Context.— Artificial intelligence is a transforming technology for anatomic pathology. Involvement within the workforce will foster support for algorithm development and implementation. Objective.— To develop a supportive ecosystem that enables pathologists with variable expertise in artificial intelligence to create algorithms in a development environment with seamless transition to a production environment. Design.— Platform requirements included (1) approachable and intuitive user interface, (2) diverse algorithmic modeling options, (3) support capability for internal and external collaborations, (4) a seamless mechanism for discovery to clinical deployment transition, (5) the ability to meet minimum institutional requirements for information technology (IT) review, and (6) the ability to be scalable over time. The ecosystem required platform education, data science guidance, project management structure, and ongoing leadership. Results.— The development team considered internal development and vended solutions. Because of the extended timeline and resource requirements for internal development, a decision was made to use a vended solution. Vendor proposals were solicited and reviewed by pathologists, IT, and security groups. A vendor was selected and pipelines for development and production were established. Proposals for development were solicited from the pathology department. Eighty-four investigators were selected for the initial cohort, receiving training and access to dedicated subject matter experts. A total of 30 of 31 projects progressed through the model development process of annotating, training, and validation. Based on these projects, 15 abstracts were submitted to national meetings. Conclusions.— Democratizing artificial intelligence by creating an ecosystem to support pathologists with varying levels of expertise can break down entry barriers, reduce overall cost of algorithm development, improve algorithm quality, and enhance the speed of adoption.

PURPOSE:Neoadjuvant ± adjuvant immunotherapy improves event-free survival (EFS) versus adjuvant immunotherapy alone for high-risk resectable Stage III melanoma. However, the optimal regimen balancing efficacy and tolerability is unknown. T-cell immunoglobulin and ITIM domain (TIGIT) is a promising immune checkpoint whose therapeutic potential in Stage III melanoma is underexplored. METHODS:In this phase II trial, patients with resectable, macroscopic Stage III melanoma received four 21-day neoadjuvant cycles of 1200 mg IV atezolizumab (anti-PD-L1) + 600 mg IV tiragolumab (anti-TIGIT), therapeutic lymph node dissection (TLND) and eight adjuvant cycles of atezolizumab. Primary endpoints were pathologic response and recurrence-free survival (RFS) after TLND in patients receiving adjuvant therapy; secondary and exploratory endpoints included adverse events (AEs), EFS and distant metastasis-free survival (DMFS). RESULTS:Thirty-four patients initiated neoadjuvant atezolizumab/tiragolumab and were evaluable for AEs, pathologic response and EFS. 76.5 % had > 1 metastatic lymph node at baseline and 73.5 % were Stage IIIC. Thirty patients had per protocol TLND. Major pathologic responses (MPR, ≤10 % viable tumor) were observed in 16/34 patients (47.1 %). With 19.9 months median follow-up, 12-month EFS was 72.0 % (95 %CI 57.9-89.5 %), 12-month RFS was 73.3 % (95 %CI 56.9-94.5 %), and 12-month DMFS 86.0 % (95 %CI 72.2-100 %). 12-month RFS and DMFS for patients with an MPR were both 91.7 % (95 %CI: 77.3-100 %). Grade 3 + AEs at least possibly neoadjuvant treatment-related occurred in 2 patients (5.9 %). CONCLUSION:Among patients with high-risk resectable Stage III melanoma, neoadjuvant atezolizumab/tiragolumab was a promising regimen with a favorable safety profile and warrants further study. A strategy to personalize neoadjuvant regimen selection is needed.

BACKGROUND:Clinical stage III melanoma has a high recurrence rate following surgical dissection. The use of neoadjuvant treatment has been shown to improve long-term outcomes when compared with up-front surgery, with the pathological response being the best predictor. Biomarkers predicting responses and possibly allowing surgical de-escalation are missing. We aimed at assessing the correlation between post-treatment positron emission tomography-computed tomography (PET-CT) results and the pathological and long-term outcomes in a cohort of patients treated with a variety of regimens. PATIENTS AND METHODS:With IRB approval, we collected data on all patients with stage III melanoma treated with neoadjuvant intent at our institution since 2011 and assessed the association between their baseline features, post-treatment PET-CT findings, surgical pathological findings, and long-term outcomes. RESULTS:In total, 60 patients were identified; the median age was 61 years and 50% were males. The most common regimens used were immunotherapy (IO) in 42% and IO combined with targeted therapy (TT) in 48%. The imaging and pathological response rates were correlated at 60% and 65%, respectively. Both imaging and pathological response rates were higher for those treated with IO+TT than those with IO only, yet the presence of a major pathological response was predictive of a decreased risk of recurrence only in patients treated with IO. Overall, 67% had discordance in the number of involved lymph nodes between the PET-CT and pathological findings. PET-CT underestimation of lymph node involvement was associated with recurrence in those treated with IO only. CONCLUSIONS:Post-neoadjuvant PET-CT findings should be interpreted with caution when considering the extent of the surgical dissection.

BACKGROUND AND AIMS: Effective approaches for prevention of hepatocellular carcinoma (HCC) will have a significant impact on HCC-related mortality. There are strong preclinical data and rationale to support targeting epidermal growth factor receptor (EGFR) for HCC chemoprevention. Small molecule inhibitors of EGFR have been Food and Drug Administration-approved for cancer therapy, which provides an opportunity to repurpose one of these drugs for chemoprevention of HCC. Unfortunately, the frequency of side effects associated with administration of these drugs at oncology doses renders them ineffective for chemoprevention. This clinical trial assesses whether lower doses of one of these inhibitors, erlotinib, still engages EGFR in the liver to block signaling (eg, EGFR phosphorylation). The objective of this clinical trial was determination of a safe and minimum effective dose of erlorinib for which >= 50% reduction phospho-EGFR immunohistochemical staining in the liver was observed. METHODS: Forty six participants were preregistered and 25 participants were registered in this multicenter trial. By dose de-escalation trial design, cohorts of participants received a 7-day course of erlotinib 75 mg/day, 50 mg/day or 25 mg/day with liver tissue acquisition prior to and after erlotinib. RESULTS: A >= 50% reduction phospho-EGFR immunohistochemical staining in the liver was observed in a minimum of 40% of participants (predetermined threshhold) at each of the dose levels. Erlotinib was very well tolerated with few side effects observed, particularly at the dose of 25 mg/day. Favorable modulation of the Prognostic Liver Signature was observed in participants who received erlotinib. CONCLUSION: These data support the selection of erlotinib doses as low as 25 mg/day of for a longer intervention to assess for evidence of efficacy as an HCC chemoprevention drug (ClinicalTrials.gov NCT02273362).

Both targeted therapies and immunotherapies provide benefit in resected Stage III melanoma. We hypothesized that the combination of targeted and immunotherapy given prior to therapeutic lymph node dissection (TLND) would be tolerable and drive robust pathologic responses. In NeoACTIVATE (NCT03554083), a Phase II trial, patients with clinically evident resectable Stage III melanoma received either 12 weeks of neoadjuvant vemurafenib, cobimetinib, and atezolizumab ( BRAF -mutated, Cohort A, n = 15), or cobimetinib and atezolizumab ( BRAF -wild-type, Cohort B, n = 15) followed by TLND and 24 weeks of adjuvant atezolizumab. Here, we report outcomes from the neoadjuvant portion of the trial. Based on intent to treat analysis, pathologic response (≤50% viable tumor) and major pathologic response (complete or near-complete, ≤10% viable tumor) were observed in 86.7% and 66.7% of BRAF -mutated and 53.3% and 33.3% of BRAF -wild-type patients, respectively (primary outcome); these exceeded pre-specified benchmarks of 50% and 30% for major pathologic response. Grade 3 and higher toxicities, primarily dermatologic, occurred in 63% during neoadjuvant treatment (secondary outcome). No surgical delays nor progression to regional unresectability occurred (secondary outcome). Peripheral blood CD8 + T CM cell expansion associated with favorable pathologic responses (exploratory outcome).

Antibody reagents and validation for MxIF panel design

BACKGROUND: Effective approaches for prevention of hepatocellular carcinoma (HCC) will have a significant impact on HCC-related mortality. Data from preclinical models strongly support epidermal growth factor receptor (EGFR) as a target for chemoprevention of HCC. Erlotinib is a small-molecular EGFR tyrosine kinase inhibitor that is FDA-approved for cancer treatment. The side effects of erlotinib observed in patients receiving the oncology dose (150 mg/day) render it unacceptable as a chemoprevention agent at this dose. OBJECTIVE: The objective of this clinical trial was determination of a safe and minimum effective dose of erlortinib for which a ≥50% reduction phospho-EGFR immunohistochemical staining in the liver was observed. RESULTS: 46 participants were pre-registered and 25 participants were registered in this multicenter trial. In a dose de-escalation trial design, cohorts of participants received a 7 day course of erlotinib 75 mg/day, 50 mg/day or 25 mg/day with liver tissue acquisition prior to and after erlotinib. A ≥50% reduction phospho-EGFR immunohistochemical staining in the liver was observed in a minimum of 40% of participants at each of the dose levels. Erlotinib was very well tolerated with few side effects observed, particularly at the dose of 25 mg/day. A favorable modulation of the Prognostic Liver Signature was observed in participants that received the higher erlotinib doses (50 and 75 mg/day). CONCLUSION: These data support the selection a dose of 25 mg/day of erlotinib for a longer intervention to assess for evidence of efficacy as an HCC chemoprevention drug. Citation Format: Kenneth K. Tanabe, David Zahrieh, Carrie A. Strand, Yujin Hoshida, Thomas J. Flotte, Gary Della’Zanna, Asad Umar, Paul Limburg. Pilot study of EGFR inhibition with erlotinib in liver fibrosis for hepatocellular carcinoma prevention [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3031.

Phenotypic Cell Classification Tree

Colorectal cancer (CRC) is the second most commonly diagnosed cancer in the United States. Genetic testing is critical in assisting in the early detection of CRC and selection of individualized treatment plans, which have shown to improve the survival rate of CRC patients. The tissue slide review (TSR), a tumor tissue macro-dissection procedure, is a required pre-analytical step to perform genetic testing. Due to the subjective nature of the process, major discrepancies in CRC diagnostics by pathologists are reported, and metrics for quality are often only qualitative. Progressive context encoder anomaly detection (P-CEAD) is an anomaly detection approach to detect tumor tissue from whole slide images (WSIs), since tumor tissue is by its nature, an anomaly. P-CEAD-based CRC tumor segmentation achieves a 71% 26% sensitivity, 92% 7% specificity, and 63% 23% F1 score. The proposed approach provides an automated CRC tumor segmentation pipeline with a quantitatively reproducible quality compared with the conventional manual tumor segmentation procedure.

Direct interactions between tumor and immune cells mediate the antitumor effect of all modern cancer immunotherapeutic agents. Simultaneously, tumor cells have evolved mechanisms of evasion, including the downregulation of HLA-I, potentially disrupting the mechanism of action employed by many immune checkpoint inhibitors. And yet, the in situ interplay between these cells within the tumor immune microenvironment (TIME) remains elusive. Recent advances in histologic multiplex bioimaging platforms have enabled in-depth molecular characterization of single cells within spatially preserved and clinically archived tumor tissues. Herein, we applied multiplex immunofluorescence to excisional lymph node biopsies from 14 patients with metastatic melanoma who experienced clear objective responses to immunotherapy (7 complete response; 7 progressive disease) to determine distinguishing features of the TIME in the pretreatment setting. Distinct regions of the TIME were evaluated using 35 proteins probing tumor, immune, and vasculature components across 323 fields of view. Single-cell compositional analysis confirmed established prognostic immune cell types including increased prevalence of cytotoxic T cells within the tumor core fields of view of responders. Integrating single-cell quantification with the spatial arrangement of cellular neighborhoods surrounding tumor cells revealed novel, spatial immune signatures capable of stratifying TIME based on clinical response. Our analysis revealed dynamic cellular composition of the tumor-centric cellular neighborhood (TCCN) based on anatomic subregion, functional expression of HLA-I by the index tumor cell and ultimately clinical response to immunotherapy. Overall, this study provides an analytic framework to resolve the cellular complexity of the TIME, increasingly relevant to the outcomes of modern cancer immunotherapy. Significance: Findings from this work propose a novel approach to resolving clinical heterogeneity of the TIME by objectively quantifying the cellular interactions occurring in metastatic melanoma lymph node tissue utilizing multiplex immunofluorescence. This study provides an analytic and biologically derived unit of measure, the TCCN which is customizable for studying critical paracrine interactions within spatially preserved tissue of various cancers and across the spectrum of multiplex imaging modalities.

ABSTRACTWhole slide imaging (WSI) is transforming the practice of pathology, converting a qualitative discipline into a quantitative one. However, one must exercise caution in interpreting algorithm assertions, particularly in pathology where an incorrect classification could have profound impacts on a patient, and rare classes exist that may not have been seen by the algorithm during training. A more robust approach would be to identify areas of an image for which the pathologist should concentrate their effort to make a final diagnosis. This anomaly detection strategy would be ideal for WSI, but given the extremely high resolution and large file sizes, such an approach is difficult. Here, we combine progressive generative adversarial networks with a flexible adversarial autoencoder architecture capable of learning the “normal distribution” of WSIs of normal skin tissue at extremely high resolution and demonstrate its anomaly detection performance. Our approach yielded pixel-level accuracy of 89% for identifying melanoma, suggesting that our label-free anomaly detection pipeline is a viable strategy for generating high quality annotations - without tedious manual segmentation by pathologists. The code is publicly available at https://github.com/Steven-N-Hart/P-CEAD.

Genome-wide copy number profiling by single-nucleotide polymorphism (SNP) array is increasingly employed in the clinical diagnostic workup of melanocytic tumors. We present our SNP array results on 675 melanocytic tumors, including 615 histologically ambiguous tumors evaluated by our institution's dermatopathology consultation service and a separate validation cohort of 26 known benign nevi and 34 known malignant melanomas. The total number of somatic copy number abnormalities, sub-chromosomal copy number abnormalities, regions of homozygosity, and abnormalities at disease-associated regions was significantly associated with a diagnosis of malignancy across disease categories. In our study, the number of copy number abnormalities was the factor that best discriminated between benign versus malignant diagnoses, confirming recent published research. Histologically ambiguous tumors had a range and spectrum of abnormalities, including recurrent 11p gains, copy state transitions over kinase genes, and 3p deletions overlapping BAP1 in neoplasms with Spitzoid morphology. Our data suggest that histologically ambiguous melanocytic neoplasms and early primary melanomas have a range of abnormalities that is intermediate between unambiguous benign or malignant melanocytic neoplasms. Careful technical review and an integrated diagnostic approach are essential for the accurate interpretation of SNP array results on histologically ambiguous melanocytic tumors.