Predicting Ewing Sarcoma Treatment Outcome Using Infrared Spectroscopy and Machine Learning.

Area of Science:

• Biomedical Engineering
• Spectroscopy
• Oncology

Background:

• Ewing sarcoma is the second most common paediatric malignant bone tumor.
• Accurate outcome prediction is crucial for effective, risk-adjusted care in pediatric oncology.
• Fourier transform infrared (FTIR) spectroscopy analyzes biochemical content in biological samples for disease diagnosis.

Purpose of the Study:

• To evaluate the prognostic value of FTIR spectroscopy for predicting treatment outcomes in pediatric patients with bone Ewing sarcoma.
• To assess the effectiveness of various data-reduction and machine learning techniques in analyzing FTIR spectra for outcome prediction.

Main Methods:

• Retrospective analysis of FTIR spectra from bone biopsy tissues of 27 newly diagnosed pediatric Ewing sarcoma patients (ages 5-20).
• Application of data-reduction and machine learning models (Random Forest, Linear Discriminant Analysis, Support Vector Machine) to FTIR spectral data.
• Analysis of spectral changes before and after neoadjuvant chemotherapy.

Main Results:

• Supervised learning models (Random Forest, Linear Discriminant Analysis) achieved 92% accuracy in predicting patient mortality.
• A linear Support Vector Machine model, trained on spectral changes post-chemotherapy, predicted patient relapse with 92% accuracy.

Conclusions:

• FTIR spectroscopy of tumor biopsy samples demonstrates high accuracy in predicting treatment outcomes for pediatric Ewing sarcoma.
• This spectroscopic method offers a promising tool for improving prognostic accuracy and guiding patient care in pediatric bone cancer.