Utility of new image-derived biomarkers for autosomal dominant polycystic kidney disease prognosis using automated instance cyst segmentation

Abstract

New image-derived biomarkers for patients affected by autosomal dominant polycystic kidney disease are needed to improve current clinical management. The measurement of total kidney volume (TKV) provides critical information for clinicians to drive care decisions. However, patients with similar TKV may present with very different phenotypes, often requiring subjective decisions based on other factors (e.g., appearance of healthy kidney parenchyma, a few cysts contributing significantly to overall TKV, etc.). In this study, we describe a new technique to individually segment cysts and quantify biometric parameters including cyst volume, cyst number, parenchyma volume, and cyst parenchyma surface area. Using data from the Consortium for Radiologic Imaging Studies of Polycystic Kidney Disease (CRISP) study the utility of these new parameters was explored, both quantitatively as well as visually. Total cyst number and cyst parenchyma surface area showed superior prediction of the slope of estimated glomerular filtration rate decline, kidney failure and chronic kidney disease stages 3A, 3B, and 4, compared to TKV. In addition, presentations such as a few large cysts contributing significantly to overall kidney volume were shown to be much better stratified in terms of outcome predictions. Thus, these new image biomarkers, which can be obtained automatically, will have great utility in future studies and clinical care for patients affected by autosomal dominant polycystic kidney disease.

Keywords: disease prognosis; glomerular filtration rate; imaging biomarkers; instance segmentation; outcome prediction; polycystic kidney disease.

Figure 1.

Flowchart of the study participants – from baseline MR images (top) – to follow-up outcomes (bottom).

Figure 2.

Representative MR images of excluded cases. a) Case showing low cyst-contrast every other slice likely due to poorly prescribed fat saturation pulse. b) Slice interleaving is often used to increase the signal-to-noise ratio; however, patient motion can cause misalignments between slices. c) Image acquired with a 9mm slice thickness which limits the differentiation of adjacent cysts in the axial and sagittal planes as well as limits measurement of small cysts.

Figure 3.

Bland-Altman plots between (a) the deep learning-based (planimetry) and the ellipsoid methods, and (b) the deep learning-based (planimetry) and stereology methods. The y-axis shows the percentage difference between the methods.

Figure 4.

(a) Abdominal MR image of a 41yo female patient with a PKD1 truncating mutation and a baseline eGFR of 50.7 ml/min/1.73 m². (b) The patient’s TKVp was 1912 ml. (c) The automated cyst segmentation shows each cyst labeled with a different color (TCV = 905 ml; TCN = 595 cysts). (d) The renal parenchyma segmentation resulted from the subtraction of kidney and cyst segmentations (RPV = 1007 ml). The cyst-parenchyma surface representation is shown in (e), where the surfaces between renal parenchyma and cysts are depicted with the red outline (CPSA = 3196 cm²).

Figure 5.

Linear regression analysis between the 8-year slope of eGFR and the individual image biomarkers. (a) ht-TKVp, (b) ht-TCV, (c) TCN, and (d) CPSA. To illustrate the results six pairs of patients with similar ht-TKV but different slope of eGFR were randomly selected. The colored diamonds and dots correspond to the 6 cases shown above and below the plots.

Figure 6.

ROC analysis to determine the individual biomarker predictive power of progression to KF after a 20-year follow-up period. The image biomarkers with highest predictive power were TCN and CPSA (lines pink and gray).