Biparametric prostate MRI: impact of a deep learning-based software and of quantitative ADC values on the inter-reader agreement of experienced and inexperienced readers

A total of 204 mpMRI studies were evaluated for inclusion in the study. Thirty-four MRI examinations were excluded due to inadequate quality on at least one of the bpMRI sequences. The 170 patients included in the study were homogeneous between the bpMRI, Quantib, and qADC groups with regard to median age (65 vs. 68 vs. 64 years old, p = 0.10), median total PSA value (7.6 ng/ml vs. 7.9 ng/ml vs. 7.3 ng/ml, p = 0.72), and median PSA density (0.13 ng/ml/ml vs. 0.11 ng/ml/ml vs. 0.13 ng/ml/ml, p = 0.64).

At mpMRI, a total of 238 lesions were identified by the senior radiologist (median of 1 lesion per patient), of which 64 (26.9%) PI-RADS 2 lesions, 73 (30.7%) PI-RADS 3 lesions, 90 (37.8%) PI-RADS 4 lesions, and 12 (5.0%) PI-RADS 5 lesions. Overall, 128/170 patients (75.3%) had at least one lesion classified as PI-RADS 3 or greater. The three study groups had a similar distribution with regards to the PI-RADS scoring of the lesions (p = 0.99, Table 2). One hundred and two patients underwent MRI-TRUS fusion targeted biopsy, of which 35/102 (34.3%) patients had negative results, 17/102 (16.7%) patients had Gleason Score 3 + 3 PCa, and 50/102 (49.0%) patients had Gleason Score ≥ 3 + 4 PCa. There were no statistically significant differences among the study groups in the proportion of overall PCa and clinically significant PCa (p = 0.91, Table 2).

Table 2 Patient characteristics

The mean qADC value of identified lesions was 0.56 (± 0.16 SD) for the inexperienced reader and 0.58 (± 0.18 SD) for the experienced reader.

The clinical, radiologic, and pathologic characteristics of the three different patient groups are summarized in Table 2.

Inter-reader agreement between experienced and inexperienced readers

The inter-reader agreement for lesion scoring between the experienced and inexperienced readers was fair (k = 0.38, p < 0.00001) for bpMRI, moderate (k = 0.41, p < 0.00001) for Quantib and moderate (k = 0.41, p < 0.00001) for qADC.

The inter-reader agreement for lesion significance between the two readers was moderate (k = 0.40, p < 0.00001) for bpMRI, moderate (k = 0.41, p < 0.00001) for Quantib, and fair (k = 0.39, p < 0.00001) for qADC.

The inter-reader agreement for patient significance between the two readers was moderate for both bpMRI (0.42, p < 0.00001), Quantib (0.44, p < 0.00001), and qADC (0.41, p < 0.00001).

Inexperienced radiologist analysis

With respect to mpMRI, the inter-reader agreement for the inexperienced radiologist was moderate at both the per-lesion analysis (k = 0.42, p < 0.00001 for bpMRI, k = 0.45, p < 0.00001 for Quantib, k = 0.41, p < 0.00001 for qADC), the significant lesion analysis (k = 0.44, p < 0.00001 for bpMRI, k = 0.46, p < 0.00001 for Quantib, k = 0.42, p < 0.00001 for qADC) and the per-patient analysis (k = 0.43, p < 0.00001 for bpMRI, k = 0.44, p < 0.00001 for Quantib, k = 0.42, p < 0.00001 for qADC), Table 3. Detection rate of PCa was 0.29 for mpMRI, 0.24 for bpMRI, 0.26 for Quantib, 0.23 for qADC. Detection rate of csPCa was 0,20 for mpMRI, 0.16 for bpMRI, 0.17 for Quantib, 0.14 for qADC (Table 4).

Table 3 Summary of results—inter-reader agreement between mpMRI and the three studiesTable 4 Summary of results—detection rate

The average time needed to the inexperienced radiologist for the overall reporting of MRI examinations was 8,23 min (IQR: 5,32–10,13 min) for the bpMRI, 7,11 min (IQR: 4,43–9,36 min) for Quantib, 9,87 min (IQR: 5,72–12,01 min) for qADC. The difference between the three groups was statistically significant (p < 0.00001).

Experienced radiologist analysis

The inter-reader agreement for the experienced radiologist was moderate at both the per-lesion analysis (k = 0.44, p < 0.00001 for bpMRI, k = 0.46, p < 0.00001 for Quantib, k = 0.42, p < 0.00001 for qADC), the significant lesion analysis (k = 0.43, p < 0.00001 for bpMRI, k = 0.45, p < 0.00001 for Quantib, k = 0.42, p < 0.00001 for qADC) and the per-patient analysis (k = 0.45, p < 0.00001 for bpMRI, k = 0.47, p < 0.00001 for Quantib, k = 0.43, p < 0.00001 for qADC), Table 3. Detection rate of PCa was 0.29 for mpMRI, 0.26 for bpMRI, 0.27 for Quantib, 0.27 for qADC. Detection rate of csPCa was 0,20 for mpMRI, 0.18 for bpMRI, 0.19 for Quantib, 0.16 for qADC (Table 4).

The average time needed to the experienced radiologist for the overall reporting of MRI examinations was 5,62 min (IQR: 3,54–9,13 min) for the bpMRI, 5,07 min (IQR: 3,43–8,76 min) for Quantib, 6,21 min (IQR: 4,32–12,01 min) for qADC. The difference between the three groups was statistically significant (p = 0.00001).

留言 (0)

沒有登入
gif