Description/Abstract/Artist Statement
The diabetic foot exam system aims to perform certain aspects of the dermatological and musculoskeletal assessments that are typical to a 3-minute diabetic foot exam. Utilizing the RaspberryPi computer and camera module, the system seeks to capture a series of images of the patient’s foot. It then evaluates these images for calluses, blisters, and three types of deformities: claw toe deformities, hammertoe deformities, and bunions. This evaluation is performed using a trained TensorFlow image classification model, which categorizes the image as a callus, blister, or deformity. The system was tested using six different images: four callus images, a hammertoe deformity image, and a claw toe deformity image. One of the callus images and the hammertoe image were correctly classified. The remaining images were incorrectly classified with high confidence levels, suggesting that there is overfitting in the model. These results emphasize the need for a larger, more diverse dataset for training and validation, as well as additional image processing techniques such as background subtraction, to improve system functionality.
Faculty Advisor/Mentor
Ayman Elmesalami, Soad Ibrahim
College Affiliation
College of Sciences
Presentation Type
Oral Presentation
Disciplines
Other Computer Sciences
Session Title
Colleges of Sciences UG Research #3
Location
Zoom
Start Date
3-19-2022 3:30 PM
End Date
3-19-2022 4:30 PM
Upload File
wf_yes
Included in
Diabetic Foot Exam System
Zoom
The diabetic foot exam system aims to perform certain aspects of the dermatological and musculoskeletal assessments that are typical to a 3-minute diabetic foot exam. Utilizing the RaspberryPi computer and camera module, the system seeks to capture a series of images of the patient’s foot. It then evaluates these images for calluses, blisters, and three types of deformities: claw toe deformities, hammertoe deformities, and bunions. This evaluation is performed using a trained TensorFlow image classification model, which categorizes the image as a callus, blister, or deformity. The system was tested using six different images: four callus images, a hammertoe deformity image, and a claw toe deformity image. One of the callus images and the hammertoe image were correctly classified. The remaining images were incorrectly classified with high confidence levels, suggesting that there is overfitting in the model. These results emphasize the need for a larger, more diverse dataset for training and validation, as well as additional image processing techniques such as background subtraction, to improve system functionality.