Computer-assisted measurement of primary tumor area is prognostic of recurrence-free survival in stage IB melanoma patients

Abstract

Current staging guidelines are insufficient to predict which patients with thin primary melanoma are at high risk of recurrence. Computer-assisted image analysis may allow for more practical and objective histopathological analysis of primary tumors than traditional light microscopy. We studied a prospective cohort of stage IB melanoma patients treated at NYU Langone Medical Center from 2002–2014. Primary tumor width, manual area, digital area, and conformation were evaluated in a patient subset via computer-assisted image analysis. The associations between histologic variables and survival were evaluated using Cox proportional hazards model. Logistic regressions were used to build a classifier with clinicopathological characteristics to predict recurrence status. Of the 655 patients with stage IB melanoma studied, a subset of 149 patient tumors (63 recurred, 86 did not recur) underwent computer-assisted histopathological analysis. Increasing tumor width (HR 1.17, p=0.01) and digital area (HR 1.08, p<0.01) were significantly associated with worse recurrence-free survival, while non-contiguous conformation (HR 0.57, p=0.05) was significantly associated with better recurrence-free survival. The novel histopathological classifier composed of digital area, conformation, and baseline variables effectively distinguished recurrent cases from nonrecurrent cases (AUC 0.733, 95% CI 0.647–0.818), compared to the baseline classifier alone (AUC 0.635, 95% CI 0.545–0.724). Primary tumor cross-sectional area, width, and conformation measured via computer-assisted analysis may help identify high-risk patients with stage IB melanoma.

Artificial intelligence in digital pathology — new tools for diagnosis and precision oncology

“…In digital pathology, AI approaches have been applied to a variety of image processing and classification tasks, including low-level tasks, focused around object recognition problems such as detection 27,[32][33][34][35][36][37] and segmentation [38][39][40][41] as well as higher-level tasks such as predicting disease diagnosis and prognosis of treatment response on the basis of patterns in the image [42][43][44][45][46][47][48][49][50][51][52][53] . Independently of the final application, AI approaches are built to initially extract appropriate image representations, which can then be used to train a machine classifier for a particular segmentation, diagnostic or prognostic task.…”

Optimal surveillance strategies for patients with stage 1 cutaneous melanoma post primary tumour excision: three systematic reviews and an economic model

“…A total of 11 articles reporting 11 different risk prediction models met the full inclusion criteria of this review. [105][106][107][108][109][110][111][112][113][114][115] A total of 101 studies were reviewed fully and excluded for the following reasons: used single prognostic factors for model development (21%), combined stages of the disease (28%) or were not validated (51%). Details of the excluded studies are presented in Appendix 4.…”

“…A summary of the studies and patient characteristics is presented in Table 7. Eight of the studies were conducted in the USA, [106][107][108][109]111,[113][114][115] one in the UK, 112 one in Australia 105 and one in Italy. 110 Seven studies used a retrospective cohort design, [105][106][107][108][109]112,115 three used a prospective design 111,113,114 and one used a retrospective cohort of prospectively collected data.…”

“…Eight of the studies were conducted in the USA, [106][107][108][109]111,[113][114][115] one in the UK, 112 one in Australia 105 and one in Italy. 110 Seven studies used a retrospective cohort design, [105][106][107][108][109]112,115 three used a prospective design 111,113,114 and one used a retrospective cohort of prospectively collected data. 110 Patient data for development and validation of these models were taken from cancer registries, AJCC melanoma databases, clinical data from patients diagnosed and treated for melanoma or a combination of one of these sources.…”

“…The studies concerned the development of a risk or prognostic score, 107,115 a nomogram, 110 the Melanoma Severity Index model, 105 a novel histopathological classifier, 111 an electronic prediction tool, 113 a classification tree, 114 prognostic trees, 108,109 a model focused on adding a new predictor to an established model 112 and a model used to validate AJCC staging. 106 Outcome definitions and follow-up times varied across studies.…”