Development of bi-modal transvaginal probe and evaluation of its diagnostic accuracy for a non-invasive AI-based detection of preinvasive cervical lesion in comparison to standard colposcopy in screen positive patients.

Background: Cervical cancer is the fourth highest cause of mortality around the world. However, cervical cancer progression is slow, and takes about ten years before fully developing. The current techniques, such as cytology which gives very high accuracy, are costly and requires highly skilled workforce. The colposcopy used as triage to locate the high-grade lesion involves using acetic acid, which is corrosive, resulting in patients’ pain during the procedure and needs an expert pathologist to interpret the colposcopy guided biopsies. Recommended screening with HPV DNA is far-fetched due to the cost.

Scientific Rationale: Due to the reduced activity of ferrochelatase enzyme during the heme cycle, there is a buildup of the porphyrin and lower total haemoglobin within the cancer cells. The proposed transvaginal probe estimates the cancer biomarkers, such as protein, glycogen, PpIX, HbO2, and Hb, on the tissue surfaces to differentiate the cancerous tissues from adjacent normal. In order to do so, the proposed transvaginal imaging probe uses the autofluorescence and multispectral diffuse reflectance spectroscopy modalities. The probe detects the increase in porphyrin by shining the light at 405 nm and detecting the image from the tissue surface in about 540 nm.  The probe also houses LEDs operating at 545 nm, 575 nm to target the two isosbestic points for HbO2 and Hb, and 610 nm where the absorption coefficient of Hb is about ten times more than HbO2. The camera captures the images of the tissue surface by shining the light with 545 nm, 575 nm, and 610 nm, one after other. The ratio R610/R545, R610/R575, and autofluorescence differentiate the cancerous region from the adjacent normal tissues.

The user interface will show the unprocessed images and an overlay of multiple images marking the probable region of interest. The unprocessed image will be immediately available. The processed image with the overlay of different features and AI will take about 5 mins.

Novelty and Indigenization:

• The probe using the autofluorescence (405 nm) and diffuse reflectance imaging (545 nm, 575 nm, and 610 nm) has been performed for academic research.

• The novelty of this project is the application of bi-modal (autofluorescence and diffuse reflectance imaging) for diagnosis of preinvasive lesions in cervix.

• The label-free transvaginal imaging probe housing multiple narrow bandwidth LEDs and subminiaturized high-resolution CMOS camera to perform autofluorescence and diffused reflectance imaging is to be indigenously fabricated in the lab.

Expected Outcome:

The objective of the project is to design and develop an at-community digitally connected imaging probe which can be used to capture images of the cervix and share with the clinician in remote location using WiFi or mobile internet. The images can yield rapid assessment with the help of Artificial Intelligence (AI) in identifying preinvasive cancer.

We expect the probe to identify the high-grade lesion with diagnostic accuracy as good as standard colposcopy performed by the physician. With this expectation we can in future offer point of care screen and treat at community.