ModPath Chat Modern Pathology

Flow cytometric analysis of blood & bone marrow for diagnosis of acute myelogenous leukemia (AML) relies heavily on manual intervention in the processing & analysis steps. Attention-based multi-instance learning models (ABMILMs) are deep learning models that make accurate predictions & generate interpretable insights regarding the classification of a sample from individual events.
The Drs. Olga Pozdnyakova and Joshua Lewis discuss their newly developed computational pipeline using ABMILMs for the automated diagnosis of AML cases based exclusively on flow cytometric data. The study is the first to illustrate the feasibility of using deep learning-based analysis of flow cytometric data for automated AML diagnosis & molecular characterization.

Modern Pathology is publishing a seminal series of review articles highlighting the recently completed fifth edition of the World Health Organization classification of hematolymphoid tumors (WHO-HEM5). In this episode of ModPath CHAT, Dr. Sanam Loghavi from The MD Anderson Cancer Center in Houston, discusses the major updates in the classification of myeloid neoplasms, providing a comparison with WHO-HEM4R, and offering guidance on how the new classification can be applied to the diagnosis of myeloid neoplasms in routine practice.

Polymorphous adenocarcinoma (PAC) is a common, usually low-grade salivary gland carcinoma. While conventional PACs are most associated with PRKD1 p.E710D hotspot mutations, the cribriform subtype is often associated with gene fusions in PRKD1, PRKD2, or PRKD3. The guest, Dr. Justin Bishop from UT Southwestern Medical Center in Dallas, discusses his team’s recent study characterizing the fusions associated with PAC with NGS. A diverse group of fusion partners, including 13 novel partners, were identified. The most common partners for the PRKD genes were ARID1A and ARID1B.

Endoscopic evidence of disease is a critical predictor of relapse in patients with Crohn’s disease (CD). While histologic disease activity is evolving as a similarly important end point, classical morphologic features of CD may overlap with postoperative inflammatory changes, confounding the evaluation of anastomotic biopsies.
In this episode, Dr, John Hart, Professor and Vice Chair of Anatomic Pathology at the University of Chicago discusses his team’s critical recent study in Modern Pathology showing that due to the above extensive morphologic overlap and the lack of specific histologic features of relapse, biopsies from anastomotic sites are of no value in predicting clinical CD progression. On the other hand, CD activity in biopsies obtained away from anastomotic sites should be used for guiding endoscopic sampling and clinical management.

This episode features Dr. Laura Lamps , a leader in the field of gastrointestinal pathology. Dr. Lamps is the Godfrey D. Stobbe Professor and Director of Gastrointestinal Pathology and Assistant Chair for Faculty Development and Program Director of GI Pathology Fellowship at the Department of Pathology, Michigan Medicine, at the University of Michigan. The former President of US and Canadian Academy of Pathology (USCAP) shares with the audience her perspectives as a brilliant leader on the importance of seeking a diverse mentorship and investing in leadership development resources. The informative discussion centers on how to build and empower the next generation of pathologists.

In this episode, Dr. Tamara Lotan from Johns Hopkins University discusses the potential of deep-learning (DL) algorithms trained on H and E-stained whole slide images (WSI) to screen for clinically relevant genomic alterations in prostate cancer (PCA).
Dr. Lotan reviews her team’s recent publication in Modern Pathology, where they were able to create DL algorithms to identify PCA with underlying ERG fusions or PTEN deletions. By applying the algorithms to multiple radical prostatectomy and needle biopsy cohorts, the authors demonstrated the ability of DL models to accurately predict ERG/PTEN status from H and E stained WSI.