Table 1.
List of morphological parameters (DNA content and shape descriptors) used in the study.
All units are in pixels (or pixels squared for cell area parameters). Pixels to micrometer factor is 0.624. “None” are dimensionless values.
Table 2.
Descriptive statistic (Mean ± SD) of normal and neoplastic cell nuclei with detail on Perimeter, Area and IOD values.
Values of 1000 counted neoplastic A and B cells are also reported. *Asterisk indicates mean values of morphological features significantly different (P<0.05) compared with normal haemocytes.
Fig 1.
The sequence of steps to process images of neoplastic nuclei at light disease severity (A, C, E) and heavy (B,D,F) from Feulgen-stained sections (A-B); converted to the green channel (C-D) and after binarisation (E-F).
Fig 2.
Scatter plots of 7 of the total morphological parameters extracted from normal and neoplastic nuclei.
Fig 3.
Histopathology coupled with morphometry and ploidy values.
A. Normal haemocytes and three typical cases of neoplastic condition according to disease severity: A/B cell identification in histological sections and data distribution; a: normal haemocytes—granulocyte (big arrowheads) and ialinocytes (small arrowheads); b: light disease condition with A cells visible (arrowheads and insert) exhibiting marked pleomorphism (polymetrism and polymorphism) and vesicular nuclei; c: moderate disease condition showing presence of A and B cells (insert) mixed with normal haemocytes (*); d: heavy disease state primarily exhibits the presence of B cells (insert and arrowheads) rounded in shape, bigger, with nuclei showing a dense chromatin pattern. H&E. Insert Scale bar: 25 μm; B. Histograms showing IOD and Area in at different level of disease severity. Notice the second population that arises in the heaviest disease conditions; (C) Histograms showing the differences in IOD, Area and Perimeter between normal and neoplastic nuclei; (D) Detailed small fluctuations in the IOD distribution over disease progression and compared with normal cells. The data were smoothed with a running average filter of size 7 to preserve the large scale features of the plots.
Fig 4.
(A) Binned scatterplot for area and IOD in normal and neoplastic nuclei: denser area in neoplastic graph is visible at the level of normal haemocytes values for IOD and area. (B) Scatterplot of neoplastic samples divided in disease severity.
Table 3.
The percentages of correct classification by a stepwise discriminant analysis according to ‘normal’ and ‘neoplastic’ cells.
Table 4.
The percentages of correct classification by a stepwise discriminant analysis according to disease severity.