This story was produced in collaboration with PBS "Frontline" and NPR.

Listen to NPR's Morning Edition and All Things Considered (check local listings). And watch "The Child Cases" on PBS "Frontline" tonight (check local listings).

Her name was Isis Charm Vas and at 6 months old she was a slight child -- fifth percentile in height and weight.

When the ambulance sped her to Northwest Texas Hospital on a Saturday morning in October 2000, doctors and nurses feared that someone had done something awful to her delicate little body.

A constellation of bruises stretched across her pale skin. CT scans showed blood pooling on her brain and swelling. Her vagina was bleeding, as well. The damage was so severe that her body's vital organs were shutting down.

Less than 24 hours later, Isis died.

An autopsy bolstered the initial suspicions that she'd been abused. Dr. Joni McClain, a forensic pathologist, ruled Isis' death a homicide and said the baby had been sexually violated. McClain would later describe it as a "classic" case of blunt force trauma, the type of damage often done by a beating.

The police investigation that followed was constructed almost entirely from medical evidence. In the end, prosecutors indicted one of the child's babysitters: Ernie Lopez.

Today, Lopez is serving a 60-year prison term for sexual assault and is still facing capital murder charges.

But in the years since Lopez was sent to the penitentiary, a growing body of evidence has emerged suggesting that McClain and the hospital staffers were wrong about what happened to Isis -- and that her death was not the result of a criminal attack.

If Lopez is ultimately exonerated, his case will not be unique. An investigation by ProPublica, PBS "Frontline" and NPR has found that medical examiners and coroners have repeatedly mishandled cases of infant and child deaths, helping to put innocent people behind bars.

We analyzed nearly two dozen cases in the United States and Canada in which people have been accused of killing children based on flawed or biased work by forensic pathologists, and then later cleared.

Some spent years in prison before courts overturned their convictions. In 2004, San Diego prosecutors moved to dismiss charges against a man who'd been imprisoned for two decades for murdering his girlfriend's son.

Others were freed more swiftly but endured hardships nonetheless. An El Paso, Texas, jury acquitted a woman of killing her child in 2010, but after spending 22 months in the county jail, she still had to wage a legal battle to regain custody of her other children.

The questionable prosecutions identified in our investigation had common elements:

Often, authorities had little to go on other than autopsy findings. Many of the doctors who conducted post-mortem examinations failed to consult specialists in childhood injuries or ailments, or to thoroughly review medical records that could have affected their conclusions. In several cases, forensic pathologists worked so closely with authorities, they effectively became agents of law enforcement, rather than objective arbiters of scientific evidence.

Some experts in the field say worries about mistakes in child death cases are overstated. "The vast majority of forensic pathologists recognize a child abuse case when they see it, and it's not because they want to persecute people," said Dr. Mary Case, chief medical examiner for four Missouri counties including St. Louis County.

But others say the criminal justice system has yet to confront the full scope of the problem, and that, as a result, more innocent people may be serving time for crimes they didn't commit. "I think it's time to look at these cases again," said Dr. Michael Laposata, chief pathologist at Vanderbilt University Medical Center, adding that this could "result in the liberation of a number of falsely accused people."

Lopez, 40, a soft-spoken man with a slight twang, still can't quite believe he may spend the rest of his life locked up for something he says he didn't do: harming the infant he nicknamed "Little Bird."

"Sometimes I wake up and I look at my cell and man, it just hits me: You know, I'm in prison," he said in an interview. "I never thought I would be in prison, never in a hundred years."

* * *

At 10:55 a.m. on Oct. 28, 2000, Ernie Lopez grabbed the cordless phone at his house and punched in the numbers 9-1-1.

"What's going on? What's going on?" asked the operator.

"OK, my ... We're babysitting this little baby girl for Dr. Vas," said Lopez, according to a recording and transcript of the call. A spider, he explained, had bitten Isis a week earlier, "and she's been acting funny ever since."

Lopez and his wife, DeAnn, regularly babysat Isis and her two older siblings, both toddlers. The children's mother, Veronica Vas, was a physician at a nearby hospital, and on that morning she was on her way to Detroit for the weekend.

Lopez, a burly, gregarious man who worked as a mechanic, was looking after the children while DeAnn went shopping for a dress for the annual Lopez family Christmas photo, scheduled to be taken that afternoon. He had been watching the Vas children for 40 minutes when he called for an ambulance.

"Is she breathing right now?" asked the operator.

"No, she's not breathing on her own. I was fixing to put her in a bath and she stopped breathing and I have been trying to get her CPR for the last two to three minutes," Lopez replied, stress permeating his voice.

On the phone, Lopez described his efforts to revive Isis. "I tried to slap her on the bottom and slap her on the face and she won't wake up. She won't do nothing." Blood spilled from her mouth. "She was bit about 14 times. ... She's got all these bruises around her neck and on her face where she was bitten." After the ambulance arrived at his modest one-story home, Lopez rode with Isis to the emergency room.

Police detectives, alerted by hospital staffers, quickly showed up at the hospital to question Lopez. He wept as he spoke to the officers.

They weren't the only people asking questions. One of Lopez's brothers pulled him into a hospital bathroom and confronted him. Had he attacked the infant? "I said, 'No, I didn't do this,'" recalled Lopez. "Why would I do it?"

By the time Isis died a day later, police had arrested Lopez.

The body of the baby, wrapped in a colorful blanket, was transported to Dallas, where McClain performed the autopsy. To the doctor, the evidence pointed to sexual assault and murder.

"It is my opinion that Isis Charm Vas, a 6-month-old white female, died as the result of multiple blunt force injuries," McClain wrote in the autopsy report. (McClain declined to comment for this story.)

For police, solving the case was an exercise in elimination. Lopez was the only adult present when Isis collapsed. That made him the sole suspect. Who else could have done it?

In October 2001, a grand jury indicted Lopez on charges of capital murder and sexually assaulting a young child.

* * *

Forensic pathologists like McClain play a critical role at the intersection of medicine and law enforcement. Employed by medical examiners and coroners' offices, they are called in to figure out how people have died. They scrutinize corpses, searching for clues. Was the person murdered? Was it suicide? An accidental overdose? A heart attack?

Their findings carry enormous weight within the criminal justice system. As anyone who's watched an episode of "CSI" knows, if a forensic pathologist says it's a homicide, police will soon be hunting for the killer.

Though depicted as glamorous and high-tech on TV, the field of death investigation is plagued by chronic underfunding, a shortage of specialists, and a lack of national standards, according to a 2009 report by the National Academy of Sciences.

Many of the nation's morgues are staffed by doctors who aren't board-certified in forensic pathology. To become certified, doctors need an extra year of training and must pass a day-long test. Earlier this year, an investigation by ProPublica, "Frontline" and NPR showed that more than 100 physicians without board certification were working at the country's busiest coroner and medical examiner offices.

Even for the best educated and trained doctors, performing an autopsy on a baby or young child poses particular technical challenges. Their developing bodies function differently. It's why doctors who treat living children -- pediatricians -- receive different training than those who deal with adults.

"Adults are generally tougher and harder to kill then a small child. Particularly an infant," said Dr. Jon Thogmartin, chief medical examiner for Pasco and Pinellas counties in Florida, a jurisdiction that includes St. Petersburg. "So, you're looking for very subtle signs of trauma or pressure, or small amounts of bleeding that could potentially cause a kid severe illness or death."

When toddlers and infants die, autopsies frequently play a primary role in the police investigation. Adults often kill one another in public places where witnesses might catch glimpses of the violence. They tend to use guns or knives, weapons that leave obvious and distinct wounds. When adults kill children, they are more likely to use their hands and to commit their crimes out of view of anyone else.

"Often there are only two pieces of evidence," said Justice Stephen Goudge, a Canadian judge who conducted an extensive inquiry into Ontario's forensic pathology system. "The first: who had care of the infant in the hours leading up to the death, normally a parent or caretaker. And secondly, the forensic pathology, which attempts to give an opinion on what the cause of death was." If the autopsy findings are flawed, the judge said, "then the risk of a miscarriage of justice is high."

Thogmartin said the charged emotions inevitably triggered by a child's death add another layer of complexity. Forensic pathologists, in his view, can get "caught up in the anger, the emotion, the despair." Their mindset can become prosecutorial, Thogmartin said, until every child death is a "homicide until proven otherwise." When he took on his current job as chief medical examiner in 2000, he stressed the need for neutrality to his staff.

"As a forensic pathologist, I don't testify for the state. I don't testify for the defense. I testify for the decedent," he said. "They are not able to talk, so I try to talk for them."

Thogmartin overruled the autopsy conclusions in two child death cases handled by his predecessors that he said might have been colored by bias. In one case, a man was four years into a 10-year prison term for killing his infant son. In the other, a father was facing trial on murder charges for killing his 7-month-old daughter.

When Thogmartin sifted through the autopsy files and tissue samples, he was shocked: He saw no evidence of violence. In his opinion, the children had died of natural causes.

Both men were subsequently cleared by the courts, but even the one exonerated before standing trial suffered life changing consequences, Thogmartin said. "That unfortunate gentleman had his life turned upside down. ... His life was destroyed."

* * *

Ernie Lopez was born in Amarillo, a dust-swept, blue-collar city in the northern reaches of Texas and spent most of his life there.

His father, Ernest Sr., toiled as a diesel mechanic at a Caterpillar dealership. Lopez, too, was fascinated by motors. At 13 or 14, he replaced a wrecked engine cylinder on a Kawasaki dirt bike all by himself. He moved on to American muscle cars, spending weekends screaming across the asphalt at the drag strip on the edge of town and weekday evenings tuning his Ford Mustang.

His mother, Rosa, operated a daycare center for neighborhood children out of the family home. Growing up, Lopez said, "we had kids all the time in the house."

By the time he turned 30, Lopez had three children of his own, two with DeAnn and one from an earlier relationship.

Lopez and DeAnn lived across the street from the house he'd been raised in, where his mom and dad still lived. His brother Eddie lived next door to their parents. His brother Sabian lived a few minutes down the road. The whole tribe often converged at Rosa and Ernest Sr.'s home for barbeques and birthdays and holidays, the grandchildren scampering up the big willow tree out front.

Lopez "was a good dad, a very good dad, a very good uncle to my kids and to Sabian's kids," said Eddie, a heavily muscled truck driver.

Lopez worked at Hand Industrial, a company that manufactures and repairs heavy factory equipment. "At work, Ernie stood out as a very gentle person," said Becky Hand, the firm's accountant and office manager, in a court affidavit. "He would joke with the other male employees, but he was softer and kinder."

In the days before Isis died, Hand said, Lopez had asked her for advice because the baby "hadn't been eating and was lethargic," and he was worried that she might be seriously ill. Lopez was also alarmed by the marks on her face. "He said they started above one eyebrow and were almost in a pattern. ... Ernie said the bumps were strange and weren't like anything he'd seen before," Hand stated.

After Isis' death, the child protection system swung into action, tapping psychologist Edwin Basham to determine if Lopez should be separated from his own children while awaiting trial. Basham figures he has done around 4,000 such evaluations, including some on people who've admitted to killing children. Child abusers, in his experience, "have difficulty coping with relationships, with stress. They lose their temper. They blow up."

In Lopez, he saw none of the normal warning signs -- Lopez had no previous criminal record, no history of domestic dysfunction, no issues with drugs or alcohol. "He seemed to be a very concerned, family-focused kind of person," recalled Basham, who wrote in his 2001 report that he could find "no signs of serious psychological problems."

Lopez was confident he'd be cleared by the courts because he had done nothing wrong, Basham said. But after interviewing Lopez, the psychologist had an uneasy feeling. "He was caught up in this legal system that was determined to convict somebody," Basham said. "They had a dead baby. Somebody was going to get convicted of it. And he was nominated."

* * *

The trial of Ernie Lopez began in April 2003.

Potter County prosecutors decided to try him only on the sexual assault charge; the capital murder charge was left pending, allowing prosecutors to try him for that offense at any time.

There were no witnesses to the alleged attack, and Lopez had not confessed, so the prosecution's case relied heavily on medical testimony. Over five days, a stream of doctors and nurses who had treated Isis at the hospital told the jury she must have been brutalized.

Dr. Eric Levy, who treated Isis in the hospital's pediatric intensive care unit, said the child's symptoms indicated she had been the victim of a violent attack. Looking at a photo of the baby's lower half, Levy pointed out bruise after bruise.

Michelle Gorday, a veteran nurse who specialized in sexual assault examinations, said it was one of the worst cases she'd witnessed in her 20-year career. "I've never ... ever seen that kind of trauma," she testified.

The defense called no expert witnesses. Basham, the psychologist, was surprised Lopez's attorneys never asked him to testify. "I would have said that there wasn't a basis to suggest that he would be someone likely to have harmed a child," Basham said.

Lopez chose to take the stand, insisting he had never hurt Isis and testifying about the strange ailments that shadowed the last days of her life.

With each day, more health issues cropped up, he said. Blood spots speckled Isis' left eye. Congestion made it hard for her to breathe, prompting the Lopezes to treat the baby with a nebulizer. When Lopez changed her soiled diapers, her fecal matter, he testified, was "black" and "really thick and sticky."

DeAnn Lopez corroborated her husband's testimony. When Isis' mother brought the baby to the Lopez home on Oct. 25, the child had bumps on her head and bruises on her chest, DeAnn told the jury. The infant was "lethargic" and reluctant to drink from her bottle, DeAnn said, consuming about six ounces of liquid over the span of several days, far less than a healthy 6-month-old would have.

Veronica Vas, Isis' mother, disputed the Lopezes' account, maintaining that Isis was only mildly ill before she died. "She had about six little bumps on the left side of her forehead, but those were already healing up," Vas testified. The baby's energy level was "quite normal."

Addressing the jury, Assistant District Attorney J. Patrick Murphy summed up the case by saying, "Common sense tells you who had to have done it. ... This child could not fight back. This child could not consent. This child could do nothing but lay there." The jury found Lopez guilty.

McClain, the medical examiner, testified in the sentencing phase of the trial, telling the jury she had ruled Isis' death a homicide and detailing what she discovered during the autopsy. The baby, she said, suffered a "laceration of the vagina area" and injuries to her brain.

"The brain is covered by a thick fibrous covering called the dura," the forensic pathologist testified. "There was a hemorrhage beneath this dura, on top of the brain." Such bleeding, explained the doctor, can occur when a baby's brain is buffeted by a powerful impact.

"In this case," McClain continued, "we know the head has struck something, because we've got bruising in that area."

Scrutinizing Isis' eye tissue under a microscope, McClain said, she had discovered more bleeding, which she interpreted as another indicator of violent head trauma. Seven other doctors in her office had reviewed the case and concurred with her findings, McClain added.

Lopez was sentenced to 60 years.

Stunned, he turned to his brother Eddie. "It's like we're at my funeral," he said, "but I'm still alive."

* * *

Heather Kirkwood was an unlikely candidate to take on Lopez's case. She had spent the bulk of her career litigating anti-trust cases for the Federal Trade Commission, and she lives 1,700 miles away in Seattle.

After learning about Lopez from a relative living in Texas, Kirkwood agreed to represent him. For her, Isis' death presented a fascinating jigsaw puzzle to solve. Lopez struck her as "a nice young man" and the "circumstances of the case seemed weird as hell."

"My gut sense kept telling me this was a sick baby who was neglected," she said.

After taking on Lopez's case, Kirkwood started contacting physicians in hopes of getting them to analyze Isis' medical history. She sent a stack of documents to Dr. Richard Soderstrom, an emeritus professor of gynecology at the University of Washington. As an adviser to the Food and Drug Administration, Soderstrom served on a panel that studied the accuracy and safety of the colposcope, a device that can be used to take photos of injuries in sexual assault exams.

Isis Vas had been examined using a colposcope. But as Soderstrom stared at the photos taken of her, he wasn't convinced that she'd been violated. "I couldn't see anything that would say, beyond a reasonable doubt, that there was penetration," he said.

Soderstrom gave a sworn affidavit stating that, in his opinion, the photos did not suggest there had been sexual abuse. No semen or pubic hair had been found on Isis' body. Further, Soderstrom said, the lack of injuries to the child's inner thighs, labia major, and hymen, was "inconsistent with abuse."

Kirkwood also approached Dr. Michael Laposata, the chief pathologist for Vanderbilt University Medical Center in Nashville and a leading expert on blood disorders. At Laposata's sprawling lab, white-coated technologists run some 6 million tests per year, feeding a never-ending line of blood samples into an array of machines.

Looking at the file on Isis, Laposata quickly homed in on the tests run on her blood while she was in the hospital. To gauge how the blood is clotting, physicians typically begin with a pair of basic tests called the PT and PTT. In Isis, the "PT and PTT were markedly abnormal," Laposata said, adding that other tests also suggested a coagulation disorder. Where McClain had seen a "classic" case of blunt force trauma, Laposata saw something entirely different, a "classic picture" of Disseminated Intravascular Coagulation, a potentially lethal condition that can cause bleeding from sufferers' every orifice.

Based on the baby's "dark, tarry stools," elevated white blood cell count, and abnormal liver function tests, Laposata concluded, "something had to be going on for days" -- long before the 40 minutes Lopez was alone with the baby.

An infection could have led to DIC, and, eventually, to a fatal collapse, Laposata said. DIC could also explain Isis' bruises and the bumps on her head that Lopez and others believed were spider bites, he added.

"The reality is when your blood is so thin, when you're so unable to make a clot, you can just develop bruises and they can be spontaneous," he said.

In an interview, Laposata pulled up a PowerPoint presentation he uses to teach students how hard it is to distinguish child abuse from blood-clotting afflictions. One slide featured photos of two small children. Both of their faces were splotched with bruises. One had been battered. The other had idiopathic thrombocytopenic purpura, a condition that causes the blood to quit making platelets.

Without a host of lab tests, Laposata said, it would be impossible to figure out which little boy needed medical help and which one needed child protective services.

* * *

There is a growing awareness among medical practitioners of "mimics": ailments that can cause the kind of bruising and bleeding once assumed to be telltale indicators of child abuse. A 2006 textbook on head injuries in children listed literally dozens of afflictions -- including some fairly common illnesses -- that can produce hemorrhaging in the brain.

This is just one way that the science of how children die has evolved in recent years. The most notable -- and controversial -- example of this is the intense debate over "shaken baby syndrome," which has played out in scientific journals and mainstream outlets such as the New York Times Magazine.

Based on studies dating back to the 1960s, many forensic pathologists -- as well as other physicians -- came to believe that a signature trio of symptoms provided definitive proof that someone had violently shaken a child. Under the theory, certain patterns of bleeding and swelling of the brain, and hemorrhages of the retinas came to be seen as conclusive evidence that a child had been assaulted with terrible force, even if there were no other signs of trauma.

But many experts now view the diagnosis with increasing skepticism. In Canada and Britain, large-scale official reviews have uncovered at least nine cases in which people may have been wrongly convicted based on the shaken-baby theory.

Dr. Case, the Missouri medical examiner, said the controversy is a "sideshow": Typically, children who've been shaken have also suffered other serious injuries from being battered. "Yes, there is a scientific debate," she said. "I personally believe that you can shake a child and kill it."

The thinking of other doctors has undergone a radical change. Dr. Patrick Barnes, a pediatric radiologist at Stanford University, was a key prosecution witness in what is arguably the most famous shaken-baby case of all, the trial of Louise Woodward. Woodward was a 19-year-old nanny charged in 1997 with shaking an 8-month-old baby to death, hitting his head and causing fatal bleeding. With Barnes' help, the jury found Woodward guilty of second-degree murder. (She was ultimately released after serving less than a year in prison, when a judge reduced her charge to manslaughter.)

Barnes said he wouldn't give the same testimony today. There's been a "revolution" in the understanding of head injuries in the past decade, in part due to advances in MRI brain scanning technology, he said. "We started realizing there were a number of medical conditions that can affect a baby's brain and look like the findings that we used to attribute to shaken baby syndrome or child abuse," Barnes said.

The case of Melonie Ware shows how profoundly a closer reading of medical evidence can affect the outcome of a child death investigation.

Ware was convicted in 2004 of murdering a 9-month-old boy she was babysitting, based in large part on the testimony of a local forensic pathologist, Dr. Gerald Gowitt. Gowitt said someone shook the child violently, damaging his brain, and slammed his head, causing three near-identical bruises beneath the scalp.

Ware was sentenced to a life term in a Georgia prison.

After an extensive legal battle, a judge granted Ware a new trial in 2009. This time, her attorneys produced evidence from the baby's medical records overlooked in her first trial: Hospital staffers had tried multiple times to insert a probe into the child's skull, as part of their attempt to save his life. The bruises under the baby's scalp, experts for Ware testified, were likely caused by those failed attempts.

Two prominent physicians testified that shaking had nothing to do with the boy's demise. The child had died from sickle cell anemia, said the doctors, both specialists in the disease, which is known to cause cerebral bleeding.

Gowitt declined repeated requests for comment on the case.

The jury acquitted Ware, but her life is not the same. Her husband spent more than $700,000 on her defense, selling off and mortgaging real estate acquired over decades.

"We had to move in with my parents," said Ware, 38. "It's just messed us up totally."

After she was freed from prison, Ware, who had worked as a day-care provider, was rejected for job after job. "I even tried McDonald's," she recalled. She thinks potential employers were frightened off by her time in jail.

The stain of the case lingers. To this day, Ware's mug shot appears on the Georgia Department of Corrections' website, which lists her as still incarcerated.

* * *

In prosecuting Ernie Lopez, law enforcement officials focused almost exclusively on Isis Vas' final hours.

Lopez's legal team looked back further, however, marshaling evidence suggesting that the baby's deteriorating condition might have been overlooked by her mother.

Veronica Vas had moved to Amarillo in 1995 to do her residency at a branch of Texas Tech University. She began dating a doctor, with whom she had two children. Then, in a subsequent relationship, Vas, 32, became pregnant with Isis. By all accounts, Isis' father wasn't involved in her life.

Court records from a custody dispute between Vas and the father of her older children, as well as statements submitted as part of the Lopez case, depict the Vas household as chaotic in the period surrounding Isis' birth.

Dena Ammons, a nurse who worked closely with Vas during her residency, said Vas changed during her pregnancy with Isis. She began showing up late for work, her hair matted and uncombed. In a sworn statement, Ammons said that Vas drank and smoked throughout the pregnancy.

Lorrie Word worked for Vas as a live-in nanny from August 1999 until the summer of 2000, caring for Isis from the time she was born. Word said in an affidavit that, on one occasion, she returned from her night off to find Isis alone in a darkened house, crying and soaked with urine. Vas would later say she only left the child for 10 minutes. Soon after the incident, Word quit her job.

A family therapist who visited the Vas home in 2000 as part of the custody dispute described it as "extremely cluttered." "It was difficult to walk across the floor because of blankets, clothes, and toys," she wrote in a report submitted to a family court. "The home appeared extraordinarily disorganized."

Vas declined repeated requests for comment for this story. She has moved to Michigan, where the state medical board recently suspended her medical license due to alcohol abuse.

During Lopez's trial, Vas testified that in the months after Word quit she came to depend on the Lopez family to help care for her children.

According to Ernie Lopez, the day before Isis went into cardiac arrest he became so worried about the baby's health that he asked Vas for a note authorizing him or his wife to take the child to the doctor.

Vas didn't give him the note before leaving town for the weekend, he recalled in an interview. "Isis will be fine," Lopez said Vas told him.

* * *

By 2009, the new medical evidence gathered by Heather Kirkwood had captured the attention of the courts. After she filed an appeal, a habeas corpus petition, a judge granted Lopez a new evidentiary hearing. It represented a step toward possibly overturning his conviction.

The hearing lasted nearly twice as long as the original trial. This time, seven doctors testified -- for free -- on Lopez's behalf.

Kirkwood also had the chance to question Joni McClain, the forensic pathologist who ruled Isis Vas' death a homicide. McClain stood by her conclusion that Isis was killed by violence, not disease.

But she acknowledged that she'd paid little attention to Isis' blood-clotting tests and had only a vague understanding of their possible significance. "Did you look at these lab tests before reaching your conclusions?" Kirkwood asked. "I don't think I did beforehand because it was such a clear case of blunt force injury," McClain replied.

Kirkwood read through the results of five tests, starting with the PT and PTT, which measure blood coagulation in seconds.

McClain admitted the tests went beyond her expertise as they can only be run on the living. "I don't get into a PT, PTT. It's a useless test after someone's dead," the doctor said.

Four other doctors testified for the state, saying Isis had died from blunt-force injuries, not a bleeding disorder. "This is a pattern of injury that we see with trauma," said Randell Alexander, a pediatrician who heads the child abuse division at the University of Florida's College of Medicine, in Jacksonville. "This is not a bleeding death."

McClain also argued that it was possible for head injuries to cause the type of clotting problems Isis had suffered.

In an interview, Laposata agreed head trauma can have that effect but said Isis' lab results were too abnormal to have resulted from an attack that allegedly occurred about an hour before her hospitalization.

It would be nearly a year before Potter County Judge Dick Alcala issued his opinion on the case. In August 2010, Alcala made a recommendation to the state's highest criminal court that Lopez's conviction should be overturned. He found that Lopez's original attorneys had failed to "fully investigate the medical issues of whether a sexual assault had occurred" and "the cause of death of the child." If they had investigated properly, Alcala wrote, the jury might not have convicted Lopez.

The judge rejected Lopez's claim of innocence, which would have required a conclusion that "no reasonable juror would have convicted him" -- a high legal standard.

The case is now in the hands of the Texas Court of Criminal Appeals. It has the power to throw out Lopez's conviction and free him.

Potter County District Attorney Randall Sims continues to fight Lopez's appeal. In an interview, Sims said he could not discuss the case in detail because it is still ongoing. (He also said he had discouraged state witnesses, including the medical examiner, from speaking with us.) Sims said he thought Lopez had received a fair trial.

"The jury found him guilty," he said. "And we're defending that conviction."

There is no timetable for the appeals court's decision. Even if it overturns Lopez's conviction, he could remain tangled up in the criminal justice system for years. Sims could refile charges and try him a second time.

* * *

The United States is not the only country in which forensic pathologists have had difficulty investigating child deaths. Canada was rocked by a scandal that affected at least 20 criminal cases, sending officials there on a search for systemic solutions to prevent future miscarriages of justice.

The controversy centered on the work of Dr. Charles Smith, once one of Canada's leading forensic pathologists. Based at a children's hospital in Toronto, Smith specialized in performing autopsies in grisly child deaths and, over a span of 24 years, he testified regularly for prosecutors.

But by 2005, the province's chief coroner had become openly skeptical of Smith's findings and assigned five other forensic pathologists to conduct a top-to-bottom review of his work in 45 child death cases.

The results of the study were devastating: In 20 of the cases, the reviewers disagreed with Smith's autopsy reports or court testimony. Over and over, Smith cited evidence of murder where there was none, they found. (Smith would not respond to our questions.)

In a dozen cases, people were wrongly accused of killing children in Ontario based on Smith's work or testimony. Tammy Marquardt, who was sentenced to life in prison for murdering her son, spent 14 years behind bars before being exonerated.

In prison, she said, the other inmates despised her. "A baby killer would basically get the living daylights beaten out of them," she said. "A baby killer is classified as one of the lowest on the totem pole." The courts reversed her conviction earlier this year.

The Ontario chief coroner's internal review led to an official inquiry by Justice Goudge, who set out nearly 170 recommendations for remaking the province's broken death-investigation system.

Forensic pathologists who conducted child autopsies should be formally trained and board-certified, Goudge said. They should read all relevant medical records. While forensic pathologists often toil in a certain amount of isolation, Goudge recommended a more collaborative approach, saying they should consult with specialists in other medical disciplines and have other doctors review their autopsy findings.

Bias was a major concern for Goudge. In Ontario there was a mantra among forensic pathologists, he said in an interview: "Think Dirty." When doctors dealt with cases involving children who had died unexpectedly, they assumed parents or caregivers had murdered them. That outlook, the justice said, skewed the conclusions they reached in the autopsy suite. "The scientist's objective is to 'think truth' not 'think dirty,'" he said.

Many of Goudge's suggestions are being implemented in Ontario. But policy-makers in the United States have largely ignored them. There are no national standards or regulations regarding forensic pathology and practices vary widely from place to place.

Barnes, the Stanford pediatric radiologist, said it was imperative for the U.S. system to absorb the lessons from Ontario. "We need to establish the new standards at all levels, just like what is happening in Canada," he said.

* * *

After Lopez was bused off to prison, his mother would look out her kitchen window and stare across the street at his former home, her mind turning back to the day everything changed. "It was too much to bear," Rosa remembered. So she and Ernest Sr. sold their house and moved to a place on the outskirts of Amarillo. Eddie and his wife did the same.

Lopez now lives in a cell in the Connally Unit, a maximum-security lock-up about 600 miles away in the scrubby countryside south of San Antonio. Every three or four months his parents make the 10-hour drive to the facility, a journey that costs about $1,000 between gas and hotels. "When we go, we have an enthusiasm, 'We're going to go see him,' you know, 'We're going to touch him and hold him,'" Rosa said. "And then on our way back it's really emotionally hard because we have to leave him there."

Lopez's imprisonment gnaws at Eddie, who weeps repeatedly when talking about his brother, tears streaking across his broad face. "Him being away this whole time, it's like a part of me is dead, because we were that close," Eddie said.

The two men are separated in age by 11 months, but these days Ernie looks years older. His close-cropped hair, once brown, has turned the color of iron.

Behind the prison's thick cinderblock walls, Lopez struggles to hold onto what's left of his old life. Every Tuesday evening he calls his children collect, offering fatherly guidance despite the circumstances. He communicates less frequently with DeAnn, who divorced him and remarried after he was sent away. (DeAnn has participated in his appeal, giving a sworn statement that corroborated his account of the day Isis died. She declined to be interviewed for this story.)

Because of the seriousness of his offense, Lopez is barred from working a prison job. He fills his mental space with the written word, reading more than 50 books last year alone. He's a big fan of the Western pulp novelist Louis L'Amour and an avid student of the Bible, which he's been through eight times.

Many men discover religion while incarcerated, but Lopez was devout long before he found himself in handcuffs. The night before he allegedly attacked Isis, he was at his church, acting in a play about good and evil.

Today, nearly 11 years after Isis died, Lopez continues to maintain -- emphatically -- that he never harmed her.

"Why should they believe that I'm innocent?" he asked during a two-hour interview. "Well, because that's not my character. That's not who I am."

Thinking back to that Saturday, Lopez paused and went silent, anguish filling his face. He exhaled heavily. "Her heart was beating a hundred miles an hour and she wasn't breathing. I put my ear to her chest, and I heard her heart just beating, just racing and ..." His head tilted downward and he stared at the floor. "She was there and then she wasn't there."

Lopez's voice grew quiet as the words trickled out slowly. "So many times," he said, "I think about what I could have done different to help her more."

Additional reporting contributed by Catherine Upin of PBS "Frontline." Lisa Schwartz, Sergio Hernandez and Liz Day contributed research to this story.

Watch the full episode. See more FRONTLINE.