Exploring BiomedCLIP’s Capabilities in Medical Image Analysis: A Focus on Scoliosis Detection and Severity Assessment
Background/Objectives: Open-source artificial intelligence models (OSAIMs), such as BiomedCLIP, hold great potential for medical image analysis. While OSAIMs are increasingly utilized for general image interpretation, their adaptation for specialized medical tasks, such as evaluating scoliosis on po...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-01-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/15/1/398 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841549385635201024 |
|---|---|
| author | Bartosz Polis Agnieszka Zawadzka-Fabijan Robert Fabijan Róża Kosińska Emilia Nowosławska Artur Fabijan |
| author_facet | Bartosz Polis Agnieszka Zawadzka-Fabijan Robert Fabijan Róża Kosińska Emilia Nowosławska Artur Fabijan |
| author_sort | Bartosz Polis |
| collection | DOAJ |
| description | Background/Objectives: Open-source artificial intelligence models (OSAIMs), such as BiomedCLIP, hold great potential for medical image analysis. While OSAIMs are increasingly utilized for general image interpretation, their adaptation for specialized medical tasks, such as evaluating scoliosis on posturographic X-ray images, is still developing. This study aims to evaluate the effectiveness of BiomedCLIP in detecting and classifying scoliosis types (single-curve and double-curve) and in assessing scoliosis severity. Methods: The study was conducted using a dataset of 262 anonymized posturographic X-ray images from pediatric patients (ages 2–17) with diagnosed scoliosis. The images were collected between January 2021 and July 2024. Two neurosurgical experts manually analyzed the Cobb angles and scoliosis stages (mild, moderate, severe). BiomedCLIP’s performance in detecting scoliosis and its type was evaluated using metrics such as accuracy, sensitivity, specificity, and AUC (Area Under the Curve). Statistical analyses, including Pearson correlation and ROC curve analysis, were applied to assess the model’s performance. Results: BiomedCLIP demonstrated moderate sensitivity in detecting scoliosis, with stronger performance in severe cases (AUC = 0.87). However, its predictive accuracy was lower for mild and moderate stages (AUC = 0.75 and 0.74, respectively). The model struggled with correctly identifying single-curve scoliosis (sensitivity = 0.35, AUC = 0.53), while it performed better in recognizing double-curve cases (sensitivity = 0.78, AUC = 0.53). Overall, the model’s predictions correlated moderately with observed Cobb angles (r = 0.37, <i>p</i> < 0.001). Conclusions: BiomedCLIP shows promise in identifying advanced scoliosis, but its performance is limited in early-stage detection and in distinguishing between scoliosis types, particularly single-curve scoliosis. Further model refinement and broader training datasets are essential to enhance its clinical applicability in scoliosis assessment. |
| format | Article |
| id | doaj-art-df09d9ddc5af4a2a89689d3fb212628f |
| institution | Kabale University |
| issn | 2076-3417 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj-art-df09d9ddc5af4a2a89689d3fb212628f2025-01-10T13:15:25ZengMDPI AGApplied Sciences2076-34172025-01-0115139810.3390/app15010398Exploring BiomedCLIP’s Capabilities in Medical Image Analysis: A Focus on Scoliosis Detection and Severity AssessmentBartosz Polis0Agnieszka Zawadzka-Fabijan1Robert Fabijan2Róża Kosińska3Emilia Nowosławska4Artur Fabijan5Department of Neurosurgery, Polish-Mother’s Memorial Hospital Research Institute, 93-338 Lodz, PolandDepartment of Rehabilitation Medicine, Faculty of Health Sciences, Medical University of Lodz, 90-419 Lodz, PolandIndependent Researcher, Luton LU2 0GS, UKDepartment of Neurosurgery, Polish-Mother’s Memorial Hospital Research Institute, 93-338 Lodz, PolandDepartment of Neurosurgery, Polish-Mother’s Memorial Hospital Research Institute, 93-338 Lodz, PolandDepartment of Neurosurgery, Polish-Mother’s Memorial Hospital Research Institute, 93-338 Lodz, PolandBackground/Objectives: Open-source artificial intelligence models (OSAIMs), such as BiomedCLIP, hold great potential for medical image analysis. While OSAIMs are increasingly utilized for general image interpretation, their adaptation for specialized medical tasks, such as evaluating scoliosis on posturographic X-ray images, is still developing. This study aims to evaluate the effectiveness of BiomedCLIP in detecting and classifying scoliosis types (single-curve and double-curve) and in assessing scoliosis severity. Methods: The study was conducted using a dataset of 262 anonymized posturographic X-ray images from pediatric patients (ages 2–17) with diagnosed scoliosis. The images were collected between January 2021 and July 2024. Two neurosurgical experts manually analyzed the Cobb angles and scoliosis stages (mild, moderate, severe). BiomedCLIP’s performance in detecting scoliosis and its type was evaluated using metrics such as accuracy, sensitivity, specificity, and AUC (Area Under the Curve). Statistical analyses, including Pearson correlation and ROC curve analysis, were applied to assess the model’s performance. Results: BiomedCLIP demonstrated moderate sensitivity in detecting scoliosis, with stronger performance in severe cases (AUC = 0.87). However, its predictive accuracy was lower for mild and moderate stages (AUC = 0.75 and 0.74, respectively). The model struggled with correctly identifying single-curve scoliosis (sensitivity = 0.35, AUC = 0.53), while it performed better in recognizing double-curve cases (sensitivity = 0.78, AUC = 0.53). Overall, the model’s predictions correlated moderately with observed Cobb angles (r = 0.37, <i>p</i> < 0.001). Conclusions: BiomedCLIP shows promise in identifying advanced scoliosis, but its performance is limited in early-stage detection and in distinguishing between scoliosis types, particularly single-curve scoliosis. Further model refinement and broader training datasets are essential to enhance its clinical applicability in scoliosis assessment.https://www.mdpi.com/2076-3417/15/1/398BiomedCLIPscoliosis detectionmedical image analysismachine learning in healthcareartificial intelligence |
| spellingShingle | Bartosz Polis Agnieszka Zawadzka-Fabijan Robert Fabijan Róża Kosińska Emilia Nowosławska Artur Fabijan Exploring BiomedCLIP’s Capabilities in Medical Image Analysis: A Focus on Scoliosis Detection and Severity Assessment Applied Sciences BiomedCLIP scoliosis detection medical image analysis machine learning in healthcare artificial intelligence |
| title | Exploring BiomedCLIP’s Capabilities in Medical Image Analysis: A Focus on Scoliosis Detection and Severity Assessment |
| title_full | Exploring BiomedCLIP’s Capabilities in Medical Image Analysis: A Focus on Scoliosis Detection and Severity Assessment |
| title_fullStr | Exploring BiomedCLIP’s Capabilities in Medical Image Analysis: A Focus on Scoliosis Detection and Severity Assessment |
| title_full_unstemmed | Exploring BiomedCLIP’s Capabilities in Medical Image Analysis: A Focus on Scoliosis Detection and Severity Assessment |
| title_short | Exploring BiomedCLIP’s Capabilities in Medical Image Analysis: A Focus on Scoliosis Detection and Severity Assessment |
| title_sort | exploring biomedclip s capabilities in medical image analysis a focus on scoliosis detection and severity assessment |
| topic | BiomedCLIP scoliosis detection medical image analysis machine learning in healthcare artificial intelligence |
| url | https://www.mdpi.com/2076-3417/15/1/398 |
| work_keys_str_mv | AT bartoszpolis exploringbiomedclipscapabilitiesinmedicalimageanalysisafocusonscoliosisdetectionandseverityassessment AT agnieszkazawadzkafabijan exploringbiomedclipscapabilitiesinmedicalimageanalysisafocusonscoliosisdetectionandseverityassessment AT robertfabijan exploringbiomedclipscapabilitiesinmedicalimageanalysisafocusonscoliosisdetectionandseverityassessment AT rozakosinska exploringbiomedclipscapabilitiesinmedicalimageanalysisafocusonscoliosisdetectionandseverityassessment AT emilianowosławska exploringbiomedclipscapabilitiesinmedicalimageanalysisafocusonscoliosisdetectionandseverityassessment AT arturfabijan exploringbiomedclipscapabilitiesinmedicalimageanalysisafocusonscoliosisdetectionandseverityassessment |