Prediction of Recurrence in Patients with Stage III Colon Cancer Using Conventional Clinicopathological Factors and Peripheral Blood Test Data: A New Analysis with Artificial Intelligence

Background: Survival rate may be predicted by tumor-node-metastasis staging systems in colon cancer. In clinical practice, about 20 to 30 clinicopathological factors and blood test data have been used. Various predictive factors for recurrence have been advocated; however, the interactions are complex and remain to be established. We used artificial intelligence (AI) to examine predictive factors related to recurrence. Methods: The study group comprised 217 patients who underwent curative surgery for stage III colon cancer. Using a self-organizing map (SOM), an AI-based method, patients with only 23 clinicopathological factors, patients with 23 clinicopathological factors and 34 of preoperative blood test data (pre-data), and those with 23 clinicopathological factors and 31 of postoperative blood test data (post-data) were classified into several clusters with various rates of recurrence. Results: When only clinicopathological factors were used, the percentage of T4b disease, the percentage of N2 disease, and the number of metastatic lymph nodes were significantly higher in a cluster with a higher rate of recurrence. When clinicopathological factors and pre-data were used, three described pathological factors and the serum C-reactive protein (CRP) levels were significantly higher and the serum total protein (TP) levels, serum albumin levels, and the percentage of lymphocytes were significantly lower in a cluster with a higher rate of recurrence. When clinicopathological factors and post-data were used, three described pathological factors, serum CRP levels, and serum carcinoembryonic antigen levels were significantly higher and serum TP levels, serum albumin levels, and the percentage of lymphocytes were significantly lower in a cluster with a higher rate of recurrence. Conclusions: This AI-based analysis extracted several risk factors for recurrence from more than 50 pathological and blood test factors before and after surgery separately. This analysis may predict the risk of recurrence of a new patient by confirming which clusters this patient belongs to.

Download Full-text

Artificial intelligence-guided tissue analysis combined with immune infiltrate assessment predicts stage III colon cancer outcomes in PETACC08 study

Gut ◽

10.1136/gutjnl-2019-319292 ◽

2019 ◽

Vol 69 (4) ◽

pp. 681-690 ◽

Cited By ~ 12

Author(s):

Cynthia Reichling ◽

Julien Taieb ◽

Valentin Derangere ◽

Quentin Klopfenstein ◽

Karine Le Malicot ◽

...

Keyword(s):

Artificial Intelligence ◽

Colon Cancer ◽

Patient Outcomes ◽

Improve Patient Care ◽

Stage Iii ◽

Free Survival ◽

Relapse Risk ◽

Stage Iii Colon Cancer ◽

Healthy Mucosa ◽

Improve Patient

ObjectiveDiagnostic tests, such as Immunoscore, predict prognosis in patients with colon cancer. However, additional prognostic markers could be detected on pathological slides using artificial intelligence tools.DesignWe have developed a software to detect colon tumour, healthy mucosa, stroma and immune cells on CD3 and CD8 stained slides. The lymphocyte density and surface area were quantified automatically in the tumour core (TC) and invasive margin (IM). Using a LASSO algorithm, DGMate (DiGital tuMor pArameTErs), we detected digital parameters within the tumour cells related to patient outcomes.ResultsWithin the dataset of 1018 patients, we observed that a poorer relapse-free survival (RFS) was associated with high IM stromal area (HR 5.65; 95% CI 2.34 to 13.67; p<0.0001) and high DGMate (HR 2.72; 95% CI 1.92 to 3.85; p<0.001). Higher CD3+ TC, CD3+ IM and CD8+ TC densities were significantly associated with a longer RFS. Analysis of variance showed that CD3+ TC yielded a similar prognostic value to the classical CD3/CD8 Immunoscore (p=0.44). A combination of the IM stromal area, DGMate and CD3, designated ‘DGMuneS’, outperformed Immunoscore when used in estimating patients’ prognosis (C-index=0.601 vs 0.578, p=0.04) and was independently associated with patient outcomes following Cox multivariate analysis. A predictive nomogram based on DGMuneS and clinical variables identified a group of patients with less than 10% relapse risk and another group with a 50% relapse risk.ConclusionThese findings suggest that artificial intelligence can potentially improve patient care by assisting pathologists in better defining stage III colon cancer patients’ prognosis.

Download Full-text

Combination of tissues analysis and immune infiltrate in localized colon cancer using artificial intelligence in PETACC8 study.

Journal of Clinical Oncology ◽

10.1200/jco.2019.37.15_suppl.3574 ◽

2019 ◽

Vol 37 (15_suppl) ◽

pp. 3574-3574

Author(s):

Cynthia Reichling ◽

Julien Taieb ◽

Valentin Derangere ◽

Karine Le Malicot ◽

Jean Francois Emile ◽

...

Keyword(s):

Artificial Intelligence ◽

Colon Cancer ◽

Outcome Prediction ◽

Prediction Models ◽

Cox Regression ◽

Disease Free Survival ◽

Stage Iii ◽

Outcome Variable ◽

Stage Iii Colon Cancer ◽

Automated Method

3574 Background: We used artificial intelligence to perform tissue classification and count CD3 and CD8 in each subclass and determined their role in outcome prediction in PETACC8 cohort of stage III colon cancer treated with FOLFOX or FOLFOX plus cetuximab. Methods: We developed artificial intelligence aimed to detect tumor, healthy mucosa, stroma and immune cells on whole slide of CD3 and CD8 staining. The invasive margin (IM) was also automatically determined. Using a lasso algorithm, the software was able to detect digital parameters within the tumor core (TC) which were related to patients’ outcome (variable called DGMate for DiGital tuMor pArameTErs). CD3 and CD8 lymphocytes density were also quantified automatically by the software in TC and at IM. Associations with disease-free survival (DFS) were evaluated by multivariable Cox regression adjusting for age, T/N stage, sidedness, KRAS/BRAF, DNA mismatch repair (MMR). Results: On 1220 samples collected, data could be generated for 1018 patients. We observed that a high IM stromal area and a high DGMate were associated with a poorer DFS [HR 5.65 (95% CI, 2.34, 13.67), p < 0.0001; HR 2.72 (95% IC, 1.92, 3.85), p<0.001 respectively for the continuous variable]. A higher density of CD3+ TC, CD3+ IM and CD8+ TC were significantly associated with a longer DFS (HR 0.75 (95% IC, .66, .87), p<0.0001; HR 0.78 (95% IC, .68, .88), p<0.0001; HR 0.83 (95% IC, .71, .96), p=0.01). All these immune variables were significantly correlated with each other. ANOVA test demonstrated that CD3+ TC gave a similar prognostic value compared to the classical CD3/CD8 immunoscore (p=0.44). The combination of IM stromal area, DGMate and CD3 outperformed the classical CD3/CD8 immunoscore to estimate patients’ prognosis (C-index= 0.601 vs 0.578, p-value=0.04). Adding this new variable to classical clinical prognostic parameters we generated a nomogram which predicted the risk of relapse of stage III colon cancer with a stronger predictive value compared to clinical parameters or the immunoscore. Conclusions: We propose a new fully automated method of whole slide analysis using a software based on artificial intelligence which classify tissue and determine tumor and immune parameters on one single slide stained with CD3 antibody. This valuable strategy outperforms immunoscore and clinical outcome prediction models.

Download Full-text