Radiomics and quantitative multi-parametric MRI for predicting uterine fibroid growth

Abstract

Significance: Uterine fibroids (UFs) can pose a serious health risk to women. UFs are benign tumors that vary in clinical presentation from asymptomatic to causing debilitating symptoms. UF management is limited by our inability to predict UF growth rate and future morbidity.

Aim: We aim to develop a predictive model to identify UFs with increased growth rates and possible resultant morbidity.

Approach: We retrospectively analyzed 44 expertly outlined UFs from 20 patients who underwent two multi-parametric MR imaging exams as part of a prospective study over an average of 16 months. We identified 44 initial features by extracting quantitative magnetic resonance imaging (MRI) features plus morphological and textural radiomics features from DCE, T2, and apparent diffusion coefficient sequences. Principal component analysis reduced dimensionality, with the smallest number of components explaining over 97.5% of the variance selected. Employing a leave-one-fibroid-out scheme, a linear discriminant analysis classifier utilized these components to output a growth risk score.

Results: The classifier incorporated the first three principal components and achieved an area under the receiver operating characteristic curve of 0.80 (95% confidence interval [0.69; 0.91]), effectively distinguishing UFs growing faster than the median growth rate of $0.93{\mathrm{cm}}^3/\mathrm{year}/\mathrm{fibroid}$ from slower-growing ones within the cohort. Time-to-event analysis, dividing the cohort based on the median growth risk score, yielded a hazard ratio of 0.33 [0.15; 0.76], demonstrating potential clinical utility.

Conclusion: We developed a promising predictive model utilizing quantitative MRI features and principal component analysis to identify UFs with increased growth rates. Furthermore, the model's discrimination ability supports its potential clinical utility in developing tailored patient and fibroid-specific management once validated on a larger cohort.

Keywords: artificial intelligence; machine learning; multi-parametric magnetic resonance imaging; radiomics; uterine fibroids; women’s health.

Fig. 1

Example T2 image of a uterine fibroid without and with the expert contour (in green).

Fig. 2

(a) Growth rates of each UF ($N=44$) for each patient ($N_{p }=20$) and (b) the initial estimated volume per UF. The patients were first ordered by increasing the number of fibroids and then by total fibroid burden growth, and fibroids for a given patient were ordered by the increasing growth rate. The ordering in panels (a) and (b) is the same. Note that a negative growth rate indicates regression ($N=13$).

Fig. 3

Confusion matrices for (a) the qualitative features “rim presence on ADC” and (b) “heterogeneous enhancement on DCE-MRI” in predicting fibroid growth as faster or slower than the median (“true” = “faster” and “false” = “slower” growth). Stacked histograms of the actual growth rates color coded using the radiologist qualitative features of (c) “rim presence on ADC” and (d) “heterogeneous enhancement on DCE-MRI.” In an “ideal” situation, the bars to the left of the vertical line indicating the median growth rate would be blue, and all to the right would be orange.

Fig. 4

Stacked histograms of the actual growth rates color coded using (a) the GLCM texture feature “sum average” derived from the T2 map and (b) the GLCM texture feature IMC2 derived from the ADC map. In an “ideal” situation, the bars to the left of the vertical line indicating the median growth rate would be blue, and all to the right would be orange.

Fig. 5

(a) ROC curve (“raw” in solid blue line, fitted using the proper binormal model in bold black line) and 95% CI (in dashed lines) with “raw” operating point (blue circle) corresponding to the median as threshold for the radiomics risk score decision variable, (b) confusion matrix corresponding to this operating point, (c) stacked histogram of the actual growth rates color coded using the radiomics risk score, and (d) time-to-event analysis based on dividing the cohort by median radiomics risk score ($N=22$ at time = 0 for each sub-cohort). Note that the radiomics model’s task was to predict growth “slower” or “faster” than the median not to estimate the actual growth rate.