Radiology report generation using scanned images

Project

1. Bapuji Education Association (Regd) Bapuji Institute of Engineering and Technology, Davangere -57700 Department of Information Science and Engineering Presentation On Radiology Report Generation For Scanned Images Through Transfer Learning. GUIDE PROJECT CO-ORDINATORS PRESENTED BY Mrs. Ranjana B Jadekar . M.Tech Dr. Anitha G. Ph.D Mahima U K 4BD19IS044 Ms. Rekha B H . M.Tech Shrihari P S 4BD19IS094 Spoorti S Isloor 4BD19IS102 Varshini Bisaralli 4BD19IS114

2. CONTENTS Abstract Introduction Existing System Problem Statement Objectives Hardware And Software Requirements Methodology Use Case Diagram Results Adavantages Conclusion

3. ABSTRACT The automated generation of radiology reports has potential to know the medical state of the patient. This study demonstrates how transfer learning from deep learning models can be used to generate the report of scanned images from three most commonly used medical imaging modes x-ray, ultrasound, and CT scan. Identifying a suitable convolutional neural network (CNN) model through initial comparative study of several popular CNN models.

4. Introduction Radiology is the medical discipline that uses medical imaging to diagnose diseases and guide their treatment. When doctors need to get a better look at what's going on in their patient's bodies, they will often refer them to receive some type of diagnostic imaging. There are several different types of diagnostic imaging tests. Each creates images based on different technology. Some of the most common types of diagnostic imaging tests are X-rays and Computed Tomography scans (CT), and Ultrasound. , where the resulting images or pictures will help in making an accurate diagnosis , and choosing the best treatment. There has been a recent interest and developments in the automatic generation of radiology reports by analyzing scanned images. Such a technology would reduce the workload of radiologist, streamline the clinical process and speed up the necessary medical intervention preventing serious illnesses or undesired outcomes.

5. Existing System Every image requires a radiologist to carefully examine and write a full-text report to describe the finding’s. A more glaring issue is the amount of time it takes the radiologist to write a full-text report. So this would prove very time-consuming when considering the number of cases a radiologist should investigate per day. Many reports conclude with indecisive findings that require the patient to take further tests. A false positive diagnosis could result in an inappropriate course of treatment. The hospitals in rural area mainly face the problem of the unavailability of a specialist.

6. Problem Statement Medical imaging techniques are widely used in hospitals worldwide. The detailed information generated from medical images is necessary for diagnosing illnesses or tracking patient’s progress. Diagnosing medical images requires an appropriate amount of experience and consumes time of the radiologist to write a full-text report, so this would prove very time-consuming in crowded hospitals, regions, and cities, this would be problematic. These reasons combined provided good motives for us to develop image processing models capable of automating report generation.

7. Proposed System In order to reduce the workload of radiologist, streamline the clinical process and speedup the necessary intervention preventing serious illness and undesired outcomes. The development of such a system will make a great impact for instant diagnosis and medication by generating the automatic radiology report. Through this project a patient can get a rough idea about his health in one stop with various detection techniques such as X-Ray , CT Scan and Ultrasound thus decreasing costs for common man and by reducing the number of tests required and thus reducing time constraint .

8. Objectives Provide automatic report to doctors as well as patients. Time constraint is reduced due to automatic report generation. Provides access to reports by scanning radiology images. One stop destination for attaining Xray , CT Scan and Ultrasound report generation.

9. Requirements Functional Requirements Graphical User interface with the User. Software Requirements For developing the application the following are the Software Requirements: Python Django MySql MySql Client WampServer 2.4

10. Operating Systems supported Windows Compatible Technologies and Languages used to Develop Python Hardware Requirements For developing the application the following are the Hardware Requirements: Processor: Pentium IV or higher RAM: 256 MB Space on Hard Disk: minimum 512MB

11. Data Collection Data Pre-processing Transfer Learning Deployment Evaluation Text Generation Methodology

12. Use Case Diagram 12 Report Generation

13. 13

14. Results 14

15. 15

16. Advantages Has potential to reduce the heavy workload of both radiologist and referring physicians while including more details on patients decision making process. More efficient communication with referring physicians using optimized radiology reports. Refined patient communication through radiology report. Overcoming the problem of unavailability of specialists in rural area. This study provides timely model selection guidelines to the practitioenrs who often are resorted to utilise certain mode of imaging due to time and resource scarcity.

17. Conclusion R adiology report generation is important in healthcare. With the use of advanced technique such as image processing, it is possible to generate accurate and reliable radiology report from scanned images. It will reduce the workload of radiologists it can help improve patient outcomes by providing faster and accurate diagnoses, enabling earlier treatment and intervention 24

18. Thank You