Prediction of tumor response to neoadjuvant chemotherapy in high-grade osteosarcoma using clustering-based analysis of magnetic resonance imaging: an exploratory study

To evaluate the ability of magnetic resonance imaging (MRI)-based clustering analysis to predict the pathological response to neoadjuvant chemotherapy (NACT) in patients with primary high-grade osteosarcoma.

Twenty-two patients were included in this retrospective study. All patients underwent MRIs before and after NACT. The entire tumor volume was manually delineated on post-contrast T1-weighted images and subsegmented into three clusters using the K-means algorithm. Histogram-based parameters were calculated for each lesion. The response to NACT was obtained from the histopathological assessment of the tumor necrosis rate following resection. The Mann–Whitney test was used to compare poor and fair-to-good responders. The receiver operating characteristic curve was used to evaluate the diagnostic performance of the optimal parameters.

At baseline, poor responders showed a significantly larger volume of cluster1 (Vol1) than fair-to-good responders (p = 0.038). After NACT, they exhibited a lower 10th percentile (P10) and kurtosis (p = 0.038 and 0.002, respectively). Vol1 at baseline and P10 after NACT had an AUC of 77% (95% CI 56–98%). The kurtosis after NACT had the best discriminative power, with an AUC of 89.7% (95% CI 75–100%).

The MRI-based histogram and clustering analysis provided a good ability to differentiate between poor and fair-to-good responders before and after NACT. Further investigations using larger datasets are required to corroborate our findings.