TL;DR:
- Johns Hopkins Medicine uses machine learning to calculate tumor cell death in osteosarcoma patients.
- Their model achieved a 99% correlation with pathologists’ assessments.
- Traditional methods are labor-intensive and prone to variability.
- AI model reduces the workload for pathologists and expedites prognosis.
- Future studies will broaden the scope and improve accuracy.
Main AI News:
In a groundbreaking development, Johns Hopkins Medicine’s research team has harnessed the power of machine learning to gauge chemotherapy success in osteosarcoma patients. Osteosarcoma, a formidable bone cancer variant, has long presented challenges in accurately assessing treatment outcomes. However, the team’s machine learning model has achieved remarkable accuracy, with a 99% correlation rate when compared to the evaluations of musculoskeletal pathologists.
Elevating Patient Prognosis with Precision
The critical metric at the heart of this innovation is the calculation of percent necrosis (PN), which quantifies the extent of tumor cell death post-chemotherapy. A higher PN indicates a more successful treatment, translating to improved survival prospects for patients. Traditionally, pathologists manually analyze whole-slide images (WSIs) of bone tissue specimens, a labor-intensive process prone to interobserver variability.
Dr. Christa LiBrizzi, a resident with Johns Hopkins Medicine’s Department of Orthopaedic Surgery and co-first author of the study, notes the challenges: “Calculating the PN is a labor-intensive process that requires a lot of annotation data from the musculoskeletal pathologist. Additionally, it has low interobserver reliability, meaning that two pathologists trying to calculate a PN from the same WSIs will often report different conclusions.“
A Paradigm Shift in Machine Learning
To address these issues, the research team embarked on a mission to develop a “weakly supervised” machine learning model that would reduce the reliance on extensive annotation data. This innovation would significantly alleviate the pathologist’s workload.
The team began by collecting WSIs and relevant data from Johns Hopkins’ U.S. tertiary cancer center, focusing on patients with intramedullary osteosarcoma. These patients had undergone both chemotherapy and surgery between 2011 and 2021. A musculoskeletal pathologist partially annotated the WSIs, identifying active tumor, dead tumor, and non-tumor tissue while estimating the PN for each patient.
Zhenzhen Wang, co-first author and a doctoral student in biomedical engineering at the Johns Hopkins University School of Medicine, explains their approach: “We decided to train the model by teaching it to recognize image patterns. We segregated the WSIs into thousands of small patches, then divided the patches into groups based on how they were labeled by the pathologist. Finally, we fed these grouped patches into the model to train it. We thought this would give the model a more robust frame of reference than simply feeding it one large WSI and risking missing the forest for the trees.”
A Resounding Success
Once the model completed its training, it was put to the test alongside the musculoskeletal pathologist. Their interpretations of six WSIs from two osteosarcoma patients yielded an impressive 99% correlation rate. Notably, the model demonstrated occasional difficulty in labeling cartilage, resulting in an outlier due to an overabundance of cartilage on one WSI. However, when this outlier was eliminated, the correlation surged to 99%.
Dr. LiBrizzi envisions a transformative impact: “If this model were to be validated and produced, it could help expedite the evaluation of chemotherapy’s effectiveness on a patient — and thus, get them a prognosis estimate sooner. That would reduce health care costs, as well as labor burdens on musculoskeletal pathologists.“
In forthcoming studies, the team aims to enhance the model by including cartilage tissue in its training and broadening the scope of WSIs to encompass various osteosarcoma types beyond intramedullary. This breakthrough underscores the potential of AI in revolutionizing cancer treatment assessment, offering hope and efficiency to patients and healthcare providers alike.
Conclusion:
The adoption of AI-enhanced chemotherapy assessment in osteosarcoma treatment at Johns Hopkins Medicine signifies a promising breakthrough. It streamlines the evaluation process, reducing labor burdens on pathologists and potentially lowering healthcare costs. This innovation has the potential to reshape the market by improving the accuracy and efficiency of cancer treatment assessments, ultimately benefiting both patients and healthcare providers.
