AI based Acoustic System for Non Invasive Bone Fracture Diagnosis
Bone fracture is a discontinuity in the bone caused by applying force on it above its threshold level. The conventional methods used for diagnosing fractures are costly and difficult to access. Furthermore, methods like X-ray and Tuning Fork require a qualified medical professional to understand and
2025-06-28 16:25:02 - Adil Khan
AI based Acoustic System for Non Invasive Bone Fracture Diagnosis
Project Area of Specialization Biomedical EngineeringProject SummaryBone fracture is a discontinuity in the bone caused by applying force on it above its threshold level. The conventional methods used for diagnosing fractures are costly and difficult to access. Furthermore, methods like X-ray and Tuning Fork require a qualified medical professional to understand and interpret the results. All these factors work as a barrier in the diagnostic process.
Identification of faults through AI techniques is a growing trend these days in all fields. In this project, we will develop an Acoustic system for fracture detection in a bone by using AI. To begin, we'll use a transmitter that will send an acoustic signal inside the body. The signal will pass through the bone before being picked up by a receiver. The captured signal will be analyzed for features. We will use Machine Learning Techniques on the extracted features, to determine whether the bone is broken or intact.
The old methods of X-rays, Ultrasound diagnosis & the use of acoustic signals through Tuning forks are practical tools for fracture detection in bone. The methodology that we will use here will propose an automated and improved solution.
Project Objectives- To implement the concept of Machine learning in medical diagnostics applications.
- To devise a low-cost, accurate, time-saving alternate & accessible system for detecting bone fracture.
Our project will progress in two phases. In Phase 1, we will train our model on acoustic data collected from bones, and in Phase 2, we will test our model.
For the training of our model, we will collect fractured and intact bones from a butcher shop. We will cover these bones with gelatin. We will divide all these gelatin-covered bones into two sets of training and testing.
Then we will start assembling our hardware for the transmission of signals through the bone and receiving them. We will transmit the signal by using a Raspberry Pi/Microcontroller. The Microcontroller will be connected to a Piezoelectric transducer to produce an acoustic signal. After generating the acoustic signal, we will induce it into the bones with gelatin covering, and it will start traveling through the bone. At some distance to the transducer, we will place a receiver. The receiver will be the diaphragm of a stethoscope connected to a microphone to record the signal and send it to the computer as audio files for fractured and intact bones.
Upon receiving the files, we will provide labels for each file. Once we provide the files with the labels, we will apply different AI Algorithms to train our model. Finally, we will test our model on other bones and display our result in the form of a report on an LCD.
Benefits of the ProjectTechnical Benefits:
The system will detect the fracture in the bone with a high level of accuracy. It will generate the results instantly, thus saving time.
Social Benefits:
The design will provide a low-cost, safe, easily accessible, and time-saving option to people who want to diagnose a fracture in the bone. The design is made such that anyone can use it by themselves without any need for a medical practitioner.
Technical Details of Final Deliverable- Gelatin is used as it will give an impression of flesh on bones.
- The transmitter will generate a 128 Hz wave as this frequency is often used in medical applications. This wave will travel into the bone as vibrations.
- The stethoscope will convert vibrations into sound. A microphone module will convert the sound into an electrical signal. An electrical signal will be saved on a device and feature extraction will be applied to it.
- Labeling of the files of fractured bones signals and intact bones signals will be done.
- Then the files will be processed using Machine Learning algorithms and Deep Learning techniques.
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 31100 | |||
| Raspberry Pi 3b / Microcontroller + Kit | Equipment | 1 | 11000 | 11000 |
| Flexible Piezo MEAS | Equipment | 2 | 1200 | 2400 |
| Stethoscope | Equipment | 1 | 2250 | 2250 |
| Boost Converter | Equipment | 2 | 300 | 600 |
| LCD (16x2) | Equipment | 2 | 400 | 800 |
| Battery | Equipment | 2 | 500 | 1000 |
| Circuit charger | Miscellaneous | 2 | 500 | 1000 |
| Piezo Vibrate Module + Plate | Equipment | 1 | 400 | 400 |
| Vibration Sensor (SW-4) | Equipment | 1 | 150 | 150 |
| Pressure Sensor(FSR) | Equipment | 1 | 1000 | 1000 |
| Mic Module | Equipment | 2 | 200 | 400 |
| Meat & Bones | Equipment | 1 | 1200 | 1200 |
| Structure & Case | Miscellaneous | 1 | 2000 | 2000 |
| Dual Supply Circuit | Equipment | 1 | 400 | 400 |
| Power Amplifier Circuit | Equipment | 1 | 500 | 500 |
| PCB board | Equipment | 2 | 200 | 400 |
| Wires | Miscellaneous | 9 | 100 | 900 |
| Transport Charges | Miscellaneous | 20 | 150 | 3000 |
| Integrated Circuits | Miscellaneous | 5 | 70 | 350 |
| Resistor & Capacitor Chart | Miscellaneous | 1 | 350 | 350 |
| Damage/ limited accuracy | Miscellaneous | 1 | 1000 | 1000 |