TECHNOLOGY GENERATORS

University of the Philippines Manila
Project Leader: Alvin B. Caballes, MD, MDE, MPP

THE PROBLEM

Hirschsprung’s disease (HD) is a congenital condition that leads to intestinal obstruction, severe infection, and other life-threatening complications among affected infants and children. The condition is characterized by the absence of ganglion cells that are responsible for regulating bowel motility, resulting in the non-passage of intestinal contents and subsequent bowel obstruction, which in turn can lead to infectious complications.

The gold-standard diagnostic method for HD involves confirming the absence of ganglion cells in tissue samples obtained from intestinal or rectal biopsies, as observed through hematoxylin and eosin staining on paraffin sections. Primary treatment for HD is surgical, involving pull-through procedures where abnormal bowel segments are removed. In many cases, addressing immediate obstruction relief involves surgically creating intestinal stomas, enabling the passage of stools through the abdominal wall. These necessary procedures come with inherent risks and are undertaken only when a confirmed diagnosis is in place. This highlights the need for a more accessible and accurate diagnostic system.

THE SOLUTION

The primary diagnostic approach for identifying HD involves analyzing histological samples from intestinal biopsies of suspected patients. In this DOST-PCHRD-funded project, an image interpretation software was designed to diagnose HD with a minimum accuracy of 90%. Utilizing various convolutional neural networks (CNNs), such as Faster R-CNN, TridentNet, and Facebook’s Detection Transformer (DETR), the software predicted the locations of ganglion cells in input images. A user-friendly browser-based interface was developed alongside the machine learning software for visualization and ease of use.

The study yielded a notably high accuracy level, averaging 99%, in the automated diagnosis of HD using digitally-converted images from hematoxylin and eosin-stained histological slides. This significant advancement supplements the gold standard procedure in diagnosing HD and can therefore aid pathologists in making more efficient and timely diagnosis, potentially reducing the need for invasive procedures.

TECHNOLOGY DEVELOPMENT STATUS

The Technology Readiness Level is presently at TRL 6, and the Investment Readiness Level stands at 4. The project is currently advancing in its Phase II of development.

CURRENT NEEDS

When ready for commercialization, the technology would need a developer or distributor specializing in systems and software to promote its adoption. The primary focus would be on marketing to healthcare facilities and hospitals equipped with histopathology laboratories, where pathologists can utilize the technology to enhance their diagnostic processes.

CONTACT DETAILS

University of the Philippines Manila-Technology Transfer and Business Development Office (UPM-TTBDO)
2/F UP Manila Main Building, Joaquin Gonzales Compound, Padre Faura Street, Ermita, Manila
(632) 8-8141-293
ttbdo.upm@up.edu.ph