Artificial Intelligence Algorithm to Screen for Diabetic Neuropathy: A Pilot Study
<b>Background/Objectives</b>: Patients with type 2 diabetes (T2D) are at risk of developing multiple complications, and diabetic polyneuropathy (DPN) is by far the most common. The purpose of the present study was to assess the ability of a new algorithm based on artificial intelligence...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-04-01
|
| Series: | Biomedicines |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2227-9059/13/5/1075 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | <b>Background/Objectives</b>: Patients with type 2 diabetes (T2D) are at risk of developing multiple complications, and diabetic polyneuropathy (DPN) is by far the most common. The purpose of the present study was to assess the ability of a new algorithm based on artificial intelligence (AI) to identify patients with T2D who are at risk of DPN in order to move on to further instrumental evaluation with the biothesiometer method. <b>Methods</b>: This is a single-centre, cross-sectional study with 201 consecutive T2D patients recruited at the Diabetes Operating Unit of the ULSS 6 of Padua (Northeast Italy). The individual risk of developing DPN was calculated using the AI-based MetaClinic Prediction Algorithm and compared with the DPN diagnosis provided by the digital biothesiometer method, which measures the vibratory perception threshold (VPT) on both feet. <b>Results</b>: Of the enrolled patients, 107 (53.23%) were classified by AI software as having a low probability of developing DPN, 39 (19.40%) as having a moderate probability, 29 (14.43%) as having a high probability, and 26 (12.94%) as having a very high probability. In 63 of the total patients, biothesiometer measurement showed a VPT ≥ 25 V, indicative of DPN, while 138 patients had a non-pathological VPT value (< 25 V) (prevalence of abnormal VPT 31.34%; prevalence of normal VPT 68.66%). The overall agreement between biothesiometer results and AI risk attribution was 65%. Cohen’s κ was 0.162, and Gwet’s AC1 coefficient 0.405. <b>Conclusions</b>: The use of an optimized AI algorithm can help estimate the risk of developing DPN, thereby guiding more targeted and in-depth screening, including instrumental assessment using the biothesiometer method. |
|---|---|
| ISSN: | 2227-9059 |