Abstract

AI has been integrated in every sector. AI in health sector plays major role to boost the work efficiency of doctor’s, radiologists, and generalist. AI has been gaming changing method for automating disease detection. This survey shows how using Deep Learning can help the Radiologists to ease their work. This survey uses convolutional neural networks (CNNs) to detect brain tumour and pneumonia. We present a partially implemented prototype system named “X-Ray Vision\' that combines image classification models for brain tumour detection and pneumonia. This survey also underlines current challenges, potential improvements, and deployment through a web-based interface.

Introduction

Automation in medical image analysis is becoming essential because manual image examination by radiologists is slow and prone to errors. Deep learning—especially Convolutional Neural Networks (CNNs)—now plays a major role in detecting diseases from MRI and X-ray scans. These AI models learn visual patterns such as edges, textures, tissues, and tumor boundaries directly from raw pixel data, allowing fast and accurate diagnosis.

The proposed system, “X-Ray Vision,” is a partially developed prototype that automates the detection of brain tumors and pneumonia. The system not only analyzes medical images but also stores patient records in a web-based database, allowing easy retrieval even if files are lost. Models are trained using large, labeled datasets containing both normal and diseased scans, resulting in radiologist-level accuracy when tested repeatedly.

CNNs are effective because their layered architecture learns simple shapes in early layers and complex medical structures in deeper layers. This hierarchical learning allows consistent and precise diagnoses. By digitizing reports and storing them in a database, the system reduces radiologists’ workload and ensures secure record maintenance.

Future improvements may integrate additional disease models, further enhancing workflow automation in healthcare.

RELATED WORK & SHORTCOMINGS

Several studies have advanced AI-based disease detection:

1. Kermany et al. (2018)
   Developed an early CNN model for pneumonia detection using chest X-rays with high accuracy.
   Limitation: Required heavy computation, not suitable for real-time deployment.

2. Masoud Nickparvar (2021)
   Released a Brain Tumor MRI Dataset and experimental CNN implementations on Kaggle.
   Limitation: Not validated using hospital data; models not integrated into a unified diagnostic system.

3. Pereira et al. (2016)
   Used deep CNNs for brain tumor segmentation with improved boundary detection.
   Limitation: Required extensive manual annotations, long training time, and suffered from overfitting with smaller datasets.

Overall, past research was powerful but often required high computing resources, lacked real-time use, or did not include a deployable system.

LITERATURE SURVEY SUMMARY

Over the last decade, CNN-based deep learning has transformed medical imaging. Key contributions include:

• Kermany et al. (2018): Demonstrated clear potential of CNNs for pneumonia detection but needed strong hardware.

• Rajpurkar et al. (2017) CheXNet: Achieved radiologist-level pneumonia detection accuracy using a DenseNet model.
  Drawback: Extremely deep network, hard to interpret, computationally heavy.

• Nickparvar (2021): Provided widely used brain tumor MRI dataset and CNN models. Limited real-world validation.

• Pereira et al. (2016): Improved brain tumor segmentation accuracy. Faced challenges like overfitting and annotation requirements.

• Abdelhafiz et al. (2019): Used transfer learning with 3D CNNs for tumor classification. Needed high memory and lacked deployment features.

Common limitations across studies:

• Focus on only one disease

• High computational requirement

• No complete end-to-end system

• Lack of clinical deployment or patient data management

• Poor interpretability

The proposed X-Ray Vision system addresses these gaps by combining multiple disease detection models with a practical web interface and integrated database.

PROPOSED SYSTEM SUMMARY

The X-Ray Vision platform provides an end-to-end solution for disease detection and record management. It includes:

Key Features

• Upload medical images (brain MRI or chest X-ray)

• Automatic disease detection using CNN models

• Web-based dashboard for radiologists

• Storage and retrieval of patient records

• Fast, user-friendly interface for real-time prediction

Two AI Models

1. Brain Tumor Detection using MRI

2. Pneumonia Detection using X-ray

After uploading a scan, preprocessing is performed (normalizing, resizing, noise removal), then the suitable model is automatically selected. The final prediction (normal/diseased) and probability score are displayed, and results are saved in a database.

System Architecture

• User uploads scan → preprocessing → CNN model → prediction → results saved & displayed

• Database allows viewing, searching, and comparing patient history

• Ensures digital storage, reduces manual documentation, and supports clinical workflow

CNN ARCHITECTURE SUMMARY

Both models use CNNs with the following structure:

• Input Layer: Resizes images (e.g., 128×128 pixels) and converts them into numerical pixel values.

• Convolution Layers: Extract patterns like edges, shapes, textures.

• ReLU Activation: Introduces nonlinearity to learn complex relationships.

• Pooling Layers: Reduce image size, retain important features, and speed up computation.

• Flatten Layer: Converts feature maps into a vector.

• Dense Layers: Learn final classification boundaries.

• Output Layer: Sigmoid-based binary classification (Normal/Diseased).

Training uses Kaggle datasets for pneumonia and brain tumor detection with data augmentation to prevent overfitting.

RESULT SUMMARY

• A working web dashboard is created for radiologists.

• Pneumonia detection testing demonstrates correct classification and probability scores.

• The integrated system proves reliable, fast, and practical for clinical use.

Conclusion

Overall, this survey demonstrates how well deep learning can identify brain tumours and pneumonia from medical images. Our partial implementation shows that lightweight CNN architectures can still be used for web deployment and still achieve dependable accuracy. To improve interpretability and reliability, future research will concentrate on clinical validation, explainable AI integration, and dataset augmentation.

References

[1] Y. Lecun, Y. Bengio, and G. Hinton, Deep Learning, Nature, vol. 521, pp. 436–444, 2015. [2] K. Simonyan and A. Zisserman, Very Deep Convolutional Networks for Large-Scale Image Recognition, arXiv preprint arXiv:1409.1556, 2014. [3] K. He, X. Zhang, S. Ren, and J. Sun, Deep Residual Learning for Image Recognition, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770–778, 2016. [4] D. Kermany, M. Goldbaum, W. Cai et al., Identifying Medical Diagnoses and Treatable Diseases by Image-Based Deep Learning, Cell, vol. 172, no. 5, pp. 1122–1131, 2018. [5] P. Rajpurkar, J. Irvin, K. Zhu et al., CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning, arXiv preprint arXiv:1711.05225, 2017. [6] S. Pereira, A. Pinto, V. Alves, and C. A. Silva, Brain Tumor Segmentation Using Convolutional Neural Networks in MRI Images, IEEE Transactions on Medical Imaging, vol. 35, no. 5, pp. 1240–1251, 2016. [7] M. Abdelhafiz, J. Yang, H. Ammar, and K. Nabavi, Deep Convolutional Neural Networks for Brain Tumor Classification, IEEE Access, vol. 7, pp. 55258–55269, 2019. [8] R. R. Selvaraju, M. Cogswell, A. Das et al., Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization, Proceedings of the IEEE International Conference on Computer Vision (ICCV), pp. 618–626, 2017. [9] M. Nickparvar, Brain Tumor MRI Dataset, Kaggle Dataset, 2021. [Online]. Available: https://www.kaggle.com/datasets/masoudnickparvar/brain-tumor-mri-dataset [10] S. Jaiswal, A. Sharma, and P. Kumar, A Comparative Study on Deep Learning Approaches for Disease Detection Using Medical Imaging, International Journal for Research in Applied Science and Engineering Technology (IJRASET), vol. 11, no. 3, pp. 1500–1506, 2023.

Copyright

Copyright © 2025 Harikrishna Ambat, Juber Shaikh, Omkar Avhad, Shreya Navale. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.