PreciseDx: New York State Approves AI-Based Diagnostic Test For Breast Cancer

In January of this year, in a major regulatory milestone, New York State Department of Health approved a new AI-based digital pathology diagnostic test for breast cancer. The test, developed by PreciseDx, is performed on tumor samples removed during surgery and used to grade tumors and improve prediction of recurrence risk. The test is capable of accurately stratifying early-stage invasive breast cancer into low risk and high-risk likelihood of recurrence within 6 years of primary diagnosis. PreciseDx’s AI platform captures more information from every slide, stain, and tissue sample than humanly possible, and hones in on a set of 8-12 key features. This is not a black box. The PDxBr test provides readily explainable features.

PDxBr will reach patients this year

The PDxBr test was approved through New York State’s Clinical Laboratory Evaluation Program after rigorous evaluation for analytical performance, clinical performance, and reproducibility. PreciseDx can now begin commercially testing patient samples through its certified lab in New York. The test will be available to providers and patients throughout New York State later this year. Oncologists, pathologists, and researchers will be able to send digital scans to PreciseDx’s lab and receive results in 2-3 days.

PreciseDx has also introduced the AI breast cancer platform to the European oncology community through research collaborations in Portugal and the Netherlands. One collaboration is with the Laboratory of Pathology at Dordrecht, Netherlands which is responsible for the pathology of both the Albert Schweitzer Hospital in Dordrecht and the Beatrix Hospital in Gorinchem. The company has also entered into a research collaboration with the Breast Center at the Champalimaud Clinical Center in Lisbon, Portugal.

AI powered approach

Once breast cancer is detected, pathologists and oncologists must design an effective treatment plan which is currently limited by the subjectivity of a visual interpretation. Traditional diagnostic methods only look at a few components when trying to grade breast cancer and assess risk of recurrence. PreciseDx’s AI-powered approach examines many more data points and relationships between them to provide patient-specific risk information. The PDxBr test improves accuracy because it combines an AI generated prognostic grade with clinical-pathologic features to improve characterization, grading and prognosis. The new AI test provides readily explainable features for pathologists and oncologists. This enables integration of the test results into the treatment plan in a meaningful way.

Significant improvement

In a subset of the validation study, scientists demonstrated significant improvement over the performance of the Oncotype Recurrence Score, with a higher C-index, sensitivity, and hazard ratio than the Oncotype RS alone. This is good news for women with breast cancer because the combination of the two tests provides a much more accurate prediction of the risk of breast cancer recurrence. Recurrence score is an important datapoint that oncologists use to determine if a woman should have chemotherapy treatment. By using the PDxBr test in conjunction with the Oncotype tests, healthcare professionals should be able to provide more personalized treatment for breast cancer patients, potentially leading to improved outcomes. A paper on the study entitled, Development and validation of an AI-enabled digital breast cancer assay to predict early-stage breast cancer recurrence within six years was published in Breast Cancer Research on December 20, 2022.

Study Overview

1. Gerardo Fernandez, MD, and colleagues at PreciseDx developed a deep learning system for analyzing digital pathology images and available clinical data to assess risk of breast cancer recurrence.
2. They conducted a validation study in New York City.
3. 2,075 patients with infiltrating ductal carcinoma of the breast were enrolled.
4. The participants were patients at Mount Sinai Hospital and Mount Sinai Beth Israel Hospital in New York.
5. The patients were followed for a median of 6 years between 2004 and 2016.
6. A total of 15,000 digital images were reviewed.
7. Two pathologists blinded to outcome, reviewed all cases to confirm the diagnosis of invasive breast cancer and tumor/image quality.
8. One image per patient was advanced for feature extraction and model development.
9. Clinical and pathology data were extracted from the Mount Sinai electronic medical record system.
10. The PDxBr training model used the following data: patient age, patient age combined with tumor size, anatomic stage, lymph node status, 7 imaging features from digital pathology.

Study Results

1. The AI tools isolated and quantified individual elements of the invasive cancer, and extracted features representing tissue architecture and cell type characteristics (this is called the AI grade).
2. The platform identified discrete biologically driven tumor elements that are not captured using current breast cancer grading.
3. The two most important clinical variables were positive lymph nodes and age at diagnosis.
4. The two most significant image features were the degree of differentiation in tubule formation and tumor-adjacent lymphoid clusters. Fewer lymphocytes in the tumor increased the risk of recurrence.
5. Neither histology grade nor ER/PR/Her2 status was selected by the AI model.
6. Scientists demonstrated significant improvement over the performance of the Oncotype Recurrence Score, with a higher C-index, sensitivity, and hazard ratio than the Oncotype RS alone.
7. The study identified significant and incremental improvement when imaging features and clinical features are combined.
8. The study demonstrates that the PDxBr test can enrich breast cancer grading and improve risk categorization for predicting recurrence.
9. The study demonstrates that the PDxBr test could be effective in combination with Oncotype RS and other gene expression tests to rule-out chemotherapy.
10. This AI approach is quantifiable and reproducible, and provides readily explainable features for pathologists and oncologists. It is appropriate for risk modeling and potential predictive response.

PreciseDx spun out from the Mount Sinai Health System in New York in October 2020 and raised $10.75 million in Series A in January 2022. The company is led by leading AI scientists and experienced business leaders with deep expertise in diagnostics and pathology. The company’s investors include Merck Global Health Innovation Fund, Agilent Technologies, IBM Ventures, Hobart Group, and Mount Sinai Health System.

References

Development and validation of an AI-enabled digital breast cancer assay to predict early-stage breast cancer recurrence within 6 years, Breast Cancer Research, December 20, 2022

Authors: Gerardo Fernandez, Marcel Prastawa, Abishek Sainath Madduri, Richard Scott, Bahram Marami, Nina Shpalensky, Krystal Cascetta, Mary Sawyer, Monica Chan, Giovanni Koll, Alexander Shtabsky, Aaron Feliz, Thomas Hansen, Brandon Veremis, Carlos Cordon‑Cardo, Jack Zeineh, and Michael J. Donovan

Subscribe and Comment

I’m interested in your feedback – please leave your comments. To subscribe please click subscribe at the top of this article.

Copyright © 2023 Margaretta Colangelo. All Rights Reserved.

This article was written by Margaretta Colangelo. Margaretta is a leading AI analyst who tracks significant milestones in AI in healthcare. She’s consulting at AI healthcare companies and she writes about some of the companies she’s consulting with. Margaretta serves on the advisory board of the AI Precision Health Institute at the University of Hawaiʻi Cancer Center @realmargaretta

SOURCE: LinkedIn