Association between Nuclear Morphometry Parameters and Gleason Grade in Patients with Prostatic Cancer

Objective: Gleason scoring system remains the pathological method of choice for prostate cancer (Pca) grading. However, this method of tumor tissue architectural structure grading is still affected by subjective assessment and might succumb to several disadvantages, mainly inter-observer variability. These limitations might be diminished by determining characteristic cellular heterogeneity parameters which might improve Gleason scoring homogeneity. One of the quantitative tools of tumor assessment is the morphometric characterization of tumor cell nuclei. We aimed to test the relationship between various morphometric measures and the Gleason score assigned to different prostate cancer samples. Materials and Methods: We reviewed 60 prostate biopsy samples performed at a tertiary uro-oncology center. Each slide was assigned a Gleason grade according to the International Society of Urological Pathology contemporary grading system by a single experienced uro-pathologist. Samples were assigned into groups from grades 3 to 5. Next, the samples were digitally scanned (×400 magnification) and sampled on a computer using Image-Pro-Plus software©. Manual segmentation of approximately 100 selected tumor cells per sample was performed, and a computerized measurement of 54 predetermined morphometric properties of each cell nuclei was recorded. These characteristics were used to compare the pathological group grades assigned to each specimen. Results: Initially, of the 54 morphometric parameters evaluated, 38 were predictive of Gleason grade (p < 0.05). On multivariate analysis, 7 independent parameters were found to be discriminative of different Pca grades: minimum radius shape, intensity—minimal gray level, intensity—maximal gray level, character—gray level (green), character—gray level (blue), chromatin color, fractal dimension, and chromatin texture. A formula to predict the presence of Gleason grade 3 vs. grades 4 or 5 was developed (97.2% sensitivity, 100% specificity). Discussion: The suggested morphometry method based on seven selected parameters is highly sensitive and specific in predicting Gleason score ≥ 4. Since discriminating Gleason score 3 from ≥4 is essential for proper treatment selection, this method might be beneficial in addition to standard pathological tissue analysis in reducing variability among pathologists.


Introduction
Prostate cancer (Pca) is the second most common cancer worldwide, and the second cause of death from cancer in men [1,2]. Pca is highly subjected to over-treatment in case of localized, low grade (Gleason 6) disease. Localized Pca treatment options encompass a spectrum of treatments, including active surveillance, radical prostatectomy and radiation. Therefore, initial accurate tumor grading and staging are crucial for proper treatment selection.
Tissue diagnosis is commonly performed by a pathologist using microscopic examination of slides stained with haematoxylin and eosin. When Pca is diagnosed, the tumor Table 1 displays the univariant analysis of the various morphometric parameters analyzed. Step 3: morphometric characteristics measurements-With the aid of Image-Pro plus software, pre-determined characteristics were examined, including morphometric characteristics of tumor cell nuclei size (nucleus area, perimeter, diameter, and radius), shape (angle, axes, and roundness), and texture (chromatin density and heterogeneity).

Statistical Analysis
Morphometric analysis results are summarized as mean ± standard deviation. Morphometric variables of the tumor nuclear structure were compared between the three levels of Gleason using the one-way ANOVA test. In case of a significant difference of Gleason score grades, a Bonferroni test was used to correct for multiple variables. Next, variables showing significant differences were incorporated in a morphometric analysis test of gradual progress model (Wald stepwise forward method), ultimately selecting independent morphometric variables with a statistically significant association to Gleason score. We also used the discriminative analysis to calculate a regression coefficient in order to create a formula predictive of Gleason score. Additionally, we used the receiveroperating characteristic (ROC) curve to select the optimal point of sensitivity and specificity to predict the Gleason score for each case separately. p values ≤ 0.05 were considered to be statistically significant. All statistical analyses were processed using the SPSS 26.0 © (SPSS Inc., Chicago, IL, USA) software.

General Characteristics of the Study Population
We examined 60 pathological samples of prostate adenocarcinoma, including 24 Gleason 1 to 3, 20 Gleason 4, and 16 Gleason 5. We used each tumor biopsy slide to obtain 4 to 10 microscopic images (depending on tumor area), of which we obtained a mean average of 5.4 ± 0.9 images for Gleason 1-3 slides, 4.4 ± 0.5, and 4.7 ± 0.8 for Gleason 4 and Gleason 5, respectively. Finally, 131.8 ± 28.5, 105.9 ± 15.2, and 104.7 ± 11.9 nuclei were detected, respectively. Table 1 displays the univariant analysis of the various morphometric parameters analyzed. Step 3: morphometric characteristics measurements-With the aid of Image-Pro plus software, pre-determined characteristics were examined, including morphometric characteristics of tumor cell nuclei size (nucleus area, perimeter, diameter, and radius), shape (angle, axes, and roundness), and texture (chromatin density and heterogeneity).

Statistical Analysis
Morphometric analysis results are summarized as mean ± standard deviation. Morphometric variables of the tumor nuclear structure were compared between the three levels of Gleason using the one-way ANOVA test. In case of a significant difference of Gleason score grades, a Bonferroni test was used to correct for multiple variables. Next, variables showing significant differences were incorporated in a morphometric analysis test of gradual progress model (Wald stepwise forward method), ultimately selecting independent morphometric variables with a statistically significant association to Gleason score. We also used the discriminative analysis to calculate a regression coefficient in order to create a formula predictive of Gleason score. Additionally, we used the receiveroperating characteristic (ROC) curve to select the optimal point of sensitivity and specificity to predict the Gleason score for each case separately. p values ≤ 0.05 were considered to be statistically significant. All statistical analyses were processed using the SPSS 26.0 © (SPSS Inc., Chicago, IL, USA) software.

General Characteristics of the Study Population
We examined 60 pathological samples of prostate adenocarcinoma, including 24 Gleason 1 to 3, 20 Gleason 4, and 16 Gleason 5. We used each tumor biopsy slide to obtain 4 to 10 microscopic images (depending on tumor area), of which we obtained a mean average of 5.4 ± 0.9 images for Gleason 1-3 slides, 4.4 ± 0.5, and 4.7 ± 0.8 for Gleason 4 and Gleason 5, respectively. Finally, 131.8 ± 28.5, 105.9 ± 15.2, and 104.7 ± 11.9 nuclei were detected, respectively. Table 1 displays the univariant analysis of the various morphometric parameters analyzed. Step 3: morphometric characteristics measurements-With the aid of Image-Pro plus software, pre-determined characteristics were examined, including morphometric characteristics of tumor cell nuclei size (nucleus area, perimeter, diameter, and radius), shape (angle, axes, and roundness), and texture (chromatin density and heterogeneity).

Statistical Analysis
Morphometric analysis results are summarized as mean ± standard deviation. Morphometric variables of the tumor nuclear structure were compared between the three levels of Gleason using the one-way ANOVA test. In case of a significant difference of Gleason score grades, a Bonferroni test was used to correct for multiple variables. Next, variables showing significant differences were incorporated in a morphometric analysis test of gradual progress model (Wald stepwise forward method), ultimately selecting independent morphometric variables with a statistically significant association to Gleason score. We also used the discriminative analysis to calculate a regression coefficient in order to create a formula predictive of Gleason score. Additionally, we used the receiveroperating characteristic (ROC) curve to select the optimal point of sensitivity and specificity to predict the Gleason score for each case separately. p values ≤ 0.05 were considered to be statistically significant. All statistical analyses were processed using the SPSS 26.0 © (SPSS Inc., Chicago, IL, USA) software.

General Characteristics of the Study Population
We examined 60 pathological samples of prostate adenocarcinoma, including 24 Gleason 1 to 3, 20 Gleason 4, and 16 Gleason 5. We used each tumor biopsy slide to obtain 4 to 10 microscopic images (depending on tumor area), of which we obtained a mean average of 5.4 ± 0.9 images for Gleason 1-3 slides, 4.4 ± 0.5, and 4.7 ± 0.8 for Gleason 4 and Gleason 5, respectively. Finally, 131.8 ± 28.5, 105.9 ± 15.2, and 104.7 ± 11.9 nuclei were detected, respectively. Table 1 displays the univariant analysis of the various morphometric parameters analyzed. Following 54 morphometric properties examined for univariant analysis, 38 were found to be statistically significant in predicting the Gleason score. Not all the variables were capable of distinguishing between all three levels (1-3, 4, and 5) of the Gleason score.

Indicators of Nuclear Size
We observed statistically significant separation capacity nucleus sizes (MNA-mean nuclear area) and circumference characteristics between the lowest Gleason score 1 to 3 and Gleason 5. However, these parameters were not distinguishable compared to Gleason 4. Still, the mean nuclear area demonstrated a gradual increase when comparing Gleason 1-3 to Gleason 4 and 5. Namely, 54.01 ± 7.48, 61.16 ± 17.069, and 71.14 ± 148 µm, respectively. An additional nuclear size characteristic, the minimal radius of the nucleus had a similar trend: 3.36 µm ± 0.29, 3.63 µm ± 0.5, and 3.72 µm ± 0.518 for Gleason 1-3, 4, and 5, respectively. This gradual trend of increase in size was also observed in other nucleus measurements (Table 1).

Variables Characterizing the Optical Density (Gray Levels)
Some optical density parameters were able to differentiate between Gleason 4 and Gleason 3 and 5, for instance, the "gray level of the green channel" was dark in the Gleason 1-3 nuclei (green-gray level = 97.079 ± 22.5), lighter in the Gleason  Following 54 morphometric properties examined for univariant analysis, 38 were found to be statistically significant in predicting the Gleason score. Not all the variables were capable of distinguishing between all three levels (1-3, 4, and 5) of the Gleason score.

Indicators of Nuclear Size
We observed statistically significant separation capacity nucleus sizes (MNA-mean nuclear area) and circumference characteristics between the lowest Gleason score 1 to 3 and Gleason 5. However, these parameters were not distinguishable compared to Gleason 4. Still, the mean nuclear area demonstrated a gradual increase when comparing Gleason 1-3 to Gleason 4 and 5. Namely, 54.01 ± 7.48, 61.16 ± 17.069, and 71.14 ± 148 µm, respectively. An additional nuclear size characteristic, the minimal radius of the nucleus had a similar trend: 3.36 µm ± 0.29, 3.63 µm ± 0.5, and 3.72 µm ± 0.518 for Gleason 1-3, 4, and 5, respectively. This gradual trend of increase in size was also observed in other nucleus measurements (Table 1).

Variables Characterizing the Optical Density (Gray Levels)
Some optical density parameters were able to differentiate between Gleason 4 and Gleason 3 and 5, for instance, the "gray level of the green channel" was dark in the Gleason 1-3 nuclei (green-gray level = 97.079 ± 22.5), lighter in the Gleason  Following 54 morphometric properties examined for univariant analysis, 38 were found to be statistically significant in predicting the Gleason score. Not all the variables were capable of distinguishing between all three levels (1-3, 4, and 5) of the Gleason score.

Indicators of Nuclear Size
We observed statistically significant separation capacity nucleus sizes (MNA-mean nuclear area) and circumference characteristics between the lowest Gleason score 1 to 3 and Gleason 5. However, these parameters were not distinguishable compared to Gleason 4. Still, the mean nuclear area demonstrated a gradual increase when comparing Gleason 1-3 to Gleason 4 and 5. Namely, 54.01 ± 7.48, 61.16 ± 17.069, and 71.14 ± 148 µm, respectively. An additional nuclear size characteristic, the minimal radius of the nucleus had a similar trend: 3.36 µm ± 0.29, 3.63 µm ± 0.5, and 3.72 µm ± 0.518 for Gleason 1-3, 4, and 5, respectively. This gradual trend of increase in size was also observed in other nucleus measurements (Table 1).

Variables Characterizing the Optical Density (Gray Levels)
Some optical density parameters were able to differentiate between Gleason 4 and Gleason 3 and 5, for instance, the "gray level of the green channel" was dark in the Gleason 1-3 nuclei (green-gray level = 97.079 ± 22.5), lighter in the Gleason

Discussion
Pathological diagnosis of adenocarcinoma of the prostate is made by a pathologist using microscopic examination of tissue slides based on the glandular arrangement and basic properties of the cells in the tissue. Gleason Score remains the most reliable method to predict Pca prognosis [3][4][5][12][13][14]. However, reproducibility and agreement among pathologists remain a concern, mainly due to tissue artifacts and personal experience [5,7], subsequently affecting accurate Pca grading and choosing the proper treatment. In recent years, several methods have been reported to overcome the subjectivity of the conventional histological grading system by quantitative measurement of pathological features of cancer cells, potentially improving objectivity, accuracy, and efficiency [15][16][17]. Additional value is in samples containing a small volume of Pca within the tissue, which might not represent the true pathological grade. This challenge urged the attempts to seek morphometric characteristics (size, complexity, intensity, and texture painting chromatin) of the nuclei as predictors of pathological Gleason score.

The Relationship between the Increase in the Nucleus Size and Tumor Progression
Our method was able to produce a score capable to differentiate between pathological insignificant Pca (Gleason score 1-3) and higher grades that require treatment (Gleason 5) using the MNA and nuclear surface characteristics. However, these parameters do not differentiate between Gleason 1-3 and intermediate-risk Gleason 4, as well as between Gleason 4 and 5. Still, a trend of gradual increase in the mean nuclear area of each grade between Gleason scores 3 and 5 was observed. The minimum radius of the nucleus showed a similar trend (Table 1). This gradual increase was observed in additional measures of nucleus size, supporting findings reported by Bektas et al. in 2009, who examined 130 cases of Pca to study the relationship between Gleason score and nuclear morphometrics in 30 prostatectomy samples (only two with Gleason 5), and 100 prostate biopsies. A correlation between Gleason score and MNA in both prostatectomy and needle biopsies samples was reported [18].

Discussion
Pathological diagnosis of adenocarcinoma of the prostate is made by a pathologist using microscopic examination of tissue slides based on the glandular arrangement and basic properties of the cells in the tissue. Gleason Score remains the most reliable method to predict Pca prognosis [3][4][5][12][13][14]. However, reproducibility and agreement among pathologists remain a concern, mainly due to tissue artifacts and personal experience [5,7], subsequently affecting accurate Pca grading and choosing the proper treatment. In recent years, several methods have been reported to overcome the subjectivity of the conventional histological grading system by quantitative measurement of pathological features of cancer cells, potentially improving objectivity, accuracy, and efficiency [15][16][17]. Additional value is in samples containing a small volume of Pca within the tissue, which might not represent the true pathological grade. This challenge urged the attempts to seek morphometric characteristics (size, complexity, intensity, and texture painting chromatin) of the nuclei as predictors of pathological Gleason score.

The Relationship between the Increase in the Nucleus Size and Tumor Progression
Our method was able to produce a score capable to differentiate between pathological insignificant Pca (Gleason score 1-3) and higher grades that require treatment (Gleason 5) using the MNA and nuclear surface characteristics. However, these parameters do not differentiate between Gleason 1-3 and intermediate-risk Gleason 4, as well as between Gleason 4 and 5. Still, a trend of gradual increase in the mean nuclear area of each grade between Gleason scores 3 and 5 was observed. The minimum radius of the nucleus showed a similar trend (Table 1). This gradual increase was observed in additional measures of nucleus size, supporting findings reported by Bektas et al. in 2009, who examined 130 cases of Pca to study the relationship between Gleason score and nuclear morphometrics in 30 prostatectomy samples (only two with Gleason 5), and 100 prostate biopsies. A correlation between Gleason score and MNA in both prostatectomy and needle biopsies samples was reported [18].

The Relationship between Chromatin Density and Overall Quantity and Gleason Score
When examining optical density (gray level), we observed that chromatin density varies depending on the Gleason score. This optical density change is characterized as the "gray level of the green channel", showing dark Gleason 1-3 nuclei, lighter Gleason 4 nuclei, and again darker Gleason 5 nuclei ( Table 1). The gradual increase in nucleus size with higher Gleason score and simultaneously optical density brightening in Gleason 4 might be explained by a known phenomenon of tumors; namely, the increase in nucleus area and the appearance of vesicles (appearing as bright chromatin areas), which is related to the malignant behavior of the nuclei. As a more aggressive malignancy develops, darker nuclei are seen since the number of chromosomes and nuclear protein levels increase. Similarly, a correlation between optical density, aggressive malignancy, and lymph node metastasis has been reported [19]. While analyzing the relationship between the overall amount of cell chromatin and the Gleason score, a minimal difference in chromatin quantity between Gleason 3 and 4 was observed, but prostate tumor cells obtained from Gleason 5 areas contained a larger chromatin quantity, suggesting higher genomic instability ( Table 1). As previously reported, while genomic instability is mainly related to aggressive prostate cancer, it could be used in early stages as well [20]. This relatively simple morphometric analysis may be further used to differentiate different Gleason score prostate cancer areas within the tissue.

Discriminant Score
Combining independent morphometric variables can predict the Pca tumor Gleason score with high precision (97.2% sensitivity and 100% specificity) (Figure 1). Objective intracellular parameters measurement may assist in determining Pca Gleason score particularly in tissue samples containing a small amount of Pca. However, morphometric analysis is currently time-consuming when compared to histological examination by an experienced pathologist. The use of the suggested discriminant score may not completely replace pathological analysis but may increase accuracy and reduce inter-observer discrepancies. Future use of specific cell membrane stains enabling the isolation of each cell from its neighbors and an automated morphometric tissue analysis might provide a powerful tool in assisting pathologists.

Conclusions
Size, complexity, nuclei staining intensity, and chromatin texture of prostate cancer cells can be used to distinguish between insignificant Pca (Gleason score 1 to 3) and significant Pca (Gleason 4 and 5). Improving accuracy and reducing inter-observer variability of Gleason scoring can be achieved by computerized mathematics of selected intra-cellular characteristics, allowing better standardization of the Gleason grading system. Additional validation in a larger cohort of patients and tissue samples is required.  Informed Consent Statement: Patient consent was waived due to local IRB approval for retrospective pathological data collection.