Google Research and DeepMind have launched MedPaLM which is a large language model in the medical domain.
It brings together HealthSearchQA, a new free-response dataset of medical questions sought online, and six existing open-question answering datasets about professional medical exams, research, and consumer queries.
MedPaLM addresses multiple-choice questions and questions posed by medical professionals and non-professionals through the delivery of various datasets. These datasets come from MedQA, MedMCQA, PubMedQA, LiveQA, MedicationQA, and MMLU. A new dataset of curated, frequently searched medical inquiries called HealthSearchQA was also added to improve MultiMedQA.
The HealthsearchQA dataset consists of 3375 frequently asked consumer questions. It was collected by using seed medical diagnoses and their related symptoms. This model was developed on PaLM, a 540 billion parameter LLM, and its instruction-tuned variation Flan-PaLM to evaluate LLMs using MultiMedQA.
Med-PaLM needs to be better than a human medical expert’s judgment. So far, a group of healthcare professionals has stated that 92.6 percent of the Med-PaLM responses were identical to clinician-generated answers (92.9 percent).
For Flan-PaLM, 61.9 percent were in line with doctor assessments. Also, only 5.8 percent of Med-PaLM answers led to negative consequences, compared to 6.5 percent of clinician-generated answers and 29.7 percent of Flan-PaLM answers.
“iCAD and Google Health are united in our purpose-driven missions to elevate innovation, improve patient care, and optimize outcomes for all. By combining the power of our technologies and teams, we strengthen our fight against breast cancer and positively impact the lives of women and their loved ones across the globe,” said Stacey Stevens, president, and CEO of iCAD, at the time.