JMIR Publications

ABSTRACT

Background:

Machine learning (ML) has been gradually applied to cervical cancer research, but rarely combines both clinical parameters and image data. Meanwhile, more robust and accurate preoperative assessment of parametrial invasion and lymph node metastasis, as well as postoperative prognosis prediction are also in urgent need.

Objective:

We aimed to develop an integrated ML model that integrates clinicopathological parameters as well as MR images and includes both pre- and post-operation evaluation in cervical cancer (CC) patients.

Methods:

Data of CC patients from 2014 to 2022 in two tertiary hospitals were retrospectively collected and an exempt was granted by the Ethics Committee for this purpose. Variables were analyzed for their predictive value of parametrial invasion, lymph node metastasis, survival and recurrence using 7 ML models. The predictive performance of all 7 ML models was compared and an AI-assisted contouring and prognosis prediction system is developed based on optimal machine learning algorithms.

Results:

This study included 250 women for analysis (11 deaths, 24 recurrences): (1) In terms of evaluation of both parametrial invasions and lymph node metastasis, integrated ML models with weighted KNN outperformed other ML models, especially in the case of sensitivity. (2) An integrated model using weighted KNN achieved optimal performance in predicting recurrence and survival times for postoperative CC patients, showing high accuracy and balanced sensitivity. (3) an AI-assisted contouring and prognosis prediction system was developed that assists in lesion identification, preoperative evaluation and postoperative prognosis prediction.

Conclusions:

The integration of clinical data and MR image through ML models offers superior preoperative diagnostic and postoperative prognostic prediction capabilities, potentially reducing clinical errors and enabling tailored, precise treatment strategies.