Related Experiment Video
Updated: Jun 26, 2026
13:01
Industrialized, Artificial Intelligence-guided Laser Microdissection for Microscaled Proteomic Analysis of the Tumor Microenvironment
Published on: June 3, 2022
Hidden patterns, different outcomes: un-supervised machine-learning clustering in oligometastatic gynecologic cancer
Savino Cilla1, Francesco Deodato2, Donato Pezzulla3
1Responsible Research Hospital, Medical Physics Unit, Campobasso, Italy.
Summary
Machine learning identified three distinct groups of oligometastatic gynecologic lesions treated with stereotactic body radiation therapy. These clusters showed varying local control and distant-metastases-free survival rates, aiding in outcome stratification.
Area of Science:
- Oncology
- Radiation Oncology
- Medical Physics
- Machine Learning in Medicine
Background:
- Oligometastatic gynecologic cancer presents a unique challenge in treatment stratification.
- Stereotactic body radiation therapy (SBRT) is increasingly used for oligometastatic disease.
- Predicting treatment response in this heterogeneous group remains complex.
Purpose of the Study:
- To apply an unsupervised machine learning clustering method to stratify SBRT treatment outcomes in patients with oligometastatic gynecologic cancer.
- To identify distinct lesion clusters based on clinical and dosimetric variables.
- To evaluate the association of these clusters with local control, distant-metastases-free survival, and overall survival.
Main Methods:
- A multi-centric study included 172 patients with oligometastatic uterine tumors treated with curative-intent SBRT.
- Data collected included patient age, lesion characteristics (number, type, burden, site), and SBRT dose/volume parameters.
- K-means unsupervised clustering identified lesion groups, which were then compared for treatment outcomes.
Main Results:
- Three distinct lesion clusters were identified, primarily differentiated by planning target volume, biologically effective dose, and lesion type.
- Significant differences in local control (p = .002) and distant-metastases-free survival (p = .04) were observed between the clusters at 2 years.
- Local control rates at 2 years were 84.6%, 74.7%, and 47.5% for the high-, medium-, and low-control clusters, respectively.
Conclusions:
- Unsupervised machine learning effectively partitioned oligometastatic lesions into three clusters with differential treatment responses to SBRT.
- The identified clusters demonstrate potential for stratifying patients and predicting outcomes in oligometastatic gynecologic cancer.
- Prospective validation is recommended to confirm the predictive utility of this machine learning approach.