Explainable artificial intelligence (XAI): closing the gap between image analysis and navigation in complex invasive diagnostic procedures

Literature review

Cystoscopy is the gold standard for initial macroscopic assessments of the human urinary bladder to rule out (or diagnose) bladder cancer (BCa). Despite having guidelines, cystoscopic findings are diverse and often challenging to classify. The extent of the false negatives and false positives in cystoscopic diagnosis is currently unknown. We suspect that there is a certain degree of under-diagnosis (like the failure to detect malignant tumours) and over-diagnosis (e.g. sending the patient for unnecessary transurethral resection of bladder tumors with anesthesia) that put the patient at risk.

Conclusions

XAI robot-assisted cystoscopes would help to overcome the risks/flaws of conventional cystoscopy. Cystoscopy is considered a less life-threatening starting point for automation than open surgical procedures. Semi-autonomous cystoscopy requires standards and cystoscopy is a good procedure to establish a model that can then be exported/copied to other procedures of endoscopy and surgery. Standards also define the automation levels—an issue for medical product law. These cystoscopy skills do not give full autonomy to the machine, and represent a surgical parallel to ‘Autonomous Driving’ (where a standard requires a human supervisor to remain in the ‘vehicle’). Here in robotic cystoscopy, a human supervisor remains bedside in the ‘operating room’ as a ‘human‐in‐the‐loop’ in order to safeguard patients. The urologists will be able to delegate personal- and time-consuming cystoscopy to a specialised nurse. The result of automated diagnostic cystoscopy is a short video (with pre-processed photos from the video), which are then reviewed by the urologists at a more convenient time.

Future Platforms of Robotic Surgery

Among the various robotic devices that exist for urologic surgery, the most common are synergistic telemanipulator systems. Several have achieved clinical feasibility and have been licensed for use in humans: the standard da Vinci, Avatera, Hinotori, Revo-i, Senhance, Versius, and Surgenius. Handheld and hands-on synergistic systems are also clinically relevant for use in urologic surgeries, including minimally invasive and endoscopic approaches. Future trends of robotic innovation include an exploration of more robust haptic systems that offer kinesthetic and tactile feedback; miniaturization and microrobotics; enhanced visual feedback with greater magnification and higher fidelity detail; and autonomous robots.

A novel gaze-controlled flexible robotized endoscope; preliminary trial and report

Background

Interventional endoluminal therapy is rapidly advancing as a minimally invasive surgical technique. The expanding remit of endoscopic therapy necessitates precision control. Eye tracking is an emerging technology which allows intuitive control of devices. This was a feasibility study to establish if a novel eye gaze-controlled endoscopic system could be used to intuitively control an endoscope.

Methods

An eye gaze-control system consisting of eye tracking glasses, specialist cameras and a joystick was used to control a robotically driven endoscope allowing steering, advancement, withdrawal and retroflexion. Eight experienced and eight non-endoscopists used both the eye gaze system and a conventional endoscope to identify ten targets in two simulated environments: a sphere and an upper gastrointestinal (UGI) model.

Completion of tasks was timed. Subjective feedback was collected from each participant on task load (NASA Task Load Index) and acceptance of technology (Van der Laan scale).

Results

When using gaze-control endoscopy, non-endoscopists were significantly quicker when using gaze-control rather than conventional endoscopy (sphere task 3:54 ± 1:17 vs. 9:05 ± 5:40 min, p = 0.012, and UGI model task 1:59 ± 0:24 vs 3:45 ± 0:53 min, p < .001).

Non-endoscopists reported significantly higher NASA-TLX workload total scores using conventional endoscopy versus gaze-control (80.6 ± 11.3 vs 22.5 ± 13.8, p < .001). Endoscopists reported significantly higher total NASA-TLX workload scores using gaze control versus conventional endoscopy (54.2 ± 16 vs 26.9 ± 15.3, p = 0.012). All subjects reported that the gaze-control had positive ‘usefulness’ and ‘satisfaction’ score of 0.56 ± 0.83 and 1.43 ± 0.51 respectively.

Conclusions

The novel eye gaze-control system was significantly quicker to use and subjectively lower in workload when used by non-endoscopists. Further work is needed to see if this would translate into a shallower learning curve to proficiency versus conventional endoscopy. The eye gaze-control system appears feasible as an intuitive endoscope control system. Hybrid gaze and hand control may prove a beneficial technology to evolving endoscopic platforms.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00464-021-08556-1.

Accuracy of a Robotic Endoscopic System in Cadaver Models with Simulated Tumor Targets: ACCESS Study

BACKGROUND

Bronchoscopy for the diagnosis of peripheral pulmonary lesions continues to present clinical challenges, despite increasing experience using newer guided techniques. Robotic bronchoscopic platforms have been developed to potentially improve diagnostic yields. Previous studies in cadaver models have demonstrated increased reach into the lung periphery using robotic systems compared to similarly sized conventional bronchoscopes, although the clinical impact of additional reach is unclear.

OBJECTIVES

This study was performed to evaluate the performance of a robotic bronchoscopic system's ability to reach and access artificial tumor targets simulating peripheral nodules in human cadaveric lungs.

METHODS

Artificial tumor targets sized 10-30 mm in axial diameter were implanted into 8 human cadavers. CT scans were performed prior to procedures and all cadavers were intubated and mechanically ventilated. Electromagnetic navigation, radial probe endobronchial ultrasound, and fluoroscopy were used for all procedures. Robotic-assisted bronchoscopy was performed on each cadaver by an individual bronchoscopist to localize and biopsy peripheral lesions.

RESULTS

Sixty-seven nodules were evaluated in 8 cadavers. The mean nodule size was 20.4 mm. The overall diagnostic yield was 65/67 (97%) and there was no statistical difference in diagnostic yield for lesions <20 mm compared with lesions measuring 21-30 mm, the presence of a concentric or eccentric radial ultrasound image, or relative distance from the pleura.

CONCLUSIONS

The robotic bronchoscopic system was successful at biopsying 97% of peripheral pulmonary lesions 10-30 mm in size in human cadavers. These findings support further exploration of this technology in prospective clinical trials in live human subjects.