Summary of Outperforming Human Pathologists – New Harvard-Developed AI Tool Predicts Colon Cancer Survival, Treatment Response:
Researchers at Harvard Medical School and National Cheng Kung University in Taiwan have developed an artificial intelligence model called Multi-omics Multi-cohort Assessment (MOMA) that can predict the aggressiveness of a colorectal tumor, estimate the likelihood of survival with and without disease recurrence, and recommend optimal therapy for patients with colorectal cancer based on images of tumor samples. The model was trained on information obtained from nearly 2,000 colorectal cancer patients from diverse national patient cohorts that include more than 450,000 participants. The researchers cautioned that a patient’s prognosis depends on multiple factors; the new model could guide clinicians to follow up more closely, consider more aggressive treatments, or recommend clinical trials testing experimental therapies.
*****
Artificial Intelligence Model Can Help Clinicians Predict Colon Cancer Prognosis and Treatment
Colon cancer, the third most common cancer globally, has a dismal prognosis, with the disease causing the second highest rates of cancer-related deaths in the United States. However, a new artificial intelligence (AI) model has been developed by researchers from National Cheng Kung University in Taiwan and Harvard Medical School that could help clinicians accurately predict the aggressiveness of a colorectal tumor, the likelihood of survival with and without disease recurrence, and the optimal therapy for the patient.
Currently free and available to researchers and clinicians, the AI model can do what human pathologists cannot, based on image viewing alone, and perform a wealth of additional analysis. By detecting and interpreting patterns on microscopy images of tumor samples, which are invisible to the human eye, doctors can have a better chance of making informed decisions on the progression of the disease, thus aiding clinical decision-making in resource-constrained regions.
Furthermore, the MOMA (for Multi-omics Multi-cohort Assessment) model goes beyond many AI tools, primarily performing tasks replicating or optimizing human expertise. Instead, this tool identified indicators of how aggressive a tumor was and how likely it was to behave in response to a specific treatment. Additionally, based on the image alone, the model pinpointed characteristics associated with specific genetic mutations that typically require genomic sequencing of the tumor. Having a tool that answers such questions could ultimately save lives, as it can help clinicians and patients navigate this complex illness better.
Helping Clinicians Augment Their Expertise
Although the tool is meant to boost the sagacity of physicians, the research team points out that any individual patient’s prognosis depends on multiple factors, and no model can perfectly predict a patient’s survival. However, the latest model could help clinicians follow up more closely, consider spicier treatments, or recommend clinical trials testing experimental therapies if their patients have worse predicted prognoses based on the tool’s assessment.
The researchers say that the MOMA model will undergo periodic upgrading as science evolves and new data emerge. They also note that before deploying the model for use in hospitals or clinics, it should undergo randomized, prospective studies that assess its performance in patients over time after a diagnosis. This study would allow the true potential of the algorithm to shine by comparing its real-life performance with that of human clinicians who have access to knowledge and test results that the model does not.
Conclusions
The tool’s ability to identify patterns within tumors that suggest an increased risk of cancer return or disease progression will be beneficial in resource-limited areas where advanced pathology and tumor genetic sequencing are not readily available. The AI tool could help doctors make informed decisions about treatment and prognosis for patients with colorectal cancer worldwide, sparing some of the 1 million lives this disease claims yearly.
