Search

BackgroundThe impact and utility of machine learning (ML)-based prediction tools for cancer outcomes including assistive diagnosis, risk stratification, and adjunctive decision-making have been largely described and realized in the high income and upper-middle-income countries. However, statistical projections have estimated higher cancer incidence and mortality risks in low and lower-middle-income countries (LLMICs). Therefore, this review aimed to evaluate the utilization, model construction methods, and degree of implementation of ML-based models for cancer outcomes in LLMICs.MethodsPubMed/Medline, Scopus, and Web of Science databases were searched and articles describing the use of ML-based models for cancer among local populations in LLMICs between 2002 and 2022 were included. A total of 140 articles from 22,516 citations that met the eligibility criteria were included in this study.ResultsML-based models from LLMICs were often based on traditional ML algorithms than deep or deep hybrid learning. We found that the construction of ML-based models was skewed to particular LLMICs such as India, Iran, Pakistan, and Egypt with a paucity of applications in sub-Saharan Africa. Moreover, models for breast, head and neck, and brain cancer outcomes were frequently explored. Many models were deemed suboptimal according to the Prediction model Risk of Bias Assessment tool (PROBAST) due to sample size constraints and technical flaws in ML modeling even though their performance accuracy ranged from 0.65 to 1.00. While the development and internal validation were described for all models included (n=137), only 4.4% (6/137) have been validated in independent cohorts and 0.7% (1/137) have been assessed for clinical impact and efficacy.ConclusionOverall, the application of ML for modeling cancer outcomes in LLMICs is increasing. However, model development is largely unsatisfactory. We recommend model retraining using larger sample sizes, intensified external validation practices, and increased impact assessment studies using randomized controlled trial designsSystematic review registrationhttps://www.crd.york.ac.uk/prospero/display_record.php?RecordID=308345, identifier CRD42022308345.

BackgroundComputer-assisted surgeries (CAS) are increasingly being adopted as the treatment of choice for jaw reconstructions with osseous free flaps. Although unexpected change of surgical plans remains a major concern of CAS, there are few studies focusing on this unfavorable clinical scenario. The aim of the present study was to investigate the rate of unexpected change of surgical plans and potential influential parameters, and to discuss the contingency strategies.MethodsA retrospective study was performed to evaluate all the patients who underwent computer-assisted jaw resections and osseous free flap reconstructions. The postoperative radiographs were reviewed and compared with the preoperative surgical plans. Operating records were examined to analyze the reasons for unexpected change of surgical plans and the management. The potential influential parameters for the change of surgical plans were analyzed using Fisher-exact test. The difference was regarded as statistically significant for a p-value less than 5%.ResultsFrom Nov 2014 to Oct 2021, a total of 98 consecutive computer-assisted free flap jaw reconstruction cases with osseous free flaps were included in this study. Our experience showed that 5.1% of the patients (five cases) needed intra-operative change of the surgical plans. We summarized the unexpected change of surgical plans and the contingency strategies as four clinical scenarios, including extended resection and reconstruction, shortened resection and reconstruction, modified resection without changing reconstruction, and modified reconstruction without changed resection. None of the potential influential parameters was identified as significant in relation to unexpected change of surgical plans intraoperatively.ConclusionOur experience shows that with the comprehensive methodology for computer-assisted free flap jaw reconstruction surgery planning, we can minimize the possibility of unexpected change of surgical plans during surgery. The lessons learned from our 98 consecutive cases can help beginners prevent unexpected change of surgical plans and rationalize contingency strategies in computer-assisted free flap jaw reconstruction.

ObjectivesAlthough computer-assisted surgery using fibula flap has been widely applied for oncologic jaw reconstruction in recent years, the inaccurate positioning of the fibula harvest guide brings sliding and rotational errors, which leads to compromised accuracy in simultaneous implant placement and dental rehabilitation. This study aimed to develop a novel three-dimensional (3D)-printed patient-specific fibula malleolus cap to increase oncologic reconstruction accuracy.MethodsIn this prospective comparative study with a recent historical control cohort, patients in need of oncologic jaw reconstruction with fibula free flaps were recruited. In the study group, the fibula was harvested with the guide of the malleolus cap, whereas in the control group, without the malleolus cap. Deviations of location and angulation of distal fibula osteotomies, jaw reconstruction segments, and simultaneous dental implants were compared.ResultsTwenty patients were recruited, with 10 in each arm. The application of the malleolus cap significantly reduced the deviations in locations and angles of distal fibula osteotomies, from 9.5 to 4.1 mm and 25.3° to 8.7°. For the simultaneous dental implants placed in the fibula flaps, there was a significant increase in the accuracy of implant platform locations (the average deviation from 3.2 to 1.3 mm), apex locations (from 3.8 to 1.5 mm), and angles (from 11.3° to 4.6°). No significant difference was detected in the accuracy of fibula reconstruction segments.ConclusionsWe developed a novel fibula malleolus cap to overcome the sliding and rotational errors during fibula flap harvesting for oncologic jaw reconstruction, with increased accuracy in simultaneous dental implants. This is a step forward to achieve a satisfactory functional outcome of jaw reconstruction with dental rehabilitation.

BackgroundComputer-assisted jaw reconstruction (CAJR) has benefits in reducing operation time and improving reconstruction accuracy, compared to conventional freehand jaw reconstruction. However, no information is available regarding learning curves in CAJR with the use of 3D-printed patient-specific surgical plates (PSSP). The purpose of this study was to assess surgical outcomes and learning curve for the first 58 consecutive CAJR using 3D-printed PSSP performed by a single surgical team in a single institution.MethodsIn a prospective study, consecutive patients who underwent free flap CAJR using 3D-printed PSSP were included. The determination of proficiency, based on the cumulative sum of surgical success (no major adjustment of 3D-printed PSSP, flap survival) passing the acceptable boundary line of cumulative sum analysis, was the primary outcome. To find out any potential factors influencing the learning curve, baseline characteristics of patients were compared before and after proficiency achievement. Secondary outcomes included inflexion points of the total operation time, blood loss, length of hospital stay, and bone graft deviation, measured by the cumulative sum analysis.ResultsFrom December 2016 to November 2020, 58 consecutive cases underwent surgery performed by a single surgical team. The overall surgical success rate was 94.8% (55/58). A three-stage learning curve of primary outcome was observed. The proficiency was achieved after 23 cases. The proportions of advanced tumor staging and concomitant surgery after obtaining proficiency were significantly higher than those before achieving proficiency (p = 0.046 and p < 0.001, respectively). Mean values of operation time, intraoperative blood loss, length of hospital stay, and bone graft deviation were 532.5 ± 119.2 min, 1,006.8 ± 547.2 ml, 16.1 ± 6.3 days, and 0.9 ± 1.2 mm, respectively. Two trends of learning curve were observed in the CUSUM analyses of total operation time, length of hospital stay, and bone graft deviation, in which the first and second inflexion points occurred between 8 and 17 cases and between 43 and 46 cases, respectively.ConclusionOur results revealed a three-stage learning curve of CAJR with the use of PSSP, including initial learning, plateau, and overlearning. Based on CUSUM analysis, the surgical proficiency was achieved after 23 cases, and total operation time, length of hospital stay, and bone graft deviation stabilized after 8–17 cases.

Background: The impact and utility of machine learning (ML)-based prediction tools for cancer outcomes including assistive diagnosis, risk stratification, and adjunctive decision-making have been largely described and realized in the high income and upper-middle-income countries. However, statistical projections have estimated higher cancer incidence and mortality risks in low and lower-middle-income countries (LLMICs). Therefore, this review aimed to evaluate the utilization, model construction methods, and degree of implementation of ML-based models for cancer outcomes in LLMICs. Methods: PubMed/Medline, Scopus, and Web of Science databases were searched and articles describing the use of ML-based models for cancer among local populations in LLMICs between 2002 and 2022 were included. A total of 140 articles from 22,516 citations that met the eligibility criteria were included in this study. Results: ML-based models from LLMICs were often based on traditional ML algorithms than deep or deep hybrid learning. We found that the construction of ML-based models was skewed to particular LLMICs such as India, Iran, Pakistan, and Egypt with a paucity of applications in sub-Saharan Africa. Moreover, models for breast, head and neck, and brain cancer outcomes were frequently explored. Many models were deemed suboptimal according to the Prediction model Risk of Bias Assessment tool (PROBAST) due to sample size constraints and technical flaws in ML modeling even though their performance accuracy ranged from 0.65 to 1.00. While the development and internal validation were described for all models included (n=137), only 4.4% (6/137) have been validated in independent cohorts and 0.7% (1/137) have been assessed for clinical impact and efficacy. Conclusion: Overall, the application of ML for modeling cancer outcomes in LLMICs is increasing. However, model development is largely unsatisfactory. We recommend model retraining using larger sample sizes, intensified external validation practices, and increased impact assessment studies using randomized controlled trial designs. [ABSTRACT FROM AUTHOR]