Aims: To differentiate hepatoblastoma (HB) from hepatocellular carcinoma (HCC) by computerized image analysis. This is critical for treatment modalities and prognostic stratification but is usually difficult in small biopsy specimens.
Methods and results: Computerized image-processing technology was used to calculate the nuclear-cytoplasmic ratio (N/C), cellularity (CEL) and other cellular and nuclear parameters in HB (n = 18) and paediatric HCC (pHCC, n = 11). The proliferation index (PI) and apoptotic index (AI) were also measured. Fetal type HB (FHB) compared with pHCC had more uniform nuclei (P < or = 0.014), lower PI (P = 0.028) and AI (P = 0.009), whereas the embryonal type HB (EHB) had a higher N/C (P < 0.001), higher CEL (P = 0.043), smaller cells (P = 0.043) and higher PI (P = 0.020) than pHCC. Moreover, EHB had a higher N/C (P < 0.001), higher CEL (P = 0.021), smaller cells (P = 0.021), more nuclear pleomorphism (P < or = 0.036) and higher PI (P < 0.001) than FHB. Multivariate analysis showed that FHB, EHB and pHCC could be classified accurately by a regression model. This logistic model further correctly stratified four additional test cases from biopsy specimens.
Conclusions: These results indicate that computerized morphometric analysis can yield useful criteria to distinguish HB from pHCC in small biopsy specimens, and, compared with FHB, the poorer prognosis of EHB may result from its more undifferentiated (immature) and proliferative phenotype.