Artificial Intelligence is becoming all pervasive in the diagnostic industry. From improving workflow inefficiencies to enabling early diagnosis and improving image quality to being used as a quality control tool, there is significant potential for it to create value in patient care. 

In healthcare, AI is in a state of infancy, currently covering a handful of clinical indications with limitations compared to the ground truth in some products. There are numerous start-ups and established players invested deeply in widening the scope of indications and accuracy of AI. India is a seeding ground for many such start-ups.  

But, the main challenge in developing AI which is clinically impactful is the learning data as there are limited databases available in India that can be effectively used by these companies. There are also challenges in getting Radiologists to accurately label and teach the algorithm of the areas of abnormality, validate algorithms and integrate these into clinical used case scenarios.

Explaining a recent clinical trial at Max super specialty hospital, Saket with GE, Dr Bharat Aggarwal, Director - Radiology Services, Max Super Speciality Hospital, Saket said, “An AI-enabled bedside portable X-ray machine was used in the Emergency Room and ICU settings to assess the impact providing quaternary care to patients with diverse clinical indications. It had an embedded AI algorithm alerting the radiographer if the patient had a suspected pneumothorax.” 

During the trial, it was found that the pneumothorax algorithm was to expedite the detection of this condition seconds after taking the X-ray at the bedside. He informed, “The radiographer immediately informed the Radiologist when this alert was raised, who would validate the diagnosis, make the report and inform the intensivists managing the patients.” This was significantly faster than what has been the traditional workflow wherein cases come up for reporting to the radiologists as a part of a larger set and in otherwise clinically unsuspected cases or cases where the pneumothorax is small and early, the diagnosis can be delayed. 

Dr Aggarwal elaborates on Pneumothorax. It can be a life-threatening clinical emergency where air from the lungs escapes into the pleural cavity with a potential of collapsing the lung and resulting in acute respiratory failure. He further added, “The machine had embedded AI algorithms helping the radiographers improve the quality of X-rays acquired by ensuring that they were alerted when the entire lung was not included in the radiographs.” 

Shedding light on the hospital’s initiative, Dr Sujeet Jha, Senior Director, Endocrinology, Diabetes & Obesity, Max Super Speciality Hospital, Saket said, “Max Healthcare has been an early mover in the AI space by collaborating with the AI companies through research projects, validation studies and mentoring the companies to direct products towards solving real clinical problems. MHC has always invested in cutting edge technology and is currently evaluating ways to incubate digital healthcare companies to solve problems in patient care, improve efficiencies and reduce escalating costs of delivery.” 

In the near future, AI will be seen as an accurate tool for radiologists in pre-reading imaging studies as well as a standard quality control tool to pick up diagnostic errors in clinical practice. There are varying rates of medical errors reported in the literature, attributed to various causes including different levels of radiologist training, increasing volumes of imaging & caseloads, fatigue and the “human error” factor. 

With evolving AI algorithms, the pre-reading by AI will compensate for some of these factors by alerting the radiologists to evaluate findings which may be the raise missed. Speeding up critical alerts, evident in our study, will impact greater clinical situations helping timely treatment.