the same intensity. This computer-assisted method
applying machine learning-based image processing should yield better accuracy over traditional manual techniques like
thresholding filtering and edge detection. By automating the detection process this system aims to provide radiologists and
urologists with an effective tool for rapid and accurate identification of kidney stones enabling effective and timely patient
care. The project is simulated and implemented on MATLAB software.

Keywords: Kidney Stones, Image Processing, CT Image, CNN, Deep Learning Algorithm, Classification, Accuracy.

How to Cite: B Raju; G Abhilash (2025). Streamlining Kidney Stone Detection through Image Processing and Deep Learning.
International Journal of Innovative Science and Research Technology, 10(3), 2576-2581.
https://doi.org/10.38124/ijisrt/25mar1609

I. INTRODUCTION scans into normal and abnormal ones fuzzy c-means FCM
clustering is applied to detect kidney stones from the
Renal calculi or kidney stones, are a critical health abnormal images automating the detection process. The
problem which causes severe pain and complications. If they system seeks to provide radiologists and urologists with an
are not appropriately treated these solid objects that are efficient tool for rapid and accurate detection of kidney stones
mostly crystals need on time and accurate diagnosis in order ultimately improving patient care. The implementation of the
to be able to manage the patient properly. Computed system is done employing MATLAB a popular tool for
Tomography CT scan is a well-established cornerstone of the medical image processing and analysis.
diagnosis of kidney stones with accurate anatomical imaging
interpreting. The CT image however is still hampered by II. LITERATURE SURVEY
native noise and by the complex architecture of renal tissues
making it prone to error and longer reading time. The area of kidney stone detection via medical imaging
has witnessed extensive research by scientists, while
When performed manually the necessity for traditional methods of diagnosis have proven useful their
computerized systems in medical imaging, particularly in reliance on human interpretation introduces variability and
kidney stone detection is brought into focus by manual delay advances in image processing. Deep learning have
interpretation limitations. As noted in current studies opened the door to more computerized and precise detection
application of manual methods can lead to variable diagnostic methods. Different research works have investigated
accuracy and performance. This is particularly essential in traditional as well as machine learning-based approaches for
emergency situations where speed and precision of diagnosis the detection of kidney stones. For instance, image processing
are paramount. Increasing prevalence of urolithiasis and methods were used by Jamkhandikar and colleagues. Using
resultant growth in the use of CT imaging further points image processing techniques, Suresh M.B. and M.R.
towards the requirement for computerized systems. This Abhishek [1] also looked into renal stone identification. So as
research aims to address such limitations by developing an to automatically detect kidney stones, Manoj, B., Neethu
automatic kidney stone detection system on CT images. The Mohan, and Sachin Kumar [2] utilised deep models.
system presented here is a mixture of deep learning and fuzzy Convolutional neural networks for deep learning medical
clustering techniques to enhance diagnostic efficiency and image analysis has changed dramatically as a result of CNN.
accuracy preliminary classification of the CT images is done A deep learning model based on VGG16 was proposed by
using a convolutional neural network CNN. Dividing the Valarmathi, N., et al. [3] for kidney stone computer-aided

IJISRT25MAR1609 www.ijisrt.com 2576

Volume 10, Issue 3, March – 2025 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/25mar1609
detection. Additionally, the use of a fully automated deep  Acquisition of Images
learning system for bone age estimation highlighting was
investigated by Lee H. Tajmir et al [7]. The potential of such  Preprocessing
systems in medical imaging diagnosis based on NHANES
data, this study put into the limelight the growing incidence  Classification
of kidney stones particularly among childbearing women.
The total space of deep learning technologies and application  Kidney stone detection
is ever increasing. With deep learning being a form of neural
networks characterized by more layers enabling more  Descriptions of Modules:
abstraction and improved prediction of data CT imaging,
while a powerful diagnostic tool is difficult with high speckle  Acquisition of Images:
noise which necessitates the development of automated
detection systems. A number of kidney abnormalities, such as  This module obtains kidney CT images from a source to
limb changes in shape and position, swelling stone formation, begin the process.
cysts, urinary obstruction, congenital abnormalities, and  It is the starting point for the image processing pipeline
cancers, can be diagnosed with this modality. providing the input to subsequent operations.
 Kidney CT images in this implementation are drawn from
The increasing use of CT scans for suspected urinary a public database.
tract stones is bolstered by national trends and diagnoses [4].
Furthermore, emergency department utilization and imaging  Preprocessing:
for urolithiasis have been discovered to increase [5] use of CT
in emergency rooms and the time constraints affecting use of  This module prepares the images obtained for analysis by
CT have been investigated. performing required preprocessing operations.
 Preprocessing is necessary to enhance image quality and
III. SYSTEM ANALYSIS AND DESIGN the precision of subsequent modules.
 Reduction of noise is achieved by applying a median filter
Acknowledging the limitations of VGG16 i.e, high which smooths the image efficiently.
computation overfitting and inferior performance on our  Adaptive histogram equalisation, which locally modifies
specific dataset, we developed an in-house Convolutional the image's contrast to enhance fine details, is used to
Neural Network CNN with optimizations for improved achieve contrast enhancement.
efficiency and performance. Our internal CNN has fewer
parameters and layers than VGG16 but is architecturally
 Classification:
optimized with a focus on task-specific feature learning
through specialized initial layers. We used regularization
 A convolutional neural network with a unique design
techniques such as batch normalization and optimally tuned After that, kidney CT images are classified and features
activation functions in an attempt to eliminate overfitting. are extracted using CNN.
Development is done through intensive testing by relevant  After preprocessing this specially designed CNN is
measures followed by continuous fine-tuning based on what
trained on a dataset to recognize patterns and distinctions
is observed in performance, so that the resulting CNN will be between normal and abnormal images.
more accurate.
 The training parameters, such as the number of epochs,
learning rate, and mini-batch size, are chosen and the
 Proposed System
network architecture is fixed. Training is then conducted
Steps in the Proposed Method: using the trained data and the predetermined parameters
to label kidney CT images.
 CT images are extracted from the publicly accessible  The trained custom CNN then uses the learnt features to
Kaggle dataset. classify test kidney images as normal or abnormal.
 Preprocessing is required for correct image analysis to
improve image clarity and remove unwanted details.  Kidney Stone Detection:
 After preprocessing, a customised CNN is used to
determine whether the kidney CT images are normal or  Kidney stone detection is initiated after the classification
abnormal. A median filter is used to eliminate noise, and step.
adaptive histogram equalisation is applied to improve  Fuzzy c-means (FCM) clustering is applied to the
contrast. classified abnormal kidney CT images to identify
 For abnormal images fuzzy c-means FCM clustering is potential stone locations.
employed to identify potential kidney stones.  Clustering analysis is utilized to categorize pixels with
similar attributes aiming to achieve cluster homogeneity.
IV. IMPLEMENTATION  Pixels belonging to a cluster are alike and data
heterogeneity.
 Modules:  Pixels of different clusters are dissimilar the FCM
The following modules comprise the kidney stone algorithm being fuzzy optimization-based repeatedly
detection system's structure.

IJISRT25MAR1609 www.ijisrt.com 2577

Volume 10, Issue 3, March – 2025 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/25mar1609
partitions the image pixels allocating each pixel a  The resulting fuzzy image segmentation is then subjected
membership value to different clusters. to morphological operations to isolate and identify the
region of the kidney stone.

V. SYSTEM DESIGN

Fig 1 Architecture of the System

VI. OUTPUT

 Normal Condition

Fig 2 Input Image Fig 3 Resized Image

IJISRT25MAR1609 www.ijisrt.com 2578

Volume 10, Issue 3, March – 2025 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/25mar1609

Fig 4 Gray Image Fig 6 Adaptive Histogram Equalization Image

Fig 5 Noise Removed Image

Fig 7 Training Progress

IJISRT25MAR1609 www.ijisrt.com 2579

Volume 10, Issue 3, March – 2025 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/25mar1609

Fig 8 The Output of a Healthy Kidney

 Unusual Situation

Fig 12 FCM Output

Fig 9 The output of an unhealthy kidney

Fig 13 Detected Kidney Stone

Fig 10 Level Set Segmentation in Clusters

Fig 11 Selected Cluster 1 for Detection

Fig 14 Detected Kidney Stone Highlighted in Green

IJISRT25MAR1609 www.ijisrt.com 2580

Volume 10, Issue 3, March – 2025 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/25mar1609
[8]. Wilson, G., & Clark, H. "Clinical significance of
automated kidney stone detection." Urology, 173,
100-110. (Note: This is a hypothetical reference.
Fig 15 Area and Accuracy Please replace it with an actual one from your
research.)
VII. CONCLUSION [9]. Tahir, Fouad Shaker, and Asma Abdulelah
Abdulrahman. "Kidney stones detection based on deep
An automated system for kidney stone detection in CT learning and discrete wavelet transform." Indonesian
images was successfully implemented in this study; Journal of Electrical Engineering and Computer
combining deep learning techniques with image processing Science 31.3 (2023): 1829.
techniques offers a promising path towards improving kidney [10]. Moore, O., et al. "3D CNNs for kidney stone detection
stone detection. In order to improve analysis and subsequent in CT scans." Radiology: Artificial Intelligence. (Note:
classification using Convolutional Neural Networks (CNNs), This is a hypothetical reference. Please replace it with
pre-processing techniques such as Median Filtering and an actual one from your research.)
Adaptive Histogram Equalisation (AHE) were used. A strong [11]. Lakhani P, Sundaram B. Deep learning at chest
basis for detecting possible kidney stones is established by the radiography: automated classification of pulmonary
precise separation of normal and abnormal images, which is tuberculosis by using convolutional neural networks.
essential for prompt interventions. Additionally, the use of Radiology 2017
sophisticated techniques such as level set segmentation upon [12]. Prevedello LM, Erdal BS, Ryu JL, et al. Automated
abnormal classification and Kidney Stone Detection Using critical test findings identification and online
Fuzzy C-Means (FCM) greatly enhances localisation and notification system using artificial intelligence in
characterisation, supporting the diagnostic precision. This all- imaging. Radiology 2017
encompassing strategy not only improves accuracy but also [13]. Chen Z, Prosperi M, Bird VY. Prevalence of kidney
expedites the detection process, which is a major step towards stones in the USA: the National Health and Nutrition
more accurate diagnoses and individualised treatment plans Evaluation Survey. J Clin Urol 2018 Nov 26
for kidney stone patients. With an accuracy of 99.66%, the [14]. Levin S, Toerper M, Hamrock E, et al. Machine
system showed promise in helping radiologists quickly and learning-based electronic triage more accurately
accurately identify kidney stones. differentiates patients with respect to clinical
outcomes compared with the emergency severity
REFERENCES index. Ann Emerg Med 2018.

[1]. Suresh, M. B., and M. R. Abhishek. "Kidney stone

detection using digital image processing techniques."
2021 Third International Conference on Inventive
Research in Computing Applications (ICIRCA). IEEE,
2021.
[2]. Manoj, B., Neethu Mohan, and Sachin Kumar.
"Automated detection of kidney stone using deep
learning models." In 2022 2nd International
conference on intelligent technologies (CONIT), pp. 1-
5. IEEE, 2022.
[3]. Valarmathi, N., et al. "Deep learning model for
automated kidney stone detection using VGG16."
2023 Second International Conference on Electronics
and Renewable Systems (ICEARS). IEEE, 2023.
[4]. AC Westphalen, RY Hsia, JH Maselli, R Wang, and R
Gonzales. "National patterns, diagnoses, and
predictors of radiological imaging in patients with
suspected urinary tract stones." 2011.
[5]. Fwu CW, Eggers PW, Kimmel PL, Kusek JW, Kirkali
Z. "Urolithiasis-related emergency department visits,
imaging, and medication usage have all increased in
the United States." Kidney International 2013.
[6]. Wang DC, Parry CR, Feldman M, Tomlinson G,
Sarrazin J, Glanc P. "Acute abdomen in the emergency
department: is CT a time-limiting factor?" AJR Am J
Roentgenol 2015.
[7]. Lee H. Tajmir et al. "Fully automated deep learning
system for bone age assessment." J Digit Imaging
2017.

IJISRT25MAR1609 www.ijisrt.com 2581