Smartphone image analysis for real time adequacy assessment during kidney biopsy

点击翻译按钮获取中文摘要

SUMMARY. The goal of this project is to develop image analysis tools that can be deployed as a smartphone application to aid in the real time assessment of biopsy adequacy during percutaneous renal biopsy (PRB) procedures. PRB remains an essential tool for the diagnosis and treatment of patients with medical kidney disease and involves obtaining a small needle core of tissue from the kidney with the use of a biopsy gun. A biopsy is deemed “adequate” if sufficient renal cortical tissue is obtained to meet criteria for rendering a histopathologic diagnosis. If insufficient tissue is obtained, or if the tissue does not contain renal cortex with enough glomeruli, then the core is deemed “inadequate”. Over the last 15 years, the rate of kidney biopsies that are inadequate for complete diagnosis due to insufficient tissue has risen to 15%, representing a significant drain on the health care system as well as a direct risk to patients who must be re-biopsied or go without an accurate diagnosis. Ideally, adequacy can be assessed at the time of biopsy by examination of the obtained cores under a microscope by a pathologist. Such real time assessment enables the biopsy physician to obtain additional cores of tissue if existing cores are deemed to be inadequate. However, resource constraints have markedly limited the availability of adequacy evaluation within many biopsy suites. With the widespread adoption of smartphones, there is now an opportunity to evaluate biopsy tissue via macro images instead of a microscope. An additional step is to build image analysis capabilities into an application on the smartphone, thus enabling the biopsy physician to obtain reliable instantaneous feedback about whether the material they have collected is sufficient or whether the biopsy “missed”, and more tissue needs to be obtained. A recent study of 123,372 native kidney biopsies found that the miss rate in PRB increased markedly from 2% in 2005 to 14% in 2020, largely attributed to the increased involvement of radiologists performing the biopsies and their preference for smaller diameter biopsy needles. This increase in kidney biopsy miss rate significantly impacts patient care in the management of medical kidney disease and highlights the need for improved tools for assessing the adequacy of tissue collected before the patient leaves the biopsy suite. In Phase I, Arkana Laboratories will collaborate with the Kolachalama laboratory at Boston University using biopsy cores taken from unused deceased donor kidneys to develop a deep neural network that can accurately classify which portions of a biopsy core represent renal cortex vs non-cortex tissue, which may include renal medulla, perirenal fat, fibrous capsule, and non-renal tissue. In Phase II, we will extend these efforts using images taken from kidney biopsies received by Arkana to optimize the algorithmic assessment of adequacy and then use these results to develop a professional version of the smartphone application that will be available to all kidney biopsy practitioners using virtually any smartphone camera.